diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,330175 @@
+{
+  "best_metric": 0.8849976858484121,
+  "best_model_checkpoint": "daedra/checkpoint-55000",
+  "epoch": 2.7706412775175053,
+  "eval_steps": 5000,
+  "global_step": 55000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9999664164693636e-05,
+      "loss": 1.9312,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999932832938727e-05,
+      "loss": 1.866,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9998992494080906e-05,
+      "loss": 1.7934,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9998656658774537e-05,
+      "loss": 1.6948,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9998320823468175e-05,
+      "loss": 1.5604,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9997984988161806e-05,
+      "loss": 1.5124,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999764915285544e-05,
+      "loss": 1.3331,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9997313317549075e-05,
+      "loss": 1.2993,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999697748224271e-05,
+      "loss": 1.3535,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9996641646936345e-05,
+      "loss": 1.1182,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999630581162998e-05,
+      "loss": 1.1246,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9995969976323614e-05,
+      "loss": 0.9924,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9995634141017245e-05,
+      "loss": 0.8466,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9995298305710883e-05,
+      "loss": 0.8867,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9994962470404515e-05,
+      "loss": 1.0111,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999462663509815e-05,
+      "loss": 0.8419,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9994290799791784e-05,
+      "loss": 0.7001,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999395496448542e-05,
+      "loss": 0.9088,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9993619129179053e-05,
+      "loss": 0.9874,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9993283293872688e-05,
+      "loss": 0.7527,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9992947458566322e-05,
+      "loss": 0.7898,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9992611623259954e-05,
+      "loss": 1.0525,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.999227578795359e-05,
+      "loss": 0.9424,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9991939952647223e-05,
+      "loss": 0.7397,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9991604117340858e-05,
+      "loss": 0.8743,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9991268282034492e-05,
+      "loss": 0.823,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9990932446728127e-05,
+      "loss": 0.918,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9990596611421758e-05,
+      "loss": 0.8737,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9990260776115396e-05,
+      "loss": 0.7008,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9989924940809028e-05,
+      "loss": 0.7558,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9989589105502662e-05,
+      "loss": 0.7667,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9989253270196297e-05,
+      "loss": 0.8866,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.998891743488993e-05,
+      "loss": 0.8967,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9988581599583566e-05,
+      "loss": 0.7586,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.99882457642772e-05,
+      "loss": 0.7273,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9987909928970835e-05,
+      "loss": 0.8199,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9987574093664467e-05,
+      "loss": 0.8221,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9987238258358105e-05,
+      "loss": 0.7832,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9986902423051736e-05,
+      "loss": 0.7018,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.998656658774537e-05,
+      "loss": 0.8973,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9986230752439005e-05,
+      "loss": 1.0396,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.998589491713264e-05,
+      "loss": 0.8673,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9985559081826275e-05,
+      "loss": 0.8214,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.998522324651991e-05,
+      "loss": 0.7413,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9984887411213544e-05,
+      "loss": 0.7065,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9984551575907175e-05,
+      "loss": 0.712,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9984215740600813e-05,
+      "loss": 0.7567,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9983879905294444e-05,
+      "loss": 0.5787,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.998354406998808e-05,
+      "loss": 0.9594,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9983208234681714e-05,
+      "loss": 0.801,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.998287239937535e-05,
+      "loss": 0.831,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9982536564068983e-05,
+      "loss": 0.5409,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9982200728762618e-05,
+      "loss": 0.6149,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9981864893456252e-05,
+      "loss": 0.533,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9981529058149884e-05,
+      "loss": 0.7663,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.998119322284352e-05,
+      "loss": 0.777,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9980857387537153e-05,
+      "loss": 0.5344,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9980521552230787e-05,
+      "loss": 0.5846,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9980185716924422e-05,
+      "loss": 1.0767,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9979849881618057e-05,
+      "loss": 0.9598,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.997951404631169e-05,
+      "loss": 0.7682,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9979178211005326e-05,
+      "loss": 0.7267,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.997884237569896e-05,
+      "loss": 0.8605,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9978506540392592e-05,
+      "loss": 0.8198,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.997817070508623e-05,
+      "loss": 0.56,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.997783486977986e-05,
+      "loss": 0.6555,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9977499034473496e-05,
+      "loss": 0.3716,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.997716319916713e-05,
+      "loss": 1.0856,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9976827363860765e-05,
+      "loss": 0.8227,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9976491528554396e-05,
+      "loss": 0.5664,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9976155693248034e-05,
+      "loss": 0.8449,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9975819857941666e-05,
+      "loss": 0.7847,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.99754840226353e-05,
+      "loss": 0.8769,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9975148187328935e-05,
+      "loss": 0.7504,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.997481235202257e-05,
+      "loss": 0.6447,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9974476516716204e-05,
+      "loss": 0.6119,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.997414068140984e-05,
+      "loss": 0.5936,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9973804846103474e-05,
+      "loss": 0.5861,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9973469010797105e-05,
+      "loss": 0.5844,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9973133175490743e-05,
+      "loss": 0.6046,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9972797340184374e-05,
+      "loss": 0.5555,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.997246150487801e-05,
+      "loss": 0.6198,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9972125669571643e-05,
+      "loss": 0.5524,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9971789834265278e-05,
+      "loss": 0.5758,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9971453998958913e-05,
+      "loss": 0.5572,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9971118163652547e-05,
+      "loss": 0.7048,
+      "step": 86
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9970782328346182e-05,
+      "loss": 0.6092,
+      "step": 87
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9970446493039813e-05,
+      "loss": 0.682,
+      "step": 88
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.997011065773345e-05,
+      "loss": 0.5857,
+      "step": 89
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9969774822427083e-05,
+      "loss": 0.6204,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9969438987120717e-05,
+      "loss": 0.4966,
+      "step": 91
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9969103151814352e-05,
+      "loss": 0.6958,
+      "step": 92
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9968767316507987e-05,
+      "loss": 0.5924,
+      "step": 93
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.996843148120162e-05,
+      "loss": 0.7461,
+      "step": 94
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9968095645895252e-05,
+      "loss": 0.513,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.996775981058889e-05,
+      "loss": 0.567,
+      "step": 96
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9967423975282522e-05,
+      "loss": 0.3815,
+      "step": 97
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9967088139976156e-05,
+      "loss": 0.7295,
+      "step": 98
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.996675230466979e-05,
+      "loss": 0.8831,
+      "step": 99
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9966416469363426e-05,
+      "loss": 0.6067,
+      "step": 100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.996608063405706e-05,
+      "loss": 0.5848,
+      "step": 101
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9965744798750695e-05,
+      "loss": 0.6773,
+      "step": 102
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.996540896344433e-05,
+      "loss": 0.612,
+      "step": 103
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.996507312813796e-05,
+      "loss": 0.6617,
+      "step": 104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99647372928316e-05,
+      "loss": 0.5402,
+      "step": 105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.996440145752523e-05,
+      "loss": 0.4367,
+      "step": 106
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9964065622218865e-05,
+      "loss": 0.4955,
+      "step": 107
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99637297869125e-05,
+      "loss": 0.6476,
+      "step": 108
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9963393951606134e-05,
+      "loss": 0.3654,
+      "step": 109
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.996305811629977e-05,
+      "loss": 0.5381,
+      "step": 110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9962722280993403e-05,
+      "loss": 0.5422,
+      "step": 111
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9962386445687038e-05,
+      "loss": 0.4254,
+      "step": 112
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.996205061038067e-05,
+      "loss": 0.6781,
+      "step": 113
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9961714775074304e-05,
+      "loss": 0.7867,
+      "step": 114
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.996137893976794e-05,
+      "loss": 0.5343,
+      "step": 115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9961043104461573e-05,
+      "loss": 0.6459,
+      "step": 116
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9960707269155208e-05,
+      "loss": 0.6911,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9960371433848843e-05,
+      "loss": 0.6785,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9960035598542474e-05,
+      "loss": 0.5713,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9959699763236112e-05,
+      "loss": 0.3622,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9959363927929743e-05,
+      "loss": 0.5195,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9959028092623378e-05,
+      "loss": 0.4742,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9958692257317012e-05,
+      "loss": 0.4861,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9958356422010647e-05,
+      "loss": 0.5866,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9958020586704282e-05,
+      "loss": 0.5951,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9957684751397916e-05,
+      "loss": 0.5515,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.995734891609155e-05,
+      "loss": 0.3442,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9957013080785182e-05,
+      "loss": 0.4036,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.995667724547882e-05,
+      "loss": 0.3663,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.995634141017245e-05,
+      "loss": 0.5339,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9956005574866086e-05,
+      "loss": 0.638,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.995566973955972e-05,
+      "loss": 0.3788,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9955333904253356e-05,
+      "loss": 0.5566,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.995499806894699e-05,
+      "loss": 0.4393,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9954662233640625e-05,
+      "loss": 0.7286,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.995432639833426e-05,
+      "loss": 0.4683,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.995399056302789e-05,
+      "loss": 0.4162,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.995365472772153e-05,
+      "loss": 0.5453,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.995331889241516e-05,
+      "loss": 0.6974,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9952983057108795e-05,
+      "loss": 0.6273,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.995264722180243e-05,
+      "loss": 0.5494,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9952311386496064e-05,
+      "loss": 0.4036,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99519755511897e-05,
+      "loss": 0.427,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9951639715883333e-05,
+      "loss": 0.5991,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9951303880576968e-05,
+      "loss": 0.4029,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99509680452706e-05,
+      "loss": 0.4777,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9950632209964237e-05,
+      "loss": 0.4167,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.995029637465787e-05,
+      "loss": 0.4319,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9949960539351503e-05,
+      "loss": 0.4453,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9949624704045138e-05,
+      "loss": 0.4353,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9949288868738772e-05,
+      "loss": 0.4931,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9948953033432407e-05,
+      "loss": 0.2802,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9948617198126042e-05,
+      "loss": 0.5785,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9948281362819676e-05,
+      "loss": 0.5551,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9947945527513308e-05,
+      "loss": 0.6461,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9947609692206946e-05,
+      "loss": 0.561,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9947273856900577e-05,
+      "loss": 0.5099,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994693802159421e-05,
+      "loss": 0.5208,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9946602186287846e-05,
+      "loss": 0.4737,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994626635098148e-05,
+      "loss": 0.4427,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9945930515675112e-05,
+      "loss": 0.4421,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994559468036875e-05,
+      "loss": 0.4964,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994525884506238e-05,
+      "loss": 0.5451,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9944923009756016e-05,
+      "loss": 0.4473,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994458717444965e-05,
+      "loss": 0.5064,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9944251339143285e-05,
+      "loss": 0.461,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994391550383692e-05,
+      "loss": 0.5343,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9943579668530555e-05,
+      "loss": 0.6357,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994324383322419e-05,
+      "loss": 0.5082,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994290799791782e-05,
+      "loss": 0.3591,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994257216261146e-05,
+      "loss": 0.3948,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994223632730509e-05,
+      "loss": 0.561,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9941900491998724e-05,
+      "loss": 0.3884,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994156465669236e-05,
+      "loss": 0.485,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9941228821385994e-05,
+      "loss": 0.3288,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.994089298607963e-05,
+      "loss": 0.5968,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9940557150773263e-05,
+      "loss": 0.5795,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9940221315466898e-05,
+      "loss": 0.5519,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.993988548016053e-05,
+      "loss": 0.4679,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9939549644854167e-05,
+      "loss": 0.7021,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9939213809547798e-05,
+      "loss": 0.6653,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9938877974241433e-05,
+      "loss": 0.4693,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9938542138935068e-05,
+      "loss": 0.5441,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9938206303628702e-05,
+      "loss": 0.3663,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9937870468322337e-05,
+      "loss": 0.5835,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.993753463301597e-05,
+      "loss": 0.5046,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9937198797709606e-05,
+      "loss": 0.4528,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9936862962403237e-05,
+      "loss": 0.6136,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9936527127096875e-05,
+      "loss": 0.7547,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9936191291790507e-05,
+      "loss": 0.4462,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.993585545648414e-05,
+      "loss": 0.3685,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9935519621177776e-05,
+      "loss": 0.3914,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.993518378587141e-05,
+      "loss": 0.4558,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9934847950565045e-05,
+      "loss": 0.5582,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.993451211525868e-05,
+      "loss": 0.7092,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9934176279952315e-05,
+      "loss": 0.3619,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9933840444645946e-05,
+      "loss": 0.6515,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9933504609339584e-05,
+      "loss": 0.4156,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9933168774033215e-05,
+      "loss": 0.5733,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.993283293872685e-05,
+      "loss": 0.441,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9932497103420484e-05,
+      "loss": 0.4352,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.993216126811412e-05,
+      "loss": 0.5038,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9931825432807754e-05,
+      "loss": 0.5543,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.993148959750139e-05,
+      "loss": 0.4302,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.993115376219502e-05,
+      "loss": 0.7253,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9930817926888654e-05,
+      "loss": 0.3824,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.993048209158229e-05,
+      "loss": 0.6076,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9930146256275924e-05,
+      "loss": 0.5548,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9929810420969558e-05,
+      "loss": 0.4782,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9929474585663193e-05,
+      "loss": 0.5488,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9929138750356828e-05,
+      "loss": 0.4108,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.992880291505046e-05,
+      "loss": 0.4265,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9928467079744097e-05,
+      "loss": 0.4574,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9928131244437728e-05,
+      "loss": 0.5294,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9927795409131363e-05,
+      "loss": 0.5009,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9927459573824997e-05,
+      "loss": 0.4493,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9927123738518632e-05,
+      "loss": 0.5583,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9926787903212267e-05,
+      "loss": 0.5238,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99264520679059e-05,
+      "loss": 0.3584,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9926116232599536e-05,
+      "loss": 0.5269,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9925780397293167e-05,
+      "loss": 0.4361,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9925444561986805e-05,
+      "loss": 0.45,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9925108726680437e-05,
+      "loss": 0.3541,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.992477289137407e-05,
+      "loss": 0.4437,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9924437056067706e-05,
+      "loss": 0.5253,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.992410122076134e-05,
+      "loss": 0.3149,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9923765385454975e-05,
+      "loss": 0.3276,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.992342955014861e-05,
+      "loss": 0.4898,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9923093714842244e-05,
+      "loss": 0.4141,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9922757879535876e-05,
+      "loss": 0.4611,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9922422044229514e-05,
+      "loss": 0.7247,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9922086208923145e-05,
+      "loss": 0.5005,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.992175037361678e-05,
+      "loss": 0.5732,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9921414538310414e-05,
+      "loss": 0.5129,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.992107870300405e-05,
+      "loss": 0.5213,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9920742867697684e-05,
+      "loss": 0.4841,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9920407032391318e-05,
+      "loss": 0.3604,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9920071197084953e-05,
+      "loss": 0.4309,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9919735361778584e-05,
+      "loss": 0.3123,
+      "step": 239
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9919399526472222e-05,
+      "loss": 0.6273,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9919063691165853e-05,
+      "loss": 0.4482,
+      "step": 241
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9918727855859488e-05,
+      "loss": 0.4659,
+      "step": 242
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9918392020553123e-05,
+      "loss": 0.7033,
+      "step": 243
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9918056185246757e-05,
+      "loss": 0.5704,
+      "step": 244
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9917720349940392e-05,
+      "loss": 0.4537,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9917384514634027e-05,
+      "loss": 0.3662,
+      "step": 246
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.991704867932766e-05,
+      "loss": 0.4824,
+      "step": 247
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9916712844021293e-05,
+      "loss": 0.4538,
+      "step": 248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9916377008714927e-05,
+      "loss": 0.5213,
+      "step": 249
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9916041173408562e-05,
+      "loss": 0.4015,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9915705338102196e-05,
+      "loss": 0.5109,
+      "step": 251
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.991536950279583e-05,
+      "loss": 0.3671,
+      "step": 252
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9915033667489466e-05,
+      "loss": 0.4451,
+      "step": 253
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9914697832183097e-05,
+      "loss": 0.6542,
+      "step": 254
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9914361996876735e-05,
+      "loss": 0.7174,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9914026161570366e-05,
+      "loss": 0.3688,
+      "step": 256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9913690326264e-05,
+      "loss": 0.6846,
+      "step": 257
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9913354490957636e-05,
+      "loss": 0.5029,
+      "step": 258
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.991301865565127e-05,
+      "loss": 0.3355,
+      "step": 259
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9912682820344905e-05,
+      "loss": 0.4964,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.991234698503854e-05,
+      "loss": 0.4159,
+      "step": 261
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9912011149732174e-05,
+      "loss": 0.5923,
+      "step": 262
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9911675314425805e-05,
+      "loss": 0.3992,
+      "step": 263
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9911339479119444e-05,
+      "loss": 0.4972,
+      "step": 264
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9911003643813075e-05,
+      "loss": 0.3692,
+      "step": 265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.991066780850671e-05,
+      "loss": 0.4831,
+      "step": 266
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9910331973200344e-05,
+      "loss": 0.464,
+      "step": 267
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.990999613789398e-05,
+      "loss": 0.6181,
+      "step": 268
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9909660302587613e-05,
+      "loss": 0.6496,
+      "step": 269
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9909324467281248e-05,
+      "loss": 0.4047,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9908988631974883e-05,
+      "loss": 0.3541,
+      "step": 271
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9908652796668514e-05,
+      "loss": 0.4749,
+      "step": 272
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.990831696136215e-05,
+      "loss": 0.4964,
+      "step": 273
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9907981126055783e-05,
+      "loss": 0.3945,
+      "step": 274
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9907645290749418e-05,
+      "loss": 0.5349,
+      "step": 275
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9907309455443053e-05,
+      "loss": 0.5005,
+      "step": 276
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9906973620136687e-05,
+      "loss": 0.462,
+      "step": 277
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9906637784830322e-05,
+      "loss": 0.4104,
+      "step": 278
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9906301949523953e-05,
+      "loss": 0.2658,
+      "step": 279
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.990596611421759e-05,
+      "loss": 0.65,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9905630278911222e-05,
+      "loss": 0.4223,
+      "step": 281
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9905294443604857e-05,
+      "loss": 0.3986,
+      "step": 282
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.990495860829849e-05,
+      "loss": 0.3904,
+      "step": 283
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9904622772992126e-05,
+      "loss": 0.3997,
+      "step": 284
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.990428693768576e-05,
+      "loss": 0.6181,
+      "step": 285
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9903951102379396e-05,
+      "loss": 0.4778,
+      "step": 286
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.990361526707303e-05,
+      "loss": 0.4388,
+      "step": 287
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.990327943176666e-05,
+      "loss": 0.5043,
+      "step": 288
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99029435964603e-05,
+      "loss": 0.3402,
+      "step": 289
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.990260776115393e-05,
+      "loss": 0.3159,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9902271925847565e-05,
+      "loss": 0.4995,
+      "step": 291
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.99019360905412e-05,
+      "loss": 0.4331,
+      "step": 292
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9901600255234835e-05,
+      "loss": 0.3722,
+      "step": 293
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.990126441992847e-05,
+      "loss": 0.3618,
+      "step": 294
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9900928584622104e-05,
+      "loss": 0.4643,
+      "step": 295
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9900592749315735e-05,
+      "loss": 0.3422,
+      "step": 296
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.990025691400937e-05,
+      "loss": 0.4206,
+      "step": 297
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9899921078703005e-05,
+      "loss": 0.3853,
+      "step": 298
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989958524339664e-05,
+      "loss": 0.5037,
+      "step": 299
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9899249408090274e-05,
+      "loss": 0.3228,
+      "step": 300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989891357278391e-05,
+      "loss": 0.3995,
+      "step": 301
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9898577737477543e-05,
+      "loss": 0.4747,
+      "step": 302
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9898241902171174e-05,
+      "loss": 0.4919,
+      "step": 303
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9897906066864812e-05,
+      "loss": 0.5166,
+      "step": 304
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9897570231558444e-05,
+      "loss": 0.4715,
+      "step": 305
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989723439625208e-05,
+      "loss": 0.4055,
+      "step": 306
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9896898560945713e-05,
+      "loss": 0.483,
+      "step": 307
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9896562725639348e-05,
+      "loss": 0.396,
+      "step": 308
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9896226890332982e-05,
+      "loss": 0.3359,
+      "step": 309
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9895891055026617e-05,
+      "loss": 0.361,
+      "step": 310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989555521972025e-05,
+      "loss": 0.5287,
+      "step": 311
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9895219384413883e-05,
+      "loss": 0.3148,
+      "step": 312
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989488354910752e-05,
+      "loss": 0.4606,
+      "step": 313
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9894547713801152e-05,
+      "loss": 0.2814,
+      "step": 314
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9894211878494787e-05,
+      "loss": 0.4562,
+      "step": 315
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989387604318842e-05,
+      "loss": 0.4698,
+      "step": 316
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9893540207882056e-05,
+      "loss": 0.3841,
+      "step": 317
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989320437257569e-05,
+      "loss": 0.3355,
+      "step": 318
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9892868537269325e-05,
+      "loss": 0.4517,
+      "step": 319
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989253270196296e-05,
+      "loss": 0.4576,
+      "step": 320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989219686665659e-05,
+      "loss": 0.3698,
+      "step": 321
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989186103135023e-05,
+      "loss": 0.3321,
+      "step": 322
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989152519604386e-05,
+      "loss": 0.4779,
+      "step": 323
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9891189360737495e-05,
+      "loss": 0.3912,
+      "step": 324
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.989085352543113e-05,
+      "loss": 0.5808,
+      "step": 325
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9890517690124765e-05,
+      "loss": 0.4902,
+      "step": 326
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.98901818548184e-05,
+      "loss": 0.3202,
+      "step": 327
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9889846019512034e-05,
+      "loss": 0.4658,
+      "step": 328
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.988951018420567e-05,
+      "loss": 0.4174,
+      "step": 329
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.98891743488993e-05,
+      "loss": 0.5007,
+      "step": 330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9888838513592938e-05,
+      "loss": 0.4794,
+      "step": 331
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.988850267828657e-05,
+      "loss": 0.4904,
+      "step": 332
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9888166842980204e-05,
+      "loss": 0.6579,
+      "step": 333
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.988783100767384e-05,
+      "loss": 0.4322,
+      "step": 334
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9887495172367473e-05,
+      "loss": 0.354,
+      "step": 335
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9887159337061108e-05,
+      "loss": 0.4782,
+      "step": 336
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9886823501754742e-05,
+      "loss": 0.3517,
+      "step": 337
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9886487666448377e-05,
+      "loss": 0.4278,
+      "step": 338
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9886151831142008e-05,
+      "loss": 0.539,
+      "step": 339
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9885815995835643e-05,
+      "loss": 0.4879,
+      "step": 340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9885480160529277e-05,
+      "loss": 0.4237,
+      "step": 341
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9885144325222912e-05,
+      "loss": 0.3734,
+      "step": 342
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9884808489916547e-05,
+      "loss": 0.5317,
+      "step": 343
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.988447265461018e-05,
+      "loss": 0.5222,
+      "step": 344
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9884136819303813e-05,
+      "loss": 0.506,
+      "step": 345
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.988380098399745e-05,
+      "loss": 0.5529,
+      "step": 346
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9883465148691082e-05,
+      "loss": 0.4537,
+      "step": 347
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9883129313384717e-05,
+      "loss": 0.366,
+      "step": 348
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.988279347807835e-05,
+      "loss": 0.5011,
+      "step": 349
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9882457642771986e-05,
+      "loss": 0.5014,
+      "step": 350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.988212180746562e-05,
+      "loss": 0.4449,
+      "step": 351
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9881785972159255e-05,
+      "loss": 0.5165,
+      "step": 352
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.988145013685289e-05,
+      "loss": 0.4026,
+      "step": 353
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.988111430154652e-05,
+      "loss": 0.397,
+      "step": 354
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.988077846624016e-05,
+      "loss": 0.5013,
+      "step": 355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.988044263093379e-05,
+      "loss": 0.5231,
+      "step": 356
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9880106795627425e-05,
+      "loss": 0.5152,
+      "step": 357
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.987977096032106e-05,
+      "loss": 0.4271,
+      "step": 358
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9879435125014694e-05,
+      "loss": 0.5221,
+      "step": 359
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.987909928970833e-05,
+      "loss": 0.5343,
+      "step": 360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9878763454401964e-05,
+      "loss": 0.622,
+      "step": 361
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9878427619095598e-05,
+      "loss": 0.4199,
+      "step": 362
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.987809178378923e-05,
+      "loss": 0.4985,
+      "step": 363
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9877755948482868e-05,
+      "loss": 0.3494,
+      "step": 364
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.98774201131765e-05,
+      "loss": 0.5034,
+      "step": 365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9877084277870134e-05,
+      "loss": 0.3418,
+      "step": 366
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9876748442563768e-05,
+      "loss": 0.3539,
+      "step": 367
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9876412607257403e-05,
+      "loss": 0.5393,
+      "step": 368
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9876076771951037e-05,
+      "loss": 0.2356,
+      "step": 369
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9875740936644672e-05,
+      "loss": 0.3795,
+      "step": 370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9875405101338307e-05,
+      "loss": 0.3155,
+      "step": 371
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9875069266031938e-05,
+      "loss": 0.4164,
+      "step": 372
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9874733430725576e-05,
+      "loss": 0.5925,
+      "step": 373
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9874397595419207e-05,
+      "loss": 0.5044,
+      "step": 374
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9874061760112842e-05,
+      "loss": 0.5438,
+      "step": 375
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9873725924806477e-05,
+      "loss": 0.5462,
+      "step": 376
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.987339008950011e-05,
+      "loss": 0.4151,
+      "step": 377
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9873054254193746e-05,
+      "loss": 0.531,
+      "step": 378
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.987271841888738e-05,
+      "loss": 0.4027,
+      "step": 379
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9872382583581015e-05,
+      "loss": 0.4632,
+      "step": 380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9872046748274646e-05,
+      "loss": 0.2534,
+      "step": 381
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9871710912968284e-05,
+      "loss": 0.2965,
+      "step": 382
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9871375077661916e-05,
+      "loss": 0.4244,
+      "step": 383
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.987103924235555e-05,
+      "loss": 0.5771,
+      "step": 384
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9870703407049185e-05,
+      "loss": 0.4724,
+      "step": 385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.987036757174282e-05,
+      "loss": 0.4377,
+      "step": 386
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.987003173643645e-05,
+      "loss": 0.4015,
+      "step": 387
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.986969590113009e-05,
+      "loss": 0.546,
+      "step": 388
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.986936006582372e-05,
+      "loss": 0.5124,
+      "step": 389
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9869024230517355e-05,
+      "loss": 0.3772,
+      "step": 390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.986868839521099e-05,
+      "loss": 0.6089,
+      "step": 391
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9868352559904624e-05,
+      "loss": 0.4333,
+      "step": 392
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.986801672459826e-05,
+      "loss": 0.2928,
+      "step": 393
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9867680889291893e-05,
+      "loss": 0.5861,
+      "step": 394
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9867345053985528e-05,
+      "loss": 0.3939,
+      "step": 395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.986700921867916e-05,
+      "loss": 0.3875,
+      "step": 396
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9866673383372797e-05,
+      "loss": 0.2154,
+      "step": 397
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.986633754806643e-05,
+      "loss": 0.6155,
+      "step": 398
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9866001712760063e-05,
+      "loss": 0.3012,
+      "step": 399
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9865665877453698e-05,
+      "loss": 0.4718,
+      "step": 400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9865330042147333e-05,
+      "loss": 0.2966,
+      "step": 401
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9864994206840967e-05,
+      "loss": 0.4023,
+      "step": 402
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9864658371534602e-05,
+      "loss": 0.2724,
+      "step": 403
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9864322536228237e-05,
+      "loss": 0.5206,
+      "step": 404
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9863986700921868e-05,
+      "loss": 0.4998,
+      "step": 405
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9863650865615506e-05,
+      "loss": 0.6521,
+      "step": 406
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9863315030309137e-05,
+      "loss": 0.3889,
+      "step": 407
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9862979195002772e-05,
+      "loss": 0.5935,
+      "step": 408
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9862643359696406e-05,
+      "loss": 0.3903,
+      "step": 409
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.986230752439004e-05,
+      "loss": 0.3851,
+      "step": 410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9861971689083676e-05,
+      "loss": 0.4489,
+      "step": 411
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.986163585377731e-05,
+      "loss": 0.365,
+      "step": 412
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9861300018470945e-05,
+      "loss": 0.3923,
+      "step": 413
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9860964183164576e-05,
+      "loss": 0.317,
+      "step": 414
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9860628347858214e-05,
+      "loss": 0.3817,
+      "step": 415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9860292512551846e-05,
+      "loss": 0.4521,
+      "step": 416
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.985995667724548e-05,
+      "loss": 0.3631,
+      "step": 417
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9859620841939115e-05,
+      "loss": 0.7045,
+      "step": 418
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.985928500663275e-05,
+      "loss": 0.5154,
+      "step": 419
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9858949171326384e-05,
+      "loss": 0.5468,
+      "step": 420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.985861333602002e-05,
+      "loss": 0.4464,
+      "step": 421
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9858277500713653e-05,
+      "loss": 0.426,
+      "step": 422
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9857941665407285e-05,
+      "loss": 0.4109,
+      "step": 423
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9857605830100923e-05,
+      "loss": 0.3414,
+      "step": 424
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9857269994794554e-05,
+      "loss": 0.3486,
+      "step": 425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.985693415948819e-05,
+      "loss": 0.3339,
+      "step": 426
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9856598324181823e-05,
+      "loss": 0.3921,
+      "step": 427
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9856262488875458e-05,
+      "loss": 0.4546,
+      "step": 428
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9855926653569093e-05,
+      "loss": 0.4699,
+      "step": 429
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9855590818262727e-05,
+      "loss": 0.3476,
+      "step": 430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.985525498295636e-05,
+      "loss": 0.5769,
+      "step": 431
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9854919147649993e-05,
+      "loss": 0.4674,
+      "step": 432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9854583312343628e-05,
+      "loss": 0.321,
+      "step": 433
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9854247477037262e-05,
+      "loss": 0.4678,
+      "step": 434
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9853911641730897e-05,
+      "loss": 0.453,
+      "step": 435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9853575806424532e-05,
+      "loss": 0.3988,
+      "step": 436
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9853239971118166e-05,
+      "loss": 0.4263,
+      "step": 437
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9852904135811798e-05,
+      "loss": 0.434,
+      "step": 438
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9852568300505436e-05,
+      "loss": 0.5516,
+      "step": 439
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9852232465199067e-05,
+      "loss": 0.3347,
+      "step": 440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.98518966298927e-05,
+      "loss": 0.3668,
+      "step": 441
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9851560794586336e-05,
+      "loss": 0.4696,
+      "step": 442
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.985122495927997e-05,
+      "loss": 0.4132,
+      "step": 443
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9850889123973606e-05,
+      "loss": 0.595,
+      "step": 444
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.985055328866724e-05,
+      "loss": 0.3253,
+      "step": 445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9850217453360875e-05,
+      "loss": 0.2949,
+      "step": 446
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9849881618054506e-05,
+      "loss": 0.3951,
+      "step": 447
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9849545782748144e-05,
+      "loss": 0.4324,
+      "step": 448
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9849209947441775e-05,
+      "loss": 0.3462,
+      "step": 449
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.984887411213541e-05,
+      "loss": 0.4882,
+      "step": 450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9848538276829045e-05,
+      "loss": 0.3446,
+      "step": 451
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.984820244152268e-05,
+      "loss": 0.4183,
+      "step": 452
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9847866606216314e-05,
+      "loss": 0.4709,
+      "step": 453
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9847530770909945e-05,
+      "loss": 0.4249,
+      "step": 454
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9847194935603583e-05,
+      "loss": 0.5169,
+      "step": 455
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9846859100297214e-05,
+      "loss": 0.4948,
+      "step": 456
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.984652326499085e-05,
+      "loss": 0.4526,
+      "step": 457
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9846187429684484e-05,
+      "loss": 0.2504,
+      "step": 458
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.984585159437812e-05,
+      "loss": 0.394,
+      "step": 459
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9845515759071753e-05,
+      "loss": 0.4169,
+      "step": 460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9845179923765388e-05,
+      "loss": 0.6637,
+      "step": 461
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9844844088459022e-05,
+      "loss": 0.346,
+      "step": 462
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9844508253152654e-05,
+      "loss": 0.3985,
+      "step": 463
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.984417241784629e-05,
+      "loss": 0.3563,
+      "step": 464
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9843836582539923e-05,
+      "loss": 0.5011,
+      "step": 465
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9843500747233558e-05,
+      "loss": 0.3932,
+      "step": 466
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9843164911927192e-05,
+      "loss": 0.4125,
+      "step": 467
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9842829076620827e-05,
+      "loss": 0.4845,
+      "step": 468
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.984249324131446e-05,
+      "loss": 0.3708,
+      "step": 469
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9842157406008096e-05,
+      "loss": 0.3781,
+      "step": 470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.984182157070173e-05,
+      "loss": 0.3943,
+      "step": 471
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9841485735395362e-05,
+      "loss": 0.4168,
+      "step": 472
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9841149900089e-05,
+      "loss": 0.4021,
+      "step": 473
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.984081406478263e-05,
+      "loss": 0.3983,
+      "step": 474
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9840478229476266e-05,
+      "loss": 0.5077,
+      "step": 475
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.98401423941699e-05,
+      "loss": 0.3738,
+      "step": 476
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9839806558863535e-05,
+      "loss": 0.4809,
+      "step": 477
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9839470723557167e-05,
+      "loss": 0.4033,
+      "step": 478
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9839134888250805e-05,
+      "loss": 0.2446,
+      "step": 479
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9838799052944436e-05,
+      "loss": 0.2273,
+      "step": 480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.983846321763807e-05,
+      "loss": 0.303,
+      "step": 481
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9838127382331705e-05,
+      "loss": 0.4334,
+      "step": 482
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.983779154702534e-05,
+      "loss": 0.3682,
+      "step": 483
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9837455711718974e-05,
+      "loss": 0.3911,
+      "step": 484
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.983711987641261e-05,
+      "loss": 0.4904,
+      "step": 485
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9836784041106244e-05,
+      "loss": 0.475,
+      "step": 486
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9836448205799875e-05,
+      "loss": 0.3776,
+      "step": 487
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9836112370493513e-05,
+      "loss": 0.439,
+      "step": 488
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9835776535187144e-05,
+      "loss": 0.4163,
+      "step": 489
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.983544069988078e-05,
+      "loss": 0.5103,
+      "step": 490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9835104864574414e-05,
+      "loss": 0.4019,
+      "step": 491
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9834769029268048e-05,
+      "loss": 0.3664,
+      "step": 492
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9834433193961683e-05,
+      "loss": 0.3962,
+      "step": 493
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9834097358655318e-05,
+      "loss": 0.7205,
+      "step": 494
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9833761523348952e-05,
+      "loss": 0.5025,
+      "step": 495
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.9833425688042583e-05,
+      "loss": 0.4376,
+      "step": 496
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.983308985273622e-05,
+      "loss": 0.4599,
+      "step": 497
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9832754017429853e-05,
+      "loss": 0.4096,
+      "step": 498
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9832418182123487e-05,
+      "loss": 0.2528,
+      "step": 499
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9832082346817122e-05,
+      "loss": 0.506,
+      "step": 500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9831746511510757e-05,
+      "loss": 0.3776,
+      "step": 501
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.983141067620439e-05,
+      "loss": 0.5254,
+      "step": 502
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9831074840898026e-05,
+      "loss": 0.4878,
+      "step": 503
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.983073900559166e-05,
+      "loss": 0.5239,
+      "step": 504
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9830403170285292e-05,
+      "loss": 0.4145,
+      "step": 505
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.983006733497893e-05,
+      "loss": 0.392,
+      "step": 506
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982973149967256e-05,
+      "loss": 0.4303,
+      "step": 507
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9829395664366196e-05,
+      "loss": 0.4246,
+      "step": 508
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982905982905983e-05,
+      "loss": 0.4816,
+      "step": 509
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9828723993753465e-05,
+      "loss": 0.4493,
+      "step": 510
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.98283881584471e-05,
+      "loss": 0.5901,
+      "step": 511
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9828052323140734e-05,
+      "loss": 0.3732,
+      "step": 512
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982771648783437e-05,
+      "loss": 0.4702,
+      "step": 513
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9827380652528e-05,
+      "loss": 0.354,
+      "step": 514
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982704481722164e-05,
+      "loss": 0.4776,
+      "step": 515
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982670898191527e-05,
+      "loss": 0.4799,
+      "step": 516
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9826373146608904e-05,
+      "loss": 0.4884,
+      "step": 517
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982603731130254e-05,
+      "loss": 0.3638,
+      "step": 518
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9825701475996174e-05,
+      "loss": 0.3329,
+      "step": 519
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9825365640689805e-05,
+      "loss": 0.3806,
+      "step": 520
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9825029805383443e-05,
+      "loss": 0.6512,
+      "step": 521
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9824693970077074e-05,
+      "loss": 0.3106,
+      "step": 522
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982435813477071e-05,
+      "loss": 0.4087,
+      "step": 523
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9824022299464343e-05,
+      "loss": 0.3412,
+      "step": 524
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9823686464157978e-05,
+      "loss": 0.264,
+      "step": 525
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9823350628851613e-05,
+      "loss": 0.3339,
+      "step": 526
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9823014793545247e-05,
+      "loss": 0.533,
+      "step": 527
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9822678958238882e-05,
+      "loss": 0.4532,
+      "step": 528
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9822343122932513e-05,
+      "loss": 0.6336,
+      "step": 529
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982200728762615e-05,
+      "loss": 0.3043,
+      "step": 530
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9821671452319783e-05,
+      "loss": 0.4029,
+      "step": 531
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9821335617013417e-05,
+      "loss": 0.1887,
+      "step": 532
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9820999781707052e-05,
+      "loss": 0.5388,
+      "step": 533
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9820663946400686e-05,
+      "loss": 0.524,
+      "step": 534
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.982032811109432e-05,
+      "loss": 0.4633,
+      "step": 535
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9819992275787956e-05,
+      "loss": 0.2739,
+      "step": 536
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.981965644048159e-05,
+      "loss": 0.4787,
+      "step": 537
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9819320605175222e-05,
+      "loss": 0.363,
+      "step": 538
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.981898476986886e-05,
+      "loss": 0.3426,
+      "step": 539
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.981864893456249e-05,
+      "loss": 0.3254,
+      "step": 540
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9818313099256126e-05,
+      "loss": 0.3544,
+      "step": 541
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.981797726394976e-05,
+      "loss": 0.3314,
+      "step": 542
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9817641428643395e-05,
+      "loss": 0.4414,
+      "step": 543
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.981730559333703e-05,
+      "loss": 0.4929,
+      "step": 544
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9816969758030664e-05,
+      "loss": 0.4132,
+      "step": 545
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.98166339227243e-05,
+      "loss": 0.2673,
+      "step": 546
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.981629808741793e-05,
+      "loss": 0.3582,
+      "step": 547
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9815962252111568e-05,
+      "loss": 0.2928,
+      "step": 548
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.98156264168052e-05,
+      "loss": 0.3502,
+      "step": 549
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9815290581498834e-05,
+      "loss": 0.6426,
+      "step": 550
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.981495474619247e-05,
+      "loss": 0.2715,
+      "step": 551
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9814618910886103e-05,
+      "loss": 0.4097,
+      "step": 552
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9814283075579738e-05,
+      "loss": 0.3702,
+      "step": 553
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9813947240273373e-05,
+      "loss": 0.3428,
+      "step": 554
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9813611404967007e-05,
+      "loss": 0.4108,
+      "step": 555
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.981327556966064e-05,
+      "loss": 0.4541,
+      "step": 556
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9812939734354277e-05,
+      "loss": 0.4158,
+      "step": 557
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9812603899047908e-05,
+      "loss": 0.4404,
+      "step": 558
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9812268063741543e-05,
+      "loss": 0.3573,
+      "step": 559
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9811932228435177e-05,
+      "loss": 0.4337,
+      "step": 560
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9811596393128812e-05,
+      "loss": 0.5096,
+      "step": 561
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9811260557822446e-05,
+      "loss": 0.2398,
+      "step": 562
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.981092472251608e-05,
+      "loss": 0.4213,
+      "step": 563
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9810588887209716e-05,
+      "loss": 0.3126,
+      "step": 564
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9810253051903347e-05,
+      "loss": 0.4266,
+      "step": 565
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980991721659698e-05,
+      "loss": 0.3732,
+      "step": 566
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9809581381290616e-05,
+      "loss": 0.3957,
+      "step": 567
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980924554598425e-05,
+      "loss": 0.5271,
+      "step": 568
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9808909710677886e-05,
+      "loss": 0.2188,
+      "step": 569
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980857387537152e-05,
+      "loss": 0.5168,
+      "step": 570
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980823804006515e-05,
+      "loss": 0.4964,
+      "step": 571
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980790220475879e-05,
+      "loss": 0.2946,
+      "step": 572
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980756636945242e-05,
+      "loss": 0.4692,
+      "step": 573
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9807230534146055e-05,
+      "loss": 0.4876,
+      "step": 574
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980689469883969e-05,
+      "loss": 0.3792,
+      "step": 575
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9806558863533325e-05,
+      "loss": 0.3294,
+      "step": 576
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980622302822696e-05,
+      "loss": 0.3841,
+      "step": 577
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9805887192920594e-05,
+      "loss": 0.5561,
+      "step": 578
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980555135761423e-05,
+      "loss": 0.367,
+      "step": 579
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980521552230786e-05,
+      "loss": 0.4371,
+      "step": 580
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9804879687001498e-05,
+      "loss": 0.2683,
+      "step": 581
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980454385169513e-05,
+      "loss": 0.4546,
+      "step": 582
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9804208016388764e-05,
+      "loss": 0.4846,
+      "step": 583
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.98038721810824e-05,
+      "loss": 0.3167,
+      "step": 584
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9803536345776033e-05,
+      "loss": 0.6098,
+      "step": 585
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9803200510469668e-05,
+      "loss": 0.2789,
+      "step": 586
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9802864675163302e-05,
+      "loss": 0.3008,
+      "step": 587
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9802528839856937e-05,
+      "loss": 0.4449,
+      "step": 588
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980219300455057e-05,
+      "loss": 0.3178,
+      "step": 589
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9801857169244206e-05,
+      "loss": 0.2972,
+      "step": 590
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9801521333937838e-05,
+      "loss": 0.4682,
+      "step": 591
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9801185498631472e-05,
+      "loss": 0.3464,
+      "step": 592
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9800849663325107e-05,
+      "loss": 0.2832,
+      "step": 593
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.980051382801874e-05,
+      "loss": 0.2989,
+      "step": 594
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9800177992712376e-05,
+      "loss": 0.3999,
+      "step": 595
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.979984215740601e-05,
+      "loss": 0.5006,
+      "step": 596
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9799506322099646e-05,
+      "loss": 0.6518,
+      "step": 597
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9799170486793277e-05,
+      "loss": 0.396,
+      "step": 598
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9798834651486915e-05,
+      "loss": 0.4364,
+      "step": 599
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9798498816180546e-05,
+      "loss": 0.3913,
+      "step": 600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.979816298087418e-05,
+      "loss": 0.3926,
+      "step": 601
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9797827145567815e-05,
+      "loss": 0.493,
+      "step": 602
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.979749131026145e-05,
+      "loss": 0.2769,
+      "step": 603
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9797155474955085e-05,
+      "loss": 0.4062,
+      "step": 604
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.979681963964872e-05,
+      "loss": 0.3741,
+      "step": 605
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9796483804342354e-05,
+      "loss": 0.3245,
+      "step": 606
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9796147969035985e-05,
+      "loss": 0.6284,
+      "step": 607
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9795812133729623e-05,
+      "loss": 0.4816,
+      "step": 608
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9795476298423255e-05,
+      "loss": 0.58,
+      "step": 609
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.979514046311689e-05,
+      "loss": 0.34,
+      "step": 610
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9794804627810524e-05,
+      "loss": 0.5136,
+      "step": 611
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.979446879250416e-05,
+      "loss": 0.2633,
+      "step": 612
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.979413295719779e-05,
+      "loss": 0.2695,
+      "step": 613
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9793797121891428e-05,
+      "loss": 0.2956,
+      "step": 614
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.979346128658506e-05,
+      "loss": 0.3261,
+      "step": 615
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9793125451278694e-05,
+      "loss": 0.626,
+      "step": 616
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.979278961597233e-05,
+      "loss": 0.3182,
+      "step": 617
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9792453780665963e-05,
+      "loss": 0.2866,
+      "step": 618
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9792117945359598e-05,
+      "loss": 0.6679,
+      "step": 619
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9791782110053232e-05,
+      "loss": 0.4799,
+      "step": 620
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9791446274746867e-05,
+      "loss": 0.5891,
+      "step": 621
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9791110439440498e-05,
+      "loss": 0.3867,
+      "step": 622
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9790774604134136e-05,
+      "loss": 0.3921,
+      "step": 623
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9790438768827767e-05,
+      "loss": 0.3301,
+      "step": 624
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9790102933521402e-05,
+      "loss": 0.4581,
+      "step": 625
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9789767098215037e-05,
+      "loss": 0.5019,
+      "step": 626
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.978943126290867e-05,
+      "loss": 0.4505,
+      "step": 627
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9789095427602306e-05,
+      "loss": 0.31,
+      "step": 628
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.978875959229594e-05,
+      "loss": 0.4376,
+      "step": 629
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9788423756989575e-05,
+      "loss": 0.3257,
+      "step": 630
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9788087921683207e-05,
+      "loss": 0.2847,
+      "step": 631
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.978775208637684e-05,
+      "loss": 0.3267,
+      "step": 632
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9787416251070476e-05,
+      "loss": 0.4553,
+      "step": 633
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.978708041576411e-05,
+      "loss": 0.41,
+      "step": 634
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9786744580457745e-05,
+      "loss": 0.3851,
+      "step": 635
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.978640874515138e-05,
+      "loss": 0.3677,
+      "step": 636
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9786072909845015e-05,
+      "loss": 0.4662,
+      "step": 637
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9785737074538646e-05,
+      "loss": 0.5688,
+      "step": 638
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9785401239232284e-05,
+      "loss": 0.3867,
+      "step": 639
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9785065403925915e-05,
+      "loss": 0.2705,
+      "step": 640
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.978472956861955e-05,
+      "loss": 0.4138,
+      "step": 641
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9784393733313184e-05,
+      "loss": 0.5638,
+      "step": 642
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.978405789800682e-05,
+      "loss": 0.3054,
+      "step": 643
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9783722062700454e-05,
+      "loss": 0.484,
+      "step": 644
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.978338622739409e-05,
+      "loss": 0.4834,
+      "step": 645
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9783050392087723e-05,
+      "loss": 0.4344,
+      "step": 646
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9782714556781354e-05,
+      "loss": 0.3445,
+      "step": 647
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9782378721474992e-05,
+      "loss": 0.4803,
+      "step": 648
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9782042886168624e-05,
+      "loss": 0.309,
+      "step": 649
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9781707050862258e-05,
+      "loss": 0.433,
+      "step": 650
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9781371215555893e-05,
+      "loss": 0.4656,
+      "step": 651
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9781035380249527e-05,
+      "loss": 0.4678,
+      "step": 652
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9780699544943162e-05,
+      "loss": 0.4531,
+      "step": 653
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9780363709636797e-05,
+      "loss": 0.5558,
+      "step": 654
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9780027874330428e-05,
+      "loss": 0.5011,
+      "step": 655
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9779692039024063e-05,
+      "loss": 0.3715,
+      "step": 656
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9779356203717697e-05,
+      "loss": 0.4095,
+      "step": 657
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9779020368411332e-05,
+      "loss": 0.3598,
+      "step": 658
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9778684533104967e-05,
+      "loss": 0.3865,
+      "step": 659
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.97783486977986e-05,
+      "loss": 0.2525,
+      "step": 660
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9778012862492236e-05,
+      "loss": 0.3036,
+      "step": 661
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9777677027185867e-05,
+      "loss": 0.4236,
+      "step": 662
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9777341191879505e-05,
+      "loss": 0.246,
+      "step": 663
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9777005356573136e-05,
+      "loss": 0.3674,
+      "step": 664
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.977666952126677e-05,
+      "loss": 0.5396,
+      "step": 665
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9776333685960406e-05,
+      "loss": 0.4734,
+      "step": 666
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.977599785065404e-05,
+      "loss": 0.4836,
+      "step": 667
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9775662015347675e-05,
+      "loss": 0.6125,
+      "step": 668
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.977532618004131e-05,
+      "loss": 0.4354,
+      "step": 669
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9774990344734944e-05,
+      "loss": 0.5577,
+      "step": 670
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9774654509428576e-05,
+      "loss": 0.2616,
+      "step": 671
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9774318674122214e-05,
+      "loss": 0.3774,
+      "step": 672
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9773982838815845e-05,
+      "loss": 0.4508,
+      "step": 673
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.977364700350948e-05,
+      "loss": 0.4679,
+      "step": 674
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9773311168203114e-05,
+      "loss": 0.4502,
+      "step": 675
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.977297533289675e-05,
+      "loss": 0.4419,
+      "step": 676
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9772639497590383e-05,
+      "loss": 0.3605,
+      "step": 677
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9772303662284018e-05,
+      "loss": 0.3528,
+      "step": 678
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9771967826977653e-05,
+      "loss": 0.416,
+      "step": 679
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9771631991671284e-05,
+      "loss": 0.3041,
+      "step": 680
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9771296156364922e-05,
+      "loss": 0.5701,
+      "step": 681
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9770960321058553e-05,
+      "loss": 0.4247,
+      "step": 682
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9770624485752188e-05,
+      "loss": 0.3227,
+      "step": 683
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9770288650445823e-05,
+      "loss": 0.3014,
+      "step": 684
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9769952815139457e-05,
+      "loss": 0.3091,
+      "step": 685
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9769616979833092e-05,
+      "loss": 0.396,
+      "step": 686
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9769281144526727e-05,
+      "loss": 0.3185,
+      "step": 687
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.976894530922036e-05,
+      "loss": 0.3621,
+      "step": 688
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9768609473913992e-05,
+      "loss": 0.3982,
+      "step": 689
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.976827363860763e-05,
+      "loss": 0.4608,
+      "step": 690
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9767937803301262e-05,
+      "loss": 0.3991,
+      "step": 691
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9767601967994896e-05,
+      "loss": 0.365,
+      "step": 692
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.976726613268853e-05,
+      "loss": 0.3031,
+      "step": 693
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9766930297382166e-05,
+      "loss": 0.3131,
+      "step": 694
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.97665944620758e-05,
+      "loss": 0.1248,
+      "step": 695
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9766258626769435e-05,
+      "loss": 0.4813,
+      "step": 696
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.976592279146307e-05,
+      "loss": 0.4703,
+      "step": 697
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.97655869561567e-05,
+      "loss": 0.4189,
+      "step": 698
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.976525112085034e-05,
+      "loss": 0.4046,
+      "step": 699
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.976491528554397e-05,
+      "loss": 0.5166,
+      "step": 700
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9764579450237605e-05,
+      "loss": 0.4923,
+      "step": 701
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.976424361493124e-05,
+      "loss": 0.6355,
+      "step": 702
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9763907779624874e-05,
+      "loss": 0.6686,
+      "step": 703
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9763571944318505e-05,
+      "loss": 0.3709,
+      "step": 704
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9763236109012143e-05,
+      "loss": 0.4317,
+      "step": 705
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9762900273705775e-05,
+      "loss": 0.468,
+      "step": 706
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.976256443839941e-05,
+      "loss": 0.3784,
+      "step": 707
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9762228603093044e-05,
+      "loss": 0.3315,
+      "step": 708
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.976189276778668e-05,
+      "loss": 0.482,
+      "step": 709
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9761556932480313e-05,
+      "loss": 0.6777,
+      "step": 710
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9761221097173948e-05,
+      "loss": 0.6501,
+      "step": 711
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9760885261867583e-05,
+      "loss": 0.3864,
+      "step": 712
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9760549426561214e-05,
+      "loss": 0.3951,
+      "step": 713
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9760213591254852e-05,
+      "loss": 0.3011,
+      "step": 714
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9759877755948483e-05,
+      "loss": 0.3301,
+      "step": 715
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9759541920642118e-05,
+      "loss": 0.4597,
+      "step": 716
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9759206085335752e-05,
+      "loss": 0.2886,
+      "step": 717
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9758870250029387e-05,
+      "loss": 0.3126,
+      "step": 718
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9758534414723022e-05,
+      "loss": 0.4869,
+      "step": 719
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9758198579416656e-05,
+      "loss": 0.4369,
+      "step": 720
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.975786274411029e-05,
+      "loss": 0.2716,
+      "step": 721
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9757526908803922e-05,
+      "loss": 0.3102,
+      "step": 722
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.975719107349756e-05,
+      "loss": 0.3978,
+      "step": 723
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.975685523819119e-05,
+      "loss": 0.4315,
+      "step": 724
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9756519402884826e-05,
+      "loss": 0.5047,
+      "step": 725
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.975618356757846e-05,
+      "loss": 0.4736,
+      "step": 726
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9755847732272096e-05,
+      "loss": 0.342,
+      "step": 727
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.975551189696573e-05,
+      "loss": 0.5377,
+      "step": 728
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9755176061659365e-05,
+      "loss": 0.4737,
+      "step": 729
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9754840226353e-05,
+      "loss": 0.3971,
+      "step": 730
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.975450439104663e-05,
+      "loss": 0.4379,
+      "step": 731
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.975416855574027e-05,
+      "loss": 0.3632,
+      "step": 732
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.97538327204339e-05,
+      "loss": 0.4888,
+      "step": 733
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9753496885127535e-05,
+      "loss": 0.5546,
+      "step": 734
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.975316104982117e-05,
+      "loss": 0.4774,
+      "step": 735
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9752825214514804e-05,
+      "loss": 0.3158,
+      "step": 736
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.975248937920844e-05,
+      "loss": 0.3526,
+      "step": 737
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9752153543902073e-05,
+      "loss": 0.5133,
+      "step": 738
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9751817708595708e-05,
+      "loss": 0.2416,
+      "step": 739
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.975148187328934e-05,
+      "loss": 0.4425,
+      "step": 740
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9751146037982977e-05,
+      "loss": 0.4164,
+      "step": 741
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.975081020267661e-05,
+      "loss": 0.3961,
+      "step": 742
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9750474367370243e-05,
+      "loss": 0.2251,
+      "step": 743
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9750138532063878e-05,
+      "loss": 0.3915,
+      "step": 744
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9749802696757512e-05,
+      "loss": 0.4695,
+      "step": 745
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9749466861451144e-05,
+      "loss": 0.3302,
+      "step": 746
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974913102614478e-05,
+      "loss": 0.413,
+      "step": 747
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9748795190838413e-05,
+      "loss": 0.2141,
+      "step": 748
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9748459355532048e-05,
+      "loss": 0.5261,
+      "step": 749
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9748123520225682e-05,
+      "loss": 0.5223,
+      "step": 750
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9747787684919317e-05,
+      "loss": 0.2544,
+      "step": 751
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974745184961295e-05,
+      "loss": 0.3283,
+      "step": 752
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9747116014306586e-05,
+      "loss": 0.4954,
+      "step": 753
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974678017900022e-05,
+      "loss": 0.335,
+      "step": 754
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9746444343693852e-05,
+      "loss": 0.2919,
+      "step": 755
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974610850838749e-05,
+      "loss": 0.3636,
+      "step": 756
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974577267308112e-05,
+      "loss": 0.4436,
+      "step": 757
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9745436837774756e-05,
+      "loss": 0.4482,
+      "step": 758
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974510100246839e-05,
+      "loss": 0.4064,
+      "step": 759
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9744765167162025e-05,
+      "loss": 0.3763,
+      "step": 760
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974442933185566e-05,
+      "loss": 0.2667,
+      "step": 761
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9744093496549295e-05,
+      "loss": 0.4783,
+      "step": 762
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974375766124293e-05,
+      "loss": 0.4084,
+      "step": 763
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974342182593656e-05,
+      "loss": 0.2898,
+      "step": 764
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.97430859906302e-05,
+      "loss": 0.398,
+      "step": 765
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974275015532383e-05,
+      "loss": 0.4457,
+      "step": 766
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9742414320017464e-05,
+      "loss": 0.4258,
+      "step": 767
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.97420784847111e-05,
+      "loss": 0.3932,
+      "step": 768
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9741742649404734e-05,
+      "loss": 0.3834,
+      "step": 769
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974140681409837e-05,
+      "loss": 0.3885,
+      "step": 770
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9741070978792003e-05,
+      "loss": 0.3226,
+      "step": 771
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9740735143485638e-05,
+      "loss": 0.4173,
+      "step": 772
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.974039930817927e-05,
+      "loss": 0.4888,
+      "step": 773
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9740063472872907e-05,
+      "loss": 0.4029,
+      "step": 774
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9739727637566538e-05,
+      "loss": 0.2241,
+      "step": 775
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9739391802260173e-05,
+      "loss": 0.4375,
+      "step": 776
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9739055966953808e-05,
+      "loss": 0.2085,
+      "step": 777
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9738720131647442e-05,
+      "loss": 0.3111,
+      "step": 778
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9738384296341077e-05,
+      "loss": 0.3534,
+      "step": 779
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973804846103471e-05,
+      "loss": 0.465,
+      "step": 780
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9737712625728346e-05,
+      "loss": 0.2698,
+      "step": 781
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9737376790421977e-05,
+      "loss": 0.4297,
+      "step": 782
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9737040955115615e-05,
+      "loss": 0.4896,
+      "step": 783
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9736705119809247e-05,
+      "loss": 0.4433,
+      "step": 784
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973636928450288e-05,
+      "loss": 0.3636,
+      "step": 785
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9736033449196516e-05,
+      "loss": 0.5329,
+      "step": 786
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973569761389015e-05,
+      "loss": 0.4465,
+      "step": 787
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9735361778583785e-05,
+      "loss": 0.3258,
+      "step": 788
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973502594327742e-05,
+      "loss": 0.2862,
+      "step": 789
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973469010797105e-05,
+      "loss": 0.621,
+      "step": 790
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9734354272664686e-05,
+      "loss": 0.4218,
+      "step": 791
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973401843735832e-05,
+      "loss": 0.3824,
+      "step": 792
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9733682602051955e-05,
+      "loss": 0.2466,
+      "step": 793
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973334676674559e-05,
+      "loss": 0.3761,
+      "step": 794
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9733010931439224e-05,
+      "loss": 0.3406,
+      "step": 795
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973267509613286e-05,
+      "loss": 0.5093,
+      "step": 796
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973233926082649e-05,
+      "loss": 0.4917,
+      "step": 797
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973200342552013e-05,
+      "loss": 0.3968,
+      "step": 798
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973166759021376e-05,
+      "loss": 0.3639,
+      "step": 799
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9731331754907394e-05,
+      "loss": 0.4845,
+      "step": 800
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.973099591960103e-05,
+      "loss": 0.3246,
+      "step": 801
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9730660084294664e-05,
+      "loss": 0.3354,
+      "step": 802
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9730324248988298e-05,
+      "loss": 0.282,
+      "step": 803
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9729988413681933e-05,
+      "loss": 0.3847,
+      "step": 804
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9729652578375568e-05,
+      "loss": 0.5633,
+      "step": 805
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.97293167430692e-05,
+      "loss": 0.3132,
+      "step": 806
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9728980907762837e-05,
+      "loss": 0.378,
+      "step": 807
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9728645072456468e-05,
+      "loss": 0.2724,
+      "step": 808
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9728309237150103e-05,
+      "loss": 0.4854,
+      "step": 809
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9727973401843737e-05,
+      "loss": 0.2667,
+      "step": 810
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9727637566537372e-05,
+      "loss": 0.3005,
+      "step": 811
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9727301731231007e-05,
+      "loss": 0.5149,
+      "step": 812
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9726965895924638e-05,
+      "loss": 0.3835,
+      "step": 813
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9726630060618276e-05,
+      "loss": 0.468,
+      "step": 814
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9726294225311907e-05,
+      "loss": 0.3969,
+      "step": 815
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9725958390005542e-05,
+      "loss": 0.5154,
+      "step": 816
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9725622554699177e-05,
+      "loss": 0.3569,
+      "step": 817
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.972528671939281e-05,
+      "loss": 0.3793,
+      "step": 818
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9724950884086446e-05,
+      "loss": 0.3585,
+      "step": 819
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.972461504878008e-05,
+      "loss": 0.4226,
+      "step": 820
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9724279213473715e-05,
+      "loss": 0.3776,
+      "step": 821
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9723943378167346e-05,
+      "loss": 0.2679,
+      "step": 822
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9723607542860984e-05,
+      "loss": 0.3416,
+      "step": 823
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9723271707554616e-05,
+      "loss": 0.6008,
+      "step": 824
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.972293587224825e-05,
+      "loss": 0.4115,
+      "step": 825
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9722600036941885e-05,
+      "loss": 0.26,
+      "step": 826
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.972226420163552e-05,
+      "loss": 0.4725,
+      "step": 827
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9721928366329154e-05,
+      "loss": 0.4252,
+      "step": 828
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.972159253102279e-05,
+      "loss": 0.3769,
+      "step": 829
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9721256695716424e-05,
+      "loss": 0.3876,
+      "step": 830
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9720920860410055e-05,
+      "loss": 0.4079,
+      "step": 831
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9720585025103693e-05,
+      "loss": 0.3051,
+      "step": 832
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9720249189797324e-05,
+      "loss": 0.42,
+      "step": 833
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.971991335449096e-05,
+      "loss": 0.3878,
+      "step": 834
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9719577519184593e-05,
+      "loss": 0.3935,
+      "step": 835
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9719241683878228e-05,
+      "loss": 0.4316,
+      "step": 836
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.971890584857186e-05,
+      "loss": 0.4445,
+      "step": 837
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9718570013265497e-05,
+      "loss": 0.4943,
+      "step": 838
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.971823417795913e-05,
+      "loss": 0.422,
+      "step": 839
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9717898342652763e-05,
+      "loss": 0.3443,
+      "step": 840
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9717562507346398e-05,
+      "loss": 0.4062,
+      "step": 841
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9717226672040033e-05,
+      "loss": 0.2998,
+      "step": 842
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9716890836733667e-05,
+      "loss": 0.4377,
+      "step": 843
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9716555001427302e-05,
+      "loss": 0.4622,
+      "step": 844
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9716219166120936e-05,
+      "loss": 0.4481,
+      "step": 845
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9715883330814568e-05,
+      "loss": 0.3802,
+      "step": 846
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9715547495508206e-05,
+      "loss": 0.5058,
+      "step": 847
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9715211660201837e-05,
+      "loss": 0.4566,
+      "step": 848
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.971487582489547e-05,
+      "loss": 0.4646,
+      "step": 849
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9714539989589106e-05,
+      "loss": 0.2883,
+      "step": 850
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.971420415428274e-05,
+      "loss": 0.3287,
+      "step": 851
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9713868318976376e-05,
+      "loss": 0.4167,
+      "step": 852
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.971353248367001e-05,
+      "loss": 0.4115,
+      "step": 853
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9713196648363645e-05,
+      "loss": 0.3008,
+      "step": 854
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9712860813057276e-05,
+      "loss": 0.3092,
+      "step": 855
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9712524977750914e-05,
+      "loss": 0.4164,
+      "step": 856
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9712189142444545e-05,
+      "loss": 0.2917,
+      "step": 857
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.971185330713818e-05,
+      "loss": 0.3543,
+      "step": 858
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9711517471831815e-05,
+      "loss": 0.3537,
+      "step": 859
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.971118163652545e-05,
+      "loss": 0.3689,
+      "step": 860
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9710845801219084e-05,
+      "loss": 0.2698,
+      "step": 861
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.971050996591272e-05,
+      "loss": 0.4018,
+      "step": 862
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9710174130606353e-05,
+      "loss": 0.3917,
+      "step": 863
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9709838295299985e-05,
+      "loss": 0.4734,
+      "step": 864
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9709502459993623e-05,
+      "loss": 0.3455,
+      "step": 865
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9709166624687254e-05,
+      "loss": 0.5053,
+      "step": 866
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.970883078938089e-05,
+      "loss": 0.5746,
+      "step": 867
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9708494954074523e-05,
+      "loss": 0.5604,
+      "step": 868
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9708159118768158e-05,
+      "loss": 0.3292,
+      "step": 869
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9707823283461792e-05,
+      "loss": 0.5119,
+      "step": 870
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9707487448155427e-05,
+      "loss": 0.4381,
+      "step": 871
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9707151612849062e-05,
+      "loss": 0.4372,
+      "step": 872
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9706815777542693e-05,
+      "loss": 0.6066,
+      "step": 873
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.970647994223633e-05,
+      "loss": 0.395,
+      "step": 874
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9706144106929962e-05,
+      "loss": 0.2989,
+      "step": 875
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9705808271623597e-05,
+      "loss": 0.4247,
+      "step": 876
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.970547243631723e-05,
+      "loss": 0.3419,
+      "step": 877
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9705136601010866e-05,
+      "loss": 0.4921,
+      "step": 878
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.97048007657045e-05,
+      "loss": 0.562,
+      "step": 879
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9704464930398136e-05,
+      "loss": 0.4972,
+      "step": 880
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9704129095091767e-05,
+      "loss": 0.4437,
+      "step": 881
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.97037932597854e-05,
+      "loss": 0.3914,
+      "step": 882
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9703457424479036e-05,
+      "loss": 0.3048,
+      "step": 883
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.970312158917267e-05,
+      "loss": 0.4045,
+      "step": 884
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9702785753866305e-05,
+      "loss": 0.34,
+      "step": 885
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.970244991855994e-05,
+      "loss": 0.4608,
+      "step": 886
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9702114083253575e-05,
+      "loss": 0.6024,
+      "step": 887
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9701778247947206e-05,
+      "loss": 0.4678,
+      "step": 888
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9701442412640844e-05,
+      "loss": 0.4656,
+      "step": 889
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9701106577334475e-05,
+      "loss": 0.3182,
+      "step": 890
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.970077074202811e-05,
+      "loss": 0.3782,
+      "step": 891
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9700434906721745e-05,
+      "loss": 0.2542,
+      "step": 892
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.970009907141538e-05,
+      "loss": 0.4772,
+      "step": 893
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9699763236109014e-05,
+      "loss": 0.2756,
+      "step": 894
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969942740080265e-05,
+      "loss": 0.4224,
+      "step": 895
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9699091565496283e-05,
+      "loss": 0.5184,
+      "step": 896
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9698755730189914e-05,
+      "loss": 0.3379,
+      "step": 897
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9698419894883552e-05,
+      "loss": 0.1787,
+      "step": 898
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9698084059577184e-05,
+      "loss": 0.4246,
+      "step": 899
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969774822427082e-05,
+      "loss": 0.2216,
+      "step": 900
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9697412388964453e-05,
+      "loss": 0.2818,
+      "step": 901
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9697076553658088e-05,
+      "loss": 0.5993,
+      "step": 902
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9696740718351722e-05,
+      "loss": 0.481,
+      "step": 903
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9696404883045357e-05,
+      "loss": 0.3376,
+      "step": 904
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969606904773899e-05,
+      "loss": 0.3826,
+      "step": 905
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9695733212432623e-05,
+      "loss": 0.3997,
+      "step": 906
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969539737712626e-05,
+      "loss": 0.2339,
+      "step": 907
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9695061541819892e-05,
+      "loss": 0.2443,
+      "step": 908
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9694725706513527e-05,
+      "loss": 0.4727,
+      "step": 909
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969438987120716e-05,
+      "loss": 0.4685,
+      "step": 910
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9694054035900796e-05,
+      "loss": 0.419,
+      "step": 911
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969371820059443e-05,
+      "loss": 0.3749,
+      "step": 912
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9693382365288065e-05,
+      "loss": 0.2958,
+      "step": 913
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.96930465299817e-05,
+      "loss": 0.3075,
+      "step": 914
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969271069467533e-05,
+      "loss": 0.5428,
+      "step": 915
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969237485936897e-05,
+      "loss": 0.3772,
+      "step": 916
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.96920390240626e-05,
+      "loss": 0.4647,
+      "step": 917
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9691703188756235e-05,
+      "loss": 0.3857,
+      "step": 918
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969136735344987e-05,
+      "loss": 0.426,
+      "step": 919
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9691031518143505e-05,
+      "loss": 0.3005,
+      "step": 920
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969069568283714e-05,
+      "loss": 0.3723,
+      "step": 921
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9690359847530774e-05,
+      "loss": 0.3118,
+      "step": 922
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.969002401222441e-05,
+      "loss": 0.488,
+      "step": 923
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.968968817691804e-05,
+      "loss": 0.3766,
+      "step": 924
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9689352341611678e-05,
+      "loss": 0.3542,
+      "step": 925
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.968901650630531e-05,
+      "loss": 0.372,
+      "step": 926
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9688680670998944e-05,
+      "loss": 0.4223,
+      "step": 927
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.968834483569258e-05,
+      "loss": 0.4463,
+      "step": 928
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9688009000386213e-05,
+      "loss": 0.3311,
+      "step": 929
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9687673165079844e-05,
+      "loss": 0.5073,
+      "step": 930
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9687337329773482e-05,
+      "loss": 0.2799,
+      "step": 931
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9687001494467114e-05,
+      "loss": 0.4132,
+      "step": 932
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9686665659160748e-05,
+      "loss": 0.5257,
+      "step": 933
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9686329823854383e-05,
+      "loss": 0.3148,
+      "step": 934
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9685993988548017e-05,
+      "loss": 0.8048,
+      "step": 935
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9685658153241652e-05,
+      "loss": 0.3838,
+      "step": 936
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9685322317935287e-05,
+      "loss": 0.3123,
+      "step": 937
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.968498648262892e-05,
+      "loss": 0.4406,
+      "step": 938
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9684650647322553e-05,
+      "loss": 0.372,
+      "step": 939
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.968431481201619e-05,
+      "loss": 0.4239,
+      "step": 940
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9683978976709822e-05,
+      "loss": 0.3554,
+      "step": 941
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9683643141403457e-05,
+      "loss": 0.5699,
+      "step": 942
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.968330730609709e-05,
+      "loss": 0.3624,
+      "step": 943
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9682971470790726e-05,
+      "loss": 0.4102,
+      "step": 944
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.968263563548436e-05,
+      "loss": 0.5734,
+      "step": 945
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9682299800177995e-05,
+      "loss": 0.3055,
+      "step": 946
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.968196396487163e-05,
+      "loss": 0.4933,
+      "step": 947
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.968162812956526e-05,
+      "loss": 0.5261,
+      "step": 948
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.96812922942589e-05,
+      "loss": 0.5729,
+      "step": 949
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.968095645895253e-05,
+      "loss": 0.4149,
+      "step": 950
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9680620623646165e-05,
+      "loss": 0.3652,
+      "step": 951
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.96802847883398e-05,
+      "loss": 0.4124,
+      "step": 952
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9679948953033434e-05,
+      "loss": 0.4111,
+      "step": 953
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.967961311772707e-05,
+      "loss": 0.3067,
+      "step": 954
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9679277282420704e-05,
+      "loss": 0.307,
+      "step": 955
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9678941447114338e-05,
+      "loss": 0.2865,
+      "step": 956
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.967860561180797e-05,
+      "loss": 0.4562,
+      "step": 957
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9678269776501608e-05,
+      "loss": 0.3608,
+      "step": 958
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.967793394119524e-05,
+      "loss": 0.3323,
+      "step": 959
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9677598105888873e-05,
+      "loss": 0.597,
+      "step": 960
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9677262270582508e-05,
+      "loss": 0.29,
+      "step": 961
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9676926435276143e-05,
+      "loss": 0.4673,
+      "step": 962
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9676590599969777e-05,
+      "loss": 0.3373,
+      "step": 963
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9676254764663412e-05,
+      "loss": 0.4389,
+      "step": 964
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9675918929357047e-05,
+      "loss": 0.3641,
+      "step": 965
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9675583094050678e-05,
+      "loss": 0.3971,
+      "step": 966
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9675247258744316e-05,
+      "loss": 0.3573,
+      "step": 967
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9674911423437947e-05,
+      "loss": 0.4167,
+      "step": 968
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9674575588131582e-05,
+      "loss": 0.53,
+      "step": 969
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9674239752825217e-05,
+      "loss": 0.4877,
+      "step": 970
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.967390391751885e-05,
+      "loss": 0.5838,
+      "step": 971
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9673568082212482e-05,
+      "loss": 0.4325,
+      "step": 972
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.967323224690612e-05,
+      "loss": 0.3065,
+      "step": 973
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9672896411599752e-05,
+      "loss": 0.2925,
+      "step": 974
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9672560576293386e-05,
+      "loss": 0.3802,
+      "step": 975
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.967222474098702e-05,
+      "loss": 0.348,
+      "step": 976
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9671888905680656e-05,
+      "loss": 0.3984,
+      "step": 977
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.967155307037429e-05,
+      "loss": 0.2901,
+      "step": 978
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9671217235067925e-05,
+      "loss": 0.3002,
+      "step": 979
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.967088139976156e-05,
+      "loss": 0.4126,
+      "step": 980
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.967054556445519e-05,
+      "loss": 0.4254,
+      "step": 981
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.967020972914883e-05,
+      "loss": 0.4358,
+      "step": 982
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.966987389384246e-05,
+      "loss": 0.3653,
+      "step": 983
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9669538058536095e-05,
+      "loss": 0.4527,
+      "step": 984
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.966920222322973e-05,
+      "loss": 0.4387,
+      "step": 985
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9668866387923364e-05,
+      "loss": 0.3496,
+      "step": 986
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9668530552617e-05,
+      "loss": 0.335,
+      "step": 987
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9668194717310633e-05,
+      "loss": 0.5118,
+      "step": 988
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9667858882004268e-05,
+      "loss": 0.596,
+      "step": 989
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.96675230466979e-05,
+      "loss": 0.2374,
+      "step": 990
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9667187211391537e-05,
+      "loss": 0.4111,
+      "step": 991
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.966685137608517e-05,
+      "loss": 0.4304,
+      "step": 992
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9666515540778803e-05,
+      "loss": 0.3963,
+      "step": 993
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9666179705472438e-05,
+      "loss": 0.3646,
+      "step": 994
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9665843870166073e-05,
+      "loss": 0.2936,
+      "step": 995
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9665508034859707e-05,
+      "loss": 0.2334,
+      "step": 996
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.966517219955334e-05,
+      "loss": 0.3901,
+      "step": 997
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9664836364246977e-05,
+      "loss": 0.4816,
+      "step": 998
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9664500528940608e-05,
+      "loss": 0.3948,
+      "step": 999
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9664164693634242e-05,
+      "loss": 0.4289,
+      "step": 1000
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9663828858327877e-05,
+      "loss": 0.267,
+      "step": 1001
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9663493023021512e-05,
+      "loss": 0.3139,
+      "step": 1002
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9663157187715146e-05,
+      "loss": 0.3018,
+      "step": 1003
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.966282135240878e-05,
+      "loss": 0.3899,
+      "step": 1004
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9662485517102416e-05,
+      "loss": 0.3828,
+      "step": 1005
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9662149681796047e-05,
+      "loss": 0.2704,
+      "step": 1006
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9661813846489685e-05,
+      "loss": 0.2443,
+      "step": 1007
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9661478011183316e-05,
+      "loss": 0.2918,
+      "step": 1008
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.966114217587695e-05,
+      "loss": 0.3332,
+      "step": 1009
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9660806340570586e-05,
+      "loss": 0.3593,
+      "step": 1010
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.966047050526422e-05,
+      "loss": 0.2472,
+      "step": 1011
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9660134669957855e-05,
+      "loss": 0.4826,
+      "step": 1012
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.965979883465149e-05,
+      "loss": 0.3448,
+      "step": 1013
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9659462999345124e-05,
+      "loss": 0.406,
+      "step": 1014
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9659127164038755e-05,
+      "loss": 0.3577,
+      "step": 1015
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.965879132873239e-05,
+      "loss": 0.4292,
+      "step": 1016
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9658455493426025e-05,
+      "loss": 0.3298,
+      "step": 1017
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.965811965811966e-05,
+      "loss": 0.3384,
+      "step": 1018
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9657783822813294e-05,
+      "loss": 0.281,
+      "step": 1019
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.965744798750693e-05,
+      "loss": 0.3854,
+      "step": 1020
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.965711215220056e-05,
+      "loss": 0.4843,
+      "step": 1021
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9656776316894198e-05,
+      "loss": 0.3604,
+      "step": 1022
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.965644048158783e-05,
+      "loss": 0.334,
+      "step": 1023
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9656104646281464e-05,
+      "loss": 0.6628,
+      "step": 1024
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.96557688109751e-05,
+      "loss": 0.4328,
+      "step": 1025
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9655432975668733e-05,
+      "loss": 0.3595,
+      "step": 1026
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9655097140362368e-05,
+      "loss": 0.4079,
+      "step": 1027
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9654761305056002e-05,
+      "loss": 0.2962,
+      "step": 1028
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9654425469749637e-05,
+      "loss": 0.313,
+      "step": 1029
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.965408963444327e-05,
+      "loss": 0.3694,
+      "step": 1030
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9653753799136906e-05,
+      "loss": 0.2411,
+      "step": 1031
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9653417963830538e-05,
+      "loss": 0.4152,
+      "step": 1032
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9653082128524172e-05,
+      "loss": 0.3123,
+      "step": 1033
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9652746293217807e-05,
+      "loss": 0.3323,
+      "step": 1034
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.965241045791144e-05,
+      "loss": 0.5504,
+      "step": 1035
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9652074622605076e-05,
+      "loss": 0.4988,
+      "step": 1036
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.965173878729871e-05,
+      "loss": 0.284,
+      "step": 1037
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9651402951992345e-05,
+      "loss": 0.6631,
+      "step": 1038
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9651067116685977e-05,
+      "loss": 0.466,
+      "step": 1039
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9650731281379615e-05,
+      "loss": 0.2889,
+      "step": 1040
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9650395446073246e-05,
+      "loss": 0.2209,
+      "step": 1041
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.965005961076688e-05,
+      "loss": 0.3449,
+      "step": 1042
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9649723775460515e-05,
+      "loss": 0.3936,
+      "step": 1043
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.964938794015415e-05,
+      "loss": 0.2975,
+      "step": 1044
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9649052104847785e-05,
+      "loss": 0.611,
+      "step": 1045
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.964871626954142e-05,
+      "loss": 0.2692,
+      "step": 1046
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9648380434235054e-05,
+      "loss": 0.3081,
+      "step": 1047
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9648044598928685e-05,
+      "loss": 0.2505,
+      "step": 1048
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9647708763622323e-05,
+      "loss": 0.3103,
+      "step": 1049
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9647372928315954e-05,
+      "loss": 0.4105,
+      "step": 1050
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.964703709300959e-05,
+      "loss": 0.3798,
+      "step": 1051
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9646701257703224e-05,
+      "loss": 0.3311,
+      "step": 1052
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.964636542239686e-05,
+      "loss": 0.4623,
+      "step": 1053
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9646029587090493e-05,
+      "loss": 0.4477,
+      "step": 1054
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9645693751784128e-05,
+      "loss": 0.4059,
+      "step": 1055
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9645357916477762e-05,
+      "loss": 0.3024,
+      "step": 1056
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9645022081171394e-05,
+      "loss": 0.2963,
+      "step": 1057
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.964468624586503e-05,
+      "loss": 0.4235,
+      "step": 1058
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9644350410558663e-05,
+      "loss": 0.479,
+      "step": 1059
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9644014575252298e-05,
+      "loss": 0.3222,
+      "step": 1060
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9643678739945932e-05,
+      "loss": 0.3866,
+      "step": 1061
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9643342904639567e-05,
+      "loss": 0.4668,
+      "step": 1062
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9643007069333198e-05,
+      "loss": 0.3407,
+      "step": 1063
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9642671234026836e-05,
+      "loss": 0.514,
+      "step": 1064
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9642335398720467e-05,
+      "loss": 0.3846,
+      "step": 1065
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9641999563414102e-05,
+      "loss": 0.339,
+      "step": 1066
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9641663728107737e-05,
+      "loss": 0.222,
+      "step": 1067
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.964132789280137e-05,
+      "loss": 0.408,
+      "step": 1068
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9640992057495006e-05,
+      "loss": 0.4306,
+      "step": 1069
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.964065622218864e-05,
+      "loss": 0.3577,
+      "step": 1070
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9640320386882275e-05,
+      "loss": 0.4065,
+      "step": 1071
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9639984551575907e-05,
+      "loss": 0.4048,
+      "step": 1072
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9639648716269545e-05,
+      "loss": 0.3365,
+      "step": 1073
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9639312880963176e-05,
+      "loss": 0.4642,
+      "step": 1074
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.963897704565681e-05,
+      "loss": 0.2707,
+      "step": 1075
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9638641210350445e-05,
+      "loss": 0.5487,
+      "step": 1076
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.963830537504408e-05,
+      "loss": 0.4623,
+      "step": 1077
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9637969539737714e-05,
+      "loss": 0.5437,
+      "step": 1078
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.963763370443135e-05,
+      "loss": 0.4287,
+      "step": 1079
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9637297869124984e-05,
+      "loss": 0.457,
+      "step": 1080
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9636962033818615e-05,
+      "loss": 0.5184,
+      "step": 1081
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9636626198512253e-05,
+      "loss": 0.3271,
+      "step": 1082
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9636290363205884e-05,
+      "loss": 0.3187,
+      "step": 1083
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.963595452789952e-05,
+      "loss": 0.457,
+      "step": 1084
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9635618692593154e-05,
+      "loss": 0.3064,
+      "step": 1085
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9635282857286788e-05,
+      "loss": 0.4292,
+      "step": 1086
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9634947021980423e-05,
+      "loss": 0.4081,
+      "step": 1087
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9634611186674058e-05,
+      "loss": 0.4954,
+      "step": 1088
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9634275351367692e-05,
+      "loss": 0.276,
+      "step": 1089
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9633939516061323e-05,
+      "loss": 0.3579,
+      "step": 1090
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.963360368075496e-05,
+      "loss": 0.7952,
+      "step": 1091
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9633267845448593e-05,
+      "loss": 0.3126,
+      "step": 1092
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9632932010142227e-05,
+      "loss": 0.6557,
+      "step": 1093
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9632596174835862e-05,
+      "loss": 0.2909,
+      "step": 1094
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9632260339529497e-05,
+      "loss": 0.4253,
+      "step": 1095
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.963192450422313e-05,
+      "loss": 0.6995,
+      "step": 1096
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9631588668916766e-05,
+      "loss": 0.3463,
+      "step": 1097
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.96312528336104e-05,
+      "loss": 0.3967,
+      "step": 1098
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9630916998304032e-05,
+      "loss": 0.6452,
+      "step": 1099
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.963058116299767e-05,
+      "loss": 0.344,
+      "step": 1100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.96302453276913e-05,
+      "loss": 0.3769,
+      "step": 1101
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9629909492384936e-05,
+      "loss": 0.3229,
+      "step": 1102
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.962957365707857e-05,
+      "loss": 0.2988,
+      "step": 1103
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9629237821772205e-05,
+      "loss": 0.3824,
+      "step": 1104
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.962890198646584e-05,
+      "loss": 0.2306,
+      "step": 1105
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9628566151159474e-05,
+      "loss": 0.3295,
+      "step": 1106
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9628230315853106e-05,
+      "loss": 0.4439,
+      "step": 1107
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.962789448054674e-05,
+      "loss": 0.3913,
+      "step": 1108
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9627558645240375e-05,
+      "loss": 0.3122,
+      "step": 1109
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.962722280993401e-05,
+      "loss": 0.2193,
+      "step": 1110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9626886974627644e-05,
+      "loss": 0.3643,
+      "step": 1111
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.962655113932128e-05,
+      "loss": 0.5298,
+      "step": 1112
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9626215304014914e-05,
+      "loss": 0.4337,
+      "step": 1113
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9625879468708545e-05,
+      "loss": 0.4094,
+      "step": 1114
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9625543633402183e-05,
+      "loss": 0.3414,
+      "step": 1115
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9625207798095814e-05,
+      "loss": 0.457,
+      "step": 1116
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.962487196278945e-05,
+      "loss": 0.4076,
+      "step": 1117
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9624536127483083e-05,
+      "loss": 0.4424,
+      "step": 1118
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9624200292176718e-05,
+      "loss": 0.4617,
+      "step": 1119
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9623864456870353e-05,
+      "loss": 0.5037,
+      "step": 1120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9623528621563987e-05,
+      "loss": 0.3834,
+      "step": 1121
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9623192786257622e-05,
+      "loss": 0.328,
+      "step": 1122
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9622856950951253e-05,
+      "loss": 0.4629,
+      "step": 1123
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.962252111564489e-05,
+      "loss": 0.5513,
+      "step": 1124
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9622185280338523e-05,
+      "loss": 0.3576,
+      "step": 1125
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9621849445032157e-05,
+      "loss": 0.3823,
+      "step": 1126
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9621513609725792e-05,
+      "loss": 0.4736,
+      "step": 1127
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9621177774419426e-05,
+      "loss": 0.486,
+      "step": 1128
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.962084193911306e-05,
+      "loss": 0.1929,
+      "step": 1129
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9620506103806696e-05,
+      "loss": 0.2777,
+      "step": 1130
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.962017026850033e-05,
+      "loss": 0.432,
+      "step": 1131
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.961983443319396e-05,
+      "loss": 0.3497,
+      "step": 1132
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.96194985978876e-05,
+      "loss": 0.3052,
+      "step": 1133
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.961916276258123e-05,
+      "loss": 0.3031,
+      "step": 1134
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9618826927274866e-05,
+      "loss": 0.6239,
+      "step": 1135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.96184910919685e-05,
+      "loss": 0.3695,
+      "step": 1136
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9618155256662135e-05,
+      "loss": 0.2586,
+      "step": 1137
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.961781942135577e-05,
+      "loss": 0.3022,
+      "step": 1138
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9617483586049404e-05,
+      "loss": 0.4841,
+      "step": 1139
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.961714775074304e-05,
+      "loss": 0.3265,
+      "step": 1140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.961681191543667e-05,
+      "loss": 0.2579,
+      "step": 1141
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9616476080130308e-05,
+      "loss": 0.4685,
+      "step": 1142
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.961614024482394e-05,
+      "loss": 0.4589,
+      "step": 1143
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9615804409517574e-05,
+      "loss": 0.3472,
+      "step": 1144
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.961546857421121e-05,
+      "loss": 0.303,
+      "step": 1145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9615132738904843e-05,
+      "loss": 0.3021,
+      "step": 1146
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9614796903598478e-05,
+      "loss": 0.4385,
+      "step": 1147
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9614461068292113e-05,
+      "loss": 0.2286,
+      "step": 1148
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9614125232985747e-05,
+      "loss": 0.5522,
+      "step": 1149
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.961378939767938e-05,
+      "loss": 0.4571,
+      "step": 1150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9613453562373013e-05,
+      "loss": 0.3984,
+      "step": 1151
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9613117727066648e-05,
+      "loss": 0.3619,
+      "step": 1152
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9612781891760283e-05,
+      "loss": 0.3765,
+      "step": 1153
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9612446056453917e-05,
+      "loss": 0.6574,
+      "step": 1154
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9612110221147552e-05,
+      "loss": 0.3288,
+      "step": 1155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9611774385841183e-05,
+      "loss": 0.3425,
+      "step": 1156
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.961143855053482e-05,
+      "loss": 0.367,
+      "step": 1157
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9611102715228452e-05,
+      "loss": 0.3459,
+      "step": 1158
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9610766879922087e-05,
+      "loss": 0.5281,
+      "step": 1159
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.961043104461572e-05,
+      "loss": 0.5126,
+      "step": 1160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9610095209309356e-05,
+      "loss": 0.3599,
+      "step": 1161
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960975937400299e-05,
+      "loss": 0.4203,
+      "step": 1162
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9609423538696626e-05,
+      "loss": 0.2739,
+      "step": 1163
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960908770339026e-05,
+      "loss": 0.4484,
+      "step": 1164
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960875186808389e-05,
+      "loss": 0.3343,
+      "step": 1165
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960841603277753e-05,
+      "loss": 0.3324,
+      "step": 1166
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960808019747116e-05,
+      "loss": 0.3346,
+      "step": 1167
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9607744362164795e-05,
+      "loss": 0.5517,
+      "step": 1168
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960740852685843e-05,
+      "loss": 0.358,
+      "step": 1169
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9607072691552065e-05,
+      "loss": 0.6357,
+      "step": 1170
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.96067368562457e-05,
+      "loss": 0.543,
+      "step": 1171
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9606401020939334e-05,
+      "loss": 0.2678,
+      "step": 1172
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960606518563297e-05,
+      "loss": 0.477,
+      "step": 1173
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.96057293503266e-05,
+      "loss": 0.4155,
+      "step": 1174
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9605393515020235e-05,
+      "loss": 0.4067,
+      "step": 1175
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960505767971387e-05,
+      "loss": 0.2996,
+      "step": 1176
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9604721844407504e-05,
+      "loss": 0.3478,
+      "step": 1177
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960438600910114e-05,
+      "loss": 0.2905,
+      "step": 1178
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9604050173794773e-05,
+      "loss": 0.6133,
+      "step": 1179
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9603714338488408e-05,
+      "loss": 0.4512,
+      "step": 1180
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960337850318204e-05,
+      "loss": 0.4207,
+      "step": 1181
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9603042667875677e-05,
+      "loss": 0.406,
+      "step": 1182
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960270683256931e-05,
+      "loss": 0.3635,
+      "step": 1183
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9602370997262943e-05,
+      "loss": 0.2771,
+      "step": 1184
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9602035161956578e-05,
+      "loss": 0.5677,
+      "step": 1185
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9601699326650212e-05,
+      "loss": 0.3869,
+      "step": 1186
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9601363491343847e-05,
+      "loss": 0.4132,
+      "step": 1187
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.960102765603748e-05,
+      "loss": 0.3646,
+      "step": 1188
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9600691820731116e-05,
+      "loss": 0.408,
+      "step": 1189
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9600355985424748e-05,
+      "loss": 0.3522,
+      "step": 1190
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9600020150118386e-05,
+      "loss": 0.5202,
+      "step": 1191
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9599684314812017e-05,
+      "loss": 0.5037,
+      "step": 1192
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959934847950565e-05,
+      "loss": 0.4356,
+      "step": 1193
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9599012644199286e-05,
+      "loss": 0.395,
+      "step": 1194
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959867680889292e-05,
+      "loss": 0.2675,
+      "step": 1195
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9598340973586555e-05,
+      "loss": 0.3648,
+      "step": 1196
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959800513828019e-05,
+      "loss": 0.4211,
+      "step": 1197
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959766930297382e-05,
+      "loss": 0.2999,
+      "step": 1198
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9597333467667456e-05,
+      "loss": 0.3661,
+      "step": 1199
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959699763236109e-05,
+      "loss": 0.2978,
+      "step": 1200
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9596661797054725e-05,
+      "loss": 0.4354,
+      "step": 1201
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959632596174836e-05,
+      "loss": 0.4854,
+      "step": 1202
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9595990126441995e-05,
+      "loss": 0.2896,
+      "step": 1203
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959565429113563e-05,
+      "loss": 0.3371,
+      "step": 1204
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959531845582926e-05,
+      "loss": 0.2383,
+      "step": 1205
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.95949826205229e-05,
+      "loss": 0.3412,
+      "step": 1206
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959464678521653e-05,
+      "loss": 0.3092,
+      "step": 1207
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9594310949910164e-05,
+      "loss": 0.3859,
+      "step": 1208
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.95939751146038e-05,
+      "loss": 0.2328,
+      "step": 1209
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9593639279297434e-05,
+      "loss": 0.2475,
+      "step": 1210
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959330344399107e-05,
+      "loss": 0.3078,
+      "step": 1211
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9592967608684703e-05,
+      "loss": 0.4891,
+      "step": 1212
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9592631773378338e-05,
+      "loss": 0.476,
+      "step": 1213
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.959229593807197e-05,
+      "loss": 0.4458,
+      "step": 1214
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9591960102765607e-05,
+      "loss": 0.4892,
+      "step": 1215
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9591624267459238e-05,
+      "loss": 0.2564,
+      "step": 1216
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9591288432152873e-05,
+      "loss": 0.3373,
+      "step": 1217
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9590952596846507e-05,
+      "loss": 0.3473,
+      "step": 1218
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9590616761540142e-05,
+      "loss": 0.3877,
+      "step": 1219
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9590280926233777e-05,
+      "loss": 0.3165,
+      "step": 1220
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.958994509092741e-05,
+      "loss": 0.3176,
+      "step": 1221
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9589609255621046e-05,
+      "loss": 0.3817,
+      "step": 1222
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9589273420314677e-05,
+      "loss": 0.3277,
+      "step": 1223
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9588937585008315e-05,
+      "loss": 0.3318,
+      "step": 1224
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9588601749701947e-05,
+      "loss": 0.4345,
+      "step": 1225
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.958826591439558e-05,
+      "loss": 0.4421,
+      "step": 1226
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9587930079089216e-05,
+      "loss": 0.3591,
+      "step": 1227
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.958759424378285e-05,
+      "loss": 0.3916,
+      "step": 1228
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9587258408476485e-05,
+      "loss": 0.2881,
+      "step": 1229
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.958692257317012e-05,
+      "loss": 0.5243,
+      "step": 1230
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9586586737863755e-05,
+      "loss": 0.4449,
+      "step": 1231
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9586250902557386e-05,
+      "loss": 0.3933,
+      "step": 1232
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9585915067251024e-05,
+      "loss": 0.4169,
+      "step": 1233
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9585579231944655e-05,
+      "loss": 0.5204,
+      "step": 1234
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.958524339663829e-05,
+      "loss": 0.5118,
+      "step": 1235
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9584907561331924e-05,
+      "loss": 0.2896,
+      "step": 1236
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.958457172602556e-05,
+      "loss": 0.3798,
+      "step": 1237
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9584235890719194e-05,
+      "loss": 0.4151,
+      "step": 1238
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9583900055412828e-05,
+      "loss": 0.4548,
+      "step": 1239
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9583564220106463e-05,
+      "loss": 0.3189,
+      "step": 1240
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9583228384800094e-05,
+      "loss": 0.3027,
+      "step": 1241
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.958289254949373e-05,
+      "loss": 0.324,
+      "step": 1242
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9582556714187364e-05,
+      "loss": 0.3648,
+      "step": 1243
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9582220878880998e-05,
+      "loss": 0.3781,
+      "step": 1244
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9581885043574633e-05,
+      "loss": 0.4575,
+      "step": 1245
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9581549208268267e-05,
+      "loss": 0.3651,
+      "step": 1246
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.95812133729619e-05,
+      "loss": 0.3003,
+      "step": 1247
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9580877537655537e-05,
+      "loss": 0.2861,
+      "step": 1248
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9580541702349168e-05,
+      "loss": 0.535,
+      "step": 1249
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9580205867042803e-05,
+      "loss": 0.3376,
+      "step": 1250
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9579870031736437e-05,
+      "loss": 0.2797,
+      "step": 1251
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9579534196430072e-05,
+      "loss": 0.2668,
+      "step": 1252
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9579198361123707e-05,
+      "loss": 0.3367,
+      "step": 1253
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.957886252581734e-05,
+      "loss": 0.2722,
+      "step": 1254
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9578526690510976e-05,
+      "loss": 0.413,
+      "step": 1255
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9578190855204607e-05,
+      "loss": 0.1891,
+      "step": 1256
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9577855019898245e-05,
+      "loss": 0.5099,
+      "step": 1257
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9577519184591876e-05,
+      "loss": 0.3627,
+      "step": 1258
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.957718334928551e-05,
+      "loss": 0.2752,
+      "step": 1259
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9576847513979146e-05,
+      "loss": 0.2573,
+      "step": 1260
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.957651167867278e-05,
+      "loss": 0.3072,
+      "step": 1261
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9576175843366415e-05,
+      "loss": 0.2414,
+      "step": 1262
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.957584000806005e-05,
+      "loss": 0.5519,
+      "step": 1263
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9575504172753684e-05,
+      "loss": 0.2228,
+      "step": 1264
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9575168337447316e-05,
+      "loss": 0.3701,
+      "step": 1265
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9574832502140954e-05,
+      "loss": 0.3929,
+      "step": 1266
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9574496666834585e-05,
+      "loss": 0.3846,
+      "step": 1267
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.957416083152822e-05,
+      "loss": 0.3424,
+      "step": 1268
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9573824996221854e-05,
+      "loss": 0.3785,
+      "step": 1269
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.957348916091549e-05,
+      "loss": 0.3254,
+      "step": 1270
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9573153325609123e-05,
+      "loss": 0.4348,
+      "step": 1271
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9572817490302758e-05,
+      "loss": 0.3152,
+      "step": 1272
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9572481654996393e-05,
+      "loss": 0.2482,
+      "step": 1273
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9572145819690024e-05,
+      "loss": 0.6632,
+      "step": 1274
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9571809984383662e-05,
+      "loss": 0.4322,
+      "step": 1275
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9571474149077293e-05,
+      "loss": 0.5323,
+      "step": 1276
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9571138313770928e-05,
+      "loss": 0.655,
+      "step": 1277
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9570802478464563e-05,
+      "loss": 0.3191,
+      "step": 1278
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9570466643158197e-05,
+      "loss": 0.5164,
+      "step": 1279
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9570130807851832e-05,
+      "loss": 0.232,
+      "step": 1280
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9569794972545467e-05,
+      "loss": 0.4742,
+      "step": 1281
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.95694591372391e-05,
+      "loss": 0.3401,
+      "step": 1282
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9569123301932732e-05,
+      "loss": 0.2835,
+      "step": 1283
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.956878746662637e-05,
+      "loss": 0.4067,
+      "step": 1284
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9568451631320002e-05,
+      "loss": 0.3725,
+      "step": 1285
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9568115796013636e-05,
+      "loss": 0.6913,
+      "step": 1286
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.956777996070727e-05,
+      "loss": 0.2709,
+      "step": 1287
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9567444125400906e-05,
+      "loss": 0.4514,
+      "step": 1288
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9567108290094537e-05,
+      "loss": 0.4058,
+      "step": 1289
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9566772454788175e-05,
+      "loss": 0.3083,
+      "step": 1290
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9566436619481806e-05,
+      "loss": 0.2242,
+      "step": 1291
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.956610078417544e-05,
+      "loss": 0.3959,
+      "step": 1292
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9565764948869076e-05,
+      "loss": 0.2645,
+      "step": 1293
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.956542911356271e-05,
+      "loss": 0.3991,
+      "step": 1294
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9565093278256345e-05,
+      "loss": 0.4166,
+      "step": 1295
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.956475744294998e-05,
+      "loss": 0.3809,
+      "step": 1296
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9564421607643614e-05,
+      "loss": 0.3097,
+      "step": 1297
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9564085772337245e-05,
+      "loss": 0.2723,
+      "step": 1298
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9563749937030883e-05,
+      "loss": 0.2756,
+      "step": 1299
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9563414101724515e-05,
+      "loss": 0.4379,
+      "step": 1300
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.956307826641815e-05,
+      "loss": 0.4765,
+      "step": 1301
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9562742431111784e-05,
+      "loss": 0.5714,
+      "step": 1302
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.956240659580542e-05,
+      "loss": 0.5134,
+      "step": 1303
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9562070760499053e-05,
+      "loss": 0.3947,
+      "step": 1304
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9561734925192688e-05,
+      "loss": 0.3384,
+      "step": 1305
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9561399089886323e-05,
+      "loss": 0.2469,
+      "step": 1306
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9561063254579954e-05,
+      "loss": 0.4115,
+      "step": 1307
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9560727419273592e-05,
+      "loss": 0.3153,
+      "step": 1308
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9560391583967223e-05,
+      "loss": 0.2941,
+      "step": 1309
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9560055748660858e-05,
+      "loss": 0.4688,
+      "step": 1310
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9559719913354492e-05,
+      "loss": 0.5334,
+      "step": 1311
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9559384078048127e-05,
+      "loss": 0.2742,
+      "step": 1312
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9559048242741762e-05,
+      "loss": 0.5362,
+      "step": 1313
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9558712407435396e-05,
+      "loss": 0.3609,
+      "step": 1314
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955837657212903e-05,
+      "loss": 0.4087,
+      "step": 1315
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9558040736822662e-05,
+      "loss": 0.2979,
+      "step": 1316
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.95577049015163e-05,
+      "loss": 0.2679,
+      "step": 1317
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955736906620993e-05,
+      "loss": 0.3051,
+      "step": 1318
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9557033230903566e-05,
+      "loss": 0.34,
+      "step": 1319
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.95566973955972e-05,
+      "loss": 0.2288,
+      "step": 1320
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9556361560290836e-05,
+      "loss": 0.3867,
+      "step": 1321
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955602572498447e-05,
+      "loss": 0.2775,
+      "step": 1322
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9555689889678105e-05,
+      "loss": 0.3247,
+      "step": 1323
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955535405437174e-05,
+      "loss": 0.3857,
+      "step": 1324
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955501821906537e-05,
+      "loss": 0.5461,
+      "step": 1325
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955468238375901e-05,
+      "loss": 0.4155,
+      "step": 1326
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955434654845264e-05,
+      "loss": 0.6752,
+      "step": 1327
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9554010713146275e-05,
+      "loss": 0.2698,
+      "step": 1328
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955367487783991e-05,
+      "loss": 0.3971,
+      "step": 1329
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9553339042533544e-05,
+      "loss": 0.3487,
+      "step": 1330
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955300320722718e-05,
+      "loss": 0.2183,
+      "step": 1331
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9552667371920813e-05,
+      "loss": 0.3848,
+      "step": 1332
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9552331536614444e-05,
+      "loss": 0.473,
+      "step": 1333
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955199570130808e-05,
+      "loss": 0.3262,
+      "step": 1334
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9551659866001714e-05,
+      "loss": 0.4952,
+      "step": 1335
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.955132403069535e-05,
+      "loss": 0.5086,
+      "step": 1336
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9550988195388983e-05,
+      "loss": 0.3818,
+      "step": 1337
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9550652360082618e-05,
+      "loss": 0.3539,
+      "step": 1338
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9550316524776252e-05,
+      "loss": 0.4153,
+      "step": 1339
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9549980689469884e-05,
+      "loss": 0.4034,
+      "step": 1340
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954964485416352e-05,
+      "loss": 0.2879,
+      "step": 1341
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9549309018857153e-05,
+      "loss": 0.3939,
+      "step": 1342
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9548973183550788e-05,
+      "loss": 0.4192,
+      "step": 1343
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9548637348244422e-05,
+      "loss": 0.3226,
+      "step": 1344
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9548301512938057e-05,
+      "loss": 0.3159,
+      "step": 1345
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954796567763169e-05,
+      "loss": 0.3742,
+      "step": 1346
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9547629842325326e-05,
+      "loss": 0.2816,
+      "step": 1347
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954729400701896e-05,
+      "loss": 0.3842,
+      "step": 1348
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9546958171712592e-05,
+      "loss": 0.2976,
+      "step": 1349
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954662233640623e-05,
+      "loss": 0.4247,
+      "step": 1350
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954628650109986e-05,
+      "loss": 0.2933,
+      "step": 1351
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9545950665793496e-05,
+      "loss": 0.4571,
+      "step": 1352
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954561483048713e-05,
+      "loss": 0.2397,
+      "step": 1353
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9545278995180765e-05,
+      "loss": 0.5192,
+      "step": 1354
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.95449431598744e-05,
+      "loss": 0.4667,
+      "step": 1355
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954460732456803e-05,
+      "loss": 0.3659,
+      "step": 1356
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954427148926167e-05,
+      "loss": 0.5818,
+      "step": 1357
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.95439356539553e-05,
+      "loss": 0.5429,
+      "step": 1358
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9543599818648935e-05,
+      "loss": 0.3208,
+      "step": 1359
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954326398334257e-05,
+      "loss": 0.4856,
+      "step": 1360
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9542928148036204e-05,
+      "loss": 0.4733,
+      "step": 1361
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954259231272984e-05,
+      "loss": 0.2111,
+      "step": 1362
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9542256477423474e-05,
+      "loss": 0.5558,
+      "step": 1363
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954192064211711e-05,
+      "loss": 0.4104,
+      "step": 1364
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954158480681074e-05,
+      "loss": 0.6122,
+      "step": 1365
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9541248971504378e-05,
+      "loss": 0.3396,
+      "step": 1366
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.954091313619801e-05,
+      "loss": 0.3245,
+      "step": 1367
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9540577300891644e-05,
+      "loss": 0.4097,
+      "step": 1368
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9540241465585278e-05,
+      "loss": 0.3497,
+      "step": 1369
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9539905630278913e-05,
+      "loss": 0.3572,
+      "step": 1370
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9539569794972548e-05,
+      "loss": 0.2597,
+      "step": 1371
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9539233959666182e-05,
+      "loss": 0.379,
+      "step": 1372
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9538898124359817e-05,
+      "loss": 0.4181,
+      "step": 1373
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9538562289053448e-05,
+      "loss": 0.3169,
+      "step": 1374
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9538226453747086e-05,
+      "loss": 0.6583,
+      "step": 1375
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9537890618440717e-05,
+      "loss": 0.4702,
+      "step": 1376
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9537554783134352e-05,
+      "loss": 0.3082,
+      "step": 1377
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9537218947827987e-05,
+      "loss": 0.2441,
+      "step": 1378
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.953688311252162e-05,
+      "loss": 0.275,
+      "step": 1379
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9536547277215253e-05,
+      "loss": 0.3556,
+      "step": 1380
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.953621144190889e-05,
+      "loss": 0.3099,
+      "step": 1381
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9535875606602522e-05,
+      "loss": 0.4424,
+      "step": 1382
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9535539771296157e-05,
+      "loss": 0.2821,
+      "step": 1383
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.953520393598979e-05,
+      "loss": 0.454,
+      "step": 1384
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9534868100683426e-05,
+      "loss": 0.4442,
+      "step": 1385
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.953453226537706e-05,
+      "loss": 0.3854,
+      "step": 1386
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9534196430070695e-05,
+      "loss": 0.4872,
+      "step": 1387
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.953386059476433e-05,
+      "loss": 0.2777,
+      "step": 1388
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.953352475945796e-05,
+      "loss": 0.2576,
+      "step": 1389
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.95331889241516e-05,
+      "loss": 0.3082,
+      "step": 1390
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.953285308884523e-05,
+      "loss": 0.4404,
+      "step": 1391
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9532517253538865e-05,
+      "loss": 0.2779,
+      "step": 1392
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.95321814182325e-05,
+      "loss": 0.3621,
+      "step": 1393
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9531845582926134e-05,
+      "loss": 0.4294,
+      "step": 1394
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.953150974761977e-05,
+      "loss": 0.2604,
+      "step": 1395
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9531173912313404e-05,
+      "loss": 0.3287,
+      "step": 1396
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9530838077007038e-05,
+      "loss": 0.3389,
+      "step": 1397
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.953050224170067e-05,
+      "loss": 0.2869,
+      "step": 1398
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9530166406394308e-05,
+      "loss": 0.2569,
+      "step": 1399
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.952983057108794e-05,
+      "loss": 0.3683,
+      "step": 1400
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9529494735781573e-05,
+      "loss": 0.3916,
+      "step": 1401
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9529158900475208e-05,
+      "loss": 0.4959,
+      "step": 1402
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9528823065168843e-05,
+      "loss": 0.2079,
+      "step": 1403
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9528487229862477e-05,
+      "loss": 0.6073,
+      "step": 1404
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9528151394556112e-05,
+      "loss": 0.2759,
+      "step": 1405
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9527815559249747e-05,
+      "loss": 0.2665,
+      "step": 1406
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9527479723943378e-05,
+      "loss": 0.4775,
+      "step": 1407
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9527143888637016e-05,
+      "loss": 0.4091,
+      "step": 1408
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9526808053330647e-05,
+      "loss": 0.457,
+      "step": 1409
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9526472218024282e-05,
+      "loss": 0.489,
+      "step": 1410
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9526136382717917e-05,
+      "loss": 0.4975,
+      "step": 1411
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.952580054741155e-05,
+      "loss": 0.3576,
+      "step": 1412
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9525464712105186e-05,
+      "loss": 0.5959,
+      "step": 1413
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.952512887679882e-05,
+      "loss": 0.4649,
+      "step": 1414
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9524793041492455e-05,
+      "loss": 0.2884,
+      "step": 1415
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9524457206186086e-05,
+      "loss": 0.3629,
+      "step": 1416
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9524121370879724e-05,
+      "loss": 0.4472,
+      "step": 1417
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9523785535573356e-05,
+      "loss": 0.3436,
+      "step": 1418
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.952344970026699e-05,
+      "loss": 0.4661,
+      "step": 1419
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9523113864960625e-05,
+      "loss": 0.3105,
+      "step": 1420
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.952277802965426e-05,
+      "loss": 0.4046,
+      "step": 1421
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9522442194347894e-05,
+      "loss": 0.315,
+      "step": 1422
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.952210635904153e-05,
+      "loss": 0.2932,
+      "step": 1423
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.952177052373516e-05,
+      "loss": 0.3435,
+      "step": 1424
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9521434688428795e-05,
+      "loss": 0.2539,
+      "step": 1425
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.952109885312243e-05,
+      "loss": 0.4558,
+      "step": 1426
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9520763017816064e-05,
+      "loss": 0.303,
+      "step": 1427
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.95204271825097e-05,
+      "loss": 0.3135,
+      "step": 1428
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9520091347203333e-05,
+      "loss": 0.4238,
+      "step": 1429
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9519755511896968e-05,
+      "loss": 0.3122,
+      "step": 1430
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.95194196765906e-05,
+      "loss": 0.7233,
+      "step": 1431
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9519083841284237e-05,
+      "loss": 0.4935,
+      "step": 1432
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.951874800597787e-05,
+      "loss": 0.2748,
+      "step": 1433
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9518412170671503e-05,
+      "loss": 0.3269,
+      "step": 1434
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9518076335365138e-05,
+      "loss": 0.5718,
+      "step": 1435
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9517740500058773e-05,
+      "loss": 0.3796,
+      "step": 1436
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9517404664752407e-05,
+      "loss": 0.3672,
+      "step": 1437
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9517068829446042e-05,
+      "loss": 0.4821,
+      "step": 1438
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9516732994139676e-05,
+      "loss": 0.4213,
+      "step": 1439
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9516397158833308e-05,
+      "loss": 0.3377,
+      "step": 1440
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9516061323526946e-05,
+      "loss": 0.5676,
+      "step": 1441
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9515725488220577e-05,
+      "loss": 0.4174,
+      "step": 1442
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.951538965291421e-05,
+      "loss": 0.1747,
+      "step": 1443
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9515053817607846e-05,
+      "loss": 0.2715,
+      "step": 1444
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.951471798230148e-05,
+      "loss": 0.4277,
+      "step": 1445
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9514382146995116e-05,
+      "loss": 0.3072,
+      "step": 1446
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.951404631168875e-05,
+      "loss": 0.3671,
+      "step": 1447
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9513710476382385e-05,
+      "loss": 0.5332,
+      "step": 1448
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9513374641076016e-05,
+      "loss": 0.5041,
+      "step": 1449
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9513038805769654e-05,
+      "loss": 0.3744,
+      "step": 1450
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9512702970463285e-05,
+      "loss": 0.4214,
+      "step": 1451
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.951236713515692e-05,
+      "loss": 0.2839,
+      "step": 1452
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9512031299850555e-05,
+      "loss": 0.4457,
+      "step": 1453
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.951169546454419e-05,
+      "loss": 0.2765,
+      "step": 1454
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9511359629237824e-05,
+      "loss": 0.5396,
+      "step": 1455
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.951102379393146e-05,
+      "loss": 0.4686,
+      "step": 1456
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9510687958625093e-05,
+      "loss": 0.4611,
+      "step": 1457
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9510352123318725e-05,
+      "loss": 0.3777,
+      "step": 1458
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9510016288012363e-05,
+      "loss": 0.3054,
+      "step": 1459
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9509680452705994e-05,
+      "loss": 0.2683,
+      "step": 1460
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.950934461739963e-05,
+      "loss": 0.3216,
+      "step": 1461
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9509008782093263e-05,
+      "loss": 0.4217,
+      "step": 1462
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9508672946786898e-05,
+      "loss": 0.4213,
+      "step": 1463
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9508337111480532e-05,
+      "loss": 0.3142,
+      "step": 1464
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9508001276174167e-05,
+      "loss": 0.3352,
+      "step": 1465
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9507665440867802e-05,
+      "loss": 0.3274,
+      "step": 1466
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9507329605561433e-05,
+      "loss": 0.3162,
+      "step": 1467
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9506993770255068e-05,
+      "loss": 0.2726,
+      "step": 1468
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9506657934948702e-05,
+      "loss": 0.338,
+      "step": 1469
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9506322099642337e-05,
+      "loss": 0.291,
+      "step": 1470
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.950598626433597e-05,
+      "loss": 0.2511,
+      "step": 1471
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9505650429029606e-05,
+      "loss": 0.2254,
+      "step": 1472
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9505314593723238e-05,
+      "loss": 0.4726,
+      "step": 1473
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9504978758416876e-05,
+      "loss": 0.731,
+      "step": 1474
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9504642923110507e-05,
+      "loss": 0.4447,
+      "step": 1475
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.950430708780414e-05,
+      "loss": 0.2866,
+      "step": 1476
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9503971252497776e-05,
+      "loss": 0.6976,
+      "step": 1477
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.950363541719141e-05,
+      "loss": 0.4307,
+      "step": 1478
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9503299581885045e-05,
+      "loss": 0.4558,
+      "step": 1479
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.950296374657868e-05,
+      "loss": 0.3766,
+      "step": 1480
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9502627911272315e-05,
+      "loss": 0.3459,
+      "step": 1481
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9502292075965946e-05,
+      "loss": 0.3299,
+      "step": 1482
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9501956240659584e-05,
+      "loss": 0.4473,
+      "step": 1483
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9501620405353215e-05,
+      "loss": 0.4198,
+      "step": 1484
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.950128457004685e-05,
+      "loss": 0.5191,
+      "step": 1485
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9500948734740485e-05,
+      "loss": 0.318,
+      "step": 1486
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.950061289943412e-05,
+      "loss": 0.4234,
+      "step": 1487
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9500277064127754e-05,
+      "loss": 0.2133,
+      "step": 1488
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949994122882139e-05,
+      "loss": 0.3708,
+      "step": 1489
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9499605393515023e-05,
+      "loss": 0.4112,
+      "step": 1490
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9499269558208654e-05,
+      "loss": 0.4398,
+      "step": 1491
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9498933722902292e-05,
+      "loss": 0.3534,
+      "step": 1492
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9498597887595924e-05,
+      "loss": 0.3394,
+      "step": 1493
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949826205228956e-05,
+      "loss": 0.4761,
+      "step": 1494
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9497926216983193e-05,
+      "loss": 0.4186,
+      "step": 1495
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9497590381676828e-05,
+      "loss": 0.3218,
+      "step": 1496
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9497254546370462e-05,
+      "loss": 0.3772,
+      "step": 1497
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9496918711064097e-05,
+      "loss": 0.3595,
+      "step": 1498
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949658287575773e-05,
+      "loss": 0.4745,
+      "step": 1499
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9496247040451363e-05,
+      "loss": 0.3078,
+      "step": 1500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9495911205145e-05,
+      "loss": 0.4305,
+      "step": 1501
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9495575369838632e-05,
+      "loss": 0.2965,
+      "step": 1502
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9495239534532267e-05,
+      "loss": 0.4214,
+      "step": 1503
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.94949036992259e-05,
+      "loss": 0.4713,
+      "step": 1504
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9494567863919536e-05,
+      "loss": 0.4044,
+      "step": 1505
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949423202861317e-05,
+      "loss": 0.3255,
+      "step": 1506
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9493896193306805e-05,
+      "loss": 0.3595,
+      "step": 1507
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949356035800044e-05,
+      "loss": 0.6035,
+      "step": 1508
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949322452269407e-05,
+      "loss": 0.3296,
+      "step": 1509
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949288868738771e-05,
+      "loss": 0.2983,
+      "step": 1510
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949255285208134e-05,
+      "loss": 0.3542,
+      "step": 1511
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9492217016774975e-05,
+      "loss": 0.4118,
+      "step": 1512
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949188118146861e-05,
+      "loss": 0.5502,
+      "step": 1513
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9491545346162245e-05,
+      "loss": 0.4157,
+      "step": 1514
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9491209510855876e-05,
+      "loss": 0.3358,
+      "step": 1515
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9490873675549514e-05,
+      "loss": 0.4911,
+      "step": 1516
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9490537840243145e-05,
+      "loss": 0.4049,
+      "step": 1517
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.949020200493678e-05,
+      "loss": 0.4838,
+      "step": 1518
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9489866169630414e-05,
+      "loss": 0.4484,
+      "step": 1519
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.948953033432405e-05,
+      "loss": 0.4341,
+      "step": 1520
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9489194499017684e-05,
+      "loss": 0.3046,
+      "step": 1521
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.948885866371132e-05,
+      "loss": 0.446,
+      "step": 1522
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9488522828404953e-05,
+      "loss": 0.5028,
+      "step": 1523
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9488186993098584e-05,
+      "loss": 0.4059,
+      "step": 1524
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9487851157792222e-05,
+      "loss": 0.3048,
+      "step": 1525
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9487515322485854e-05,
+      "loss": 0.36,
+      "step": 1526
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9487179487179488e-05,
+      "loss": 0.4221,
+      "step": 1527
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9486843651873123e-05,
+      "loss": 0.4154,
+      "step": 1528
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9486507816566757e-05,
+      "loss": 0.3226,
+      "step": 1529
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9486171981260392e-05,
+      "loss": 0.2011,
+      "step": 1530
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9485836145954027e-05,
+      "loss": 0.3651,
+      "step": 1531
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.948550031064766e-05,
+      "loss": 0.2622,
+      "step": 1532
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9485164475341293e-05,
+      "loss": 0.3193,
+      "step": 1533
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9484828640034927e-05,
+      "loss": 0.3969,
+      "step": 1534
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9484492804728562e-05,
+      "loss": 0.2731,
+      "step": 1535
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9484156969422197e-05,
+      "loss": 0.2941,
+      "step": 1536
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.948382113411583e-05,
+      "loss": 0.4936,
+      "step": 1537
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9483485298809466e-05,
+      "loss": 0.1452,
+      "step": 1538
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.94831494635031e-05,
+      "loss": 0.4985,
+      "step": 1539
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9482813628196732e-05,
+      "loss": 0.3539,
+      "step": 1540
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.948247779289037e-05,
+      "loss": 0.4311,
+      "step": 1541
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9482141957584e-05,
+      "loss": 0.3093,
+      "step": 1542
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9481806122277636e-05,
+      "loss": 0.4743,
+      "step": 1543
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.948147028697127e-05,
+      "loss": 0.4264,
+      "step": 1544
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9481134451664905e-05,
+      "loss": 0.3093,
+      "step": 1545
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.948079861635854e-05,
+      "loss": 0.2028,
+      "step": 1546
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9480462781052174e-05,
+      "loss": 0.3939,
+      "step": 1547
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.948012694574581e-05,
+      "loss": 0.3776,
+      "step": 1548
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947979111043944e-05,
+      "loss": 0.3031,
+      "step": 1549
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9479455275133078e-05,
+      "loss": 0.3222,
+      "step": 1550
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947911943982671e-05,
+      "loss": 0.4656,
+      "step": 1551
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9478783604520344e-05,
+      "loss": 0.39,
+      "step": 1552
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947844776921398e-05,
+      "loss": 0.3822,
+      "step": 1553
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9478111933907613e-05,
+      "loss": 0.4393,
+      "step": 1554
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9477776098601248e-05,
+      "loss": 0.3369,
+      "step": 1555
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9477440263294883e-05,
+      "loss": 0.3214,
+      "step": 1556
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9477104427988517e-05,
+      "loss": 0.308,
+      "step": 1557
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947676859268215e-05,
+      "loss": 0.3266,
+      "step": 1558
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9476432757375783e-05,
+      "loss": 0.288,
+      "step": 1559
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9476096922069418e-05,
+      "loss": 0.3323,
+      "step": 1560
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9475761086763053e-05,
+      "loss": 0.5123,
+      "step": 1561
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9475425251456687e-05,
+      "loss": 0.2499,
+      "step": 1562
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9475089416150322e-05,
+      "loss": 0.2875,
+      "step": 1563
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9474753580843953e-05,
+      "loss": 0.2727,
+      "step": 1564
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947441774553759e-05,
+      "loss": 0.3819,
+      "step": 1565
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9474081910231222e-05,
+      "loss": 0.4077,
+      "step": 1566
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9473746074924857e-05,
+      "loss": 0.4838,
+      "step": 1567
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9473410239618492e-05,
+      "loss": 0.2699,
+      "step": 1568
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9473074404312126e-05,
+      "loss": 0.5385,
+      "step": 1569
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947273856900576e-05,
+      "loss": 0.301,
+      "step": 1570
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9472402733699396e-05,
+      "loss": 0.4397,
+      "step": 1571
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947206689839303e-05,
+      "loss": 0.4409,
+      "step": 1572
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947173106308666e-05,
+      "loss": 0.4568,
+      "step": 1573
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.94713952277803e-05,
+      "loss": 0.3127,
+      "step": 1574
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.947105939247393e-05,
+      "loss": 0.3061,
+      "step": 1575
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9470723557167566e-05,
+      "loss": 0.2875,
+      "step": 1576
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.94703877218612e-05,
+      "loss": 0.2602,
+      "step": 1577
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9470051886554835e-05,
+      "loss": 0.3385,
+      "step": 1578
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.946971605124847e-05,
+      "loss": 0.2587,
+      "step": 1579
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9469380215942104e-05,
+      "loss": 0.3298,
+      "step": 1580
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.946904438063574e-05,
+      "loss": 0.3192,
+      "step": 1581
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.946870854532937e-05,
+      "loss": 0.3927,
+      "step": 1582
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9468372710023008e-05,
+      "loss": 0.412,
+      "step": 1583
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.946803687471664e-05,
+      "loss": 0.4115,
+      "step": 1584
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9467701039410274e-05,
+      "loss": 0.5877,
+      "step": 1585
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.946736520410391e-05,
+      "loss": 0.346,
+      "step": 1586
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9467029368797543e-05,
+      "loss": 0.3701,
+      "step": 1587
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9466693533491178e-05,
+      "loss": 0.4208,
+      "step": 1588
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9466357698184813e-05,
+      "loss": 0.3624,
+      "step": 1589
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9466021862878447e-05,
+      "loss": 0.3456,
+      "step": 1590
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.946568602757208e-05,
+      "loss": 0.4675,
+      "step": 1591
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9465350192265717e-05,
+      "loss": 0.2836,
+      "step": 1592
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9465014356959348e-05,
+      "loss": 0.3601,
+      "step": 1593
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9464678521652982e-05,
+      "loss": 0.491,
+      "step": 1594
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9464342686346617e-05,
+      "loss": 0.401,
+      "step": 1595
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9464006851040252e-05,
+      "loss": 0.5152,
+      "step": 1596
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9463671015733886e-05,
+      "loss": 0.2546,
+      "step": 1597
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.946333518042752e-05,
+      "loss": 0.5156,
+      "step": 1598
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9462999345121156e-05,
+      "loss": 0.2787,
+      "step": 1599
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9462663509814787e-05,
+      "loss": 0.4007,
+      "step": 1600
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9462327674508425e-05,
+      "loss": 0.3208,
+      "step": 1601
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9461991839202056e-05,
+      "loss": 0.4241,
+      "step": 1602
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.946165600389569e-05,
+      "loss": 0.3111,
+      "step": 1603
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9461320168589326e-05,
+      "loss": 0.2906,
+      "step": 1604
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.946098433328296e-05,
+      "loss": 0.3772,
+      "step": 1605
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.946064849797659e-05,
+      "loss": 0.2876,
+      "step": 1606
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.946031266267023e-05,
+      "loss": 0.3741,
+      "step": 1607
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.945997682736386e-05,
+      "loss": 0.4814,
+      "step": 1608
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9459640992057495e-05,
+      "loss": 0.3303,
+      "step": 1609
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.945930515675113e-05,
+      "loss": 0.4452,
+      "step": 1610
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9458969321444765e-05,
+      "loss": 0.4309,
+      "step": 1611
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.94586334861384e-05,
+      "loss": 0.4779,
+      "step": 1612
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9458297650832034e-05,
+      "loss": 0.3309,
+      "step": 1613
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.945796181552567e-05,
+      "loss": 0.3197,
+      "step": 1614
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.94576259802193e-05,
+      "loss": 0.2283,
+      "step": 1615
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9457290144912938e-05,
+      "loss": 0.2238,
+      "step": 1616
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.945695430960657e-05,
+      "loss": 0.4845,
+      "step": 1617
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9456618474300204e-05,
+      "loss": 0.2553,
+      "step": 1618
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.945628263899384e-05,
+      "loss": 0.2743,
+      "step": 1619
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9455946803687473e-05,
+      "loss": 0.4088,
+      "step": 1620
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9455610968381108e-05,
+      "loss": 0.5189,
+      "step": 1621
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9455275133074742e-05,
+      "loss": 0.2325,
+      "step": 1622
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9454939297768377e-05,
+      "loss": 0.3866,
+      "step": 1623
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9454603462462008e-05,
+      "loss": 0.3999,
+      "step": 1624
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9454267627155646e-05,
+      "loss": 0.2808,
+      "step": 1625
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9453931791849278e-05,
+      "loss": 0.3892,
+      "step": 1626
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9453595956542912e-05,
+      "loss": 0.2826,
+      "step": 1627
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9453260121236547e-05,
+      "loss": 0.3278,
+      "step": 1628
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.945292428593018e-05,
+      "loss": 0.2347,
+      "step": 1629
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9452588450623816e-05,
+      "loss": 0.327,
+      "step": 1630
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.945225261531745e-05,
+      "loss": 0.4519,
+      "step": 1631
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9451916780011085e-05,
+      "loss": 0.1872,
+      "step": 1632
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9451580944704717e-05,
+      "loss": 0.2942,
+      "step": 1633
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9451245109398355e-05,
+      "loss": 0.4823,
+      "step": 1634
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9450909274091986e-05,
+      "loss": 0.4547,
+      "step": 1635
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.945057343878562e-05,
+      "loss": 0.3748,
+      "step": 1636
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9450237603479255e-05,
+      "loss": 0.4873,
+      "step": 1637
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.944990176817289e-05,
+      "loss": 0.3716,
+      "step": 1638
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9449565932866525e-05,
+      "loss": 0.2415,
+      "step": 1639
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.944923009756016e-05,
+      "loss": 0.3445,
+      "step": 1640
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9448894262253794e-05,
+      "loss": 0.5012,
+      "step": 1641
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9448558426947425e-05,
+      "loss": 0.4615,
+      "step": 1642
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9448222591641063e-05,
+      "loss": 0.414,
+      "step": 1643
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9447886756334694e-05,
+      "loss": 0.5008,
+      "step": 1644
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.944755092102833e-05,
+      "loss": 0.3349,
+      "step": 1645
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9447215085721964e-05,
+      "loss": 0.4043,
+      "step": 1646
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.94468792504156e-05,
+      "loss": 0.3514,
+      "step": 1647
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9446543415109233e-05,
+      "loss": 0.4337,
+      "step": 1648
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9446207579802868e-05,
+      "loss": 0.4854,
+      "step": 1649
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.94458717444965e-05,
+      "loss": 0.3962,
+      "step": 1650
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9445535909190134e-05,
+      "loss": 0.3674,
+      "step": 1651
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9445200073883768e-05,
+      "loss": 0.4536,
+      "step": 1652
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9444864238577403e-05,
+      "loss": 0.4156,
+      "step": 1653
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9444528403271038e-05,
+      "loss": 0.2452,
+      "step": 1654
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9444192567964672e-05,
+      "loss": 0.3204,
+      "step": 1655
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9443856732658307e-05,
+      "loss": 0.3357,
+      "step": 1656
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9443520897351938e-05,
+      "loss": 0.2944,
+      "step": 1657
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9443185062045576e-05,
+      "loss": 0.3159,
+      "step": 1658
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9442849226739207e-05,
+      "loss": 0.3132,
+      "step": 1659
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9442513391432842e-05,
+      "loss": 0.535,
+      "step": 1660
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9442177556126477e-05,
+      "loss": 0.5247,
+      "step": 1661
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.944184172082011e-05,
+      "loss": 0.3131,
+      "step": 1662
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9441505885513746e-05,
+      "loss": 0.3968,
+      "step": 1663
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.944117005020738e-05,
+      "loss": 0.4046,
+      "step": 1664
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9440834214901015e-05,
+      "loss": 0.3018,
+      "step": 1665
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9440498379594647e-05,
+      "loss": 0.4198,
+      "step": 1666
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9440162544288285e-05,
+      "loss": 0.3106,
+      "step": 1667
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9439826708981916e-05,
+      "loss": 0.3341,
+      "step": 1668
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.943949087367555e-05,
+      "loss": 0.3022,
+      "step": 1669
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9439155038369185e-05,
+      "loss": 0.2628,
+      "step": 1670
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.943881920306282e-05,
+      "loss": 0.3412,
+      "step": 1671
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9438483367756454e-05,
+      "loss": 0.4905,
+      "step": 1672
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.943814753245009e-05,
+      "loss": 0.2361,
+      "step": 1673
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9437811697143724e-05,
+      "loss": 0.3745,
+      "step": 1674
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9437475861837355e-05,
+      "loss": 0.4447,
+      "step": 1675
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9437140026530993e-05,
+      "loss": 0.3293,
+      "step": 1676
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9436804191224624e-05,
+      "loss": 0.4163,
+      "step": 1677
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.943646835591826e-05,
+      "loss": 0.4087,
+      "step": 1678
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9436132520611894e-05,
+      "loss": 0.332,
+      "step": 1679
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9435796685305528e-05,
+      "loss": 0.4591,
+      "step": 1680
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9435460849999163e-05,
+      "loss": 0.2772,
+      "step": 1681
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9435125014692798e-05,
+      "loss": 0.3004,
+      "step": 1682
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9434789179386432e-05,
+      "loss": 0.4155,
+      "step": 1683
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9434453344080063e-05,
+      "loss": 0.2732,
+      "step": 1684
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.94341175087737e-05,
+      "loss": 0.2371,
+      "step": 1685
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9433781673467333e-05,
+      "loss": 0.4529,
+      "step": 1686
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9433445838160967e-05,
+      "loss": 0.3016,
+      "step": 1687
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9433110002854602e-05,
+      "loss": 0.3238,
+      "step": 1688
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9432774167548237e-05,
+      "loss": 0.3029,
+      "step": 1689
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.943243833224187e-05,
+      "loss": 0.2631,
+      "step": 1690
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9432102496935506e-05,
+      "loss": 0.2482,
+      "step": 1691
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.943176666162914e-05,
+      "loss": 0.5367,
+      "step": 1692
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9431430826322772e-05,
+      "loss": 0.4438,
+      "step": 1693
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9431094991016407e-05,
+      "loss": 0.3931,
+      "step": 1694
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.943075915571004e-05,
+      "loss": 0.3517,
+      "step": 1695
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9430423320403676e-05,
+      "loss": 0.1877,
+      "step": 1696
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.943008748509731e-05,
+      "loss": 0.4005,
+      "step": 1697
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9429751649790945e-05,
+      "loss": 0.3457,
+      "step": 1698
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9429415814484576e-05,
+      "loss": 0.4652,
+      "step": 1699
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9429079979178214e-05,
+      "loss": 0.3812,
+      "step": 1700
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9428744143871846e-05,
+      "loss": 0.5459,
+      "step": 1701
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.942840830856548e-05,
+      "loss": 0.3473,
+      "step": 1702
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9428072473259115e-05,
+      "loss": 0.4912,
+      "step": 1703
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.942773663795275e-05,
+      "loss": 0.1891,
+      "step": 1704
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9427400802646384e-05,
+      "loss": 0.2723,
+      "step": 1705
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.942706496734002e-05,
+      "loss": 0.3732,
+      "step": 1706
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9426729132033654e-05,
+      "loss": 0.3363,
+      "step": 1707
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9426393296727285e-05,
+      "loss": 0.3933,
+      "step": 1708
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9426057461420923e-05,
+      "loss": 0.4794,
+      "step": 1709
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9425721626114554e-05,
+      "loss": 0.2973,
+      "step": 1710
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.942538579080819e-05,
+      "loss": 0.4195,
+      "step": 1711
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9425049955501823e-05,
+      "loss": 0.2898,
+      "step": 1712
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9424714120195458e-05,
+      "loss": 0.2885,
+      "step": 1713
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9424378284889093e-05,
+      "loss": 0.338,
+      "step": 1714
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9424042449582724e-05,
+      "loss": 0.4473,
+      "step": 1715
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9423706614276362e-05,
+      "loss": 0.3921,
+      "step": 1716
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9423370778969993e-05,
+      "loss": 0.3949,
+      "step": 1717
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9423034943663628e-05,
+      "loss": 0.2091,
+      "step": 1718
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9422699108357263e-05,
+      "loss": 0.3551,
+      "step": 1719
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9422363273050897e-05,
+      "loss": 0.4823,
+      "step": 1720
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9422027437744532e-05,
+      "loss": 0.2996,
+      "step": 1721
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9421691602438166e-05,
+      "loss": 0.38,
+      "step": 1722
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.94213557671318e-05,
+      "loss": 0.4258,
+      "step": 1723
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9421019931825432e-05,
+      "loss": 0.3144,
+      "step": 1724
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.942068409651907e-05,
+      "loss": 0.4204,
+      "step": 1725
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.94203482612127e-05,
+      "loss": 0.3634,
+      "step": 1726
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9420012425906336e-05,
+      "loss": 0.1999,
+      "step": 1727
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.941967659059997e-05,
+      "loss": 0.3121,
+      "step": 1728
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9419340755293606e-05,
+      "loss": 0.4009,
+      "step": 1729
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.941900491998724e-05,
+      "loss": 0.4294,
+      "step": 1730
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9418669084680875e-05,
+      "loss": 0.379,
+      "step": 1731
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.941833324937451e-05,
+      "loss": 0.31,
+      "step": 1732
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.941799741406814e-05,
+      "loss": 0.3423,
+      "step": 1733
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.941766157876178e-05,
+      "loss": 0.3033,
+      "step": 1734
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.941732574345541e-05,
+      "loss": 0.3648,
+      "step": 1735
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9416989908149045e-05,
+      "loss": 0.3583,
+      "step": 1736
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.941665407284268e-05,
+      "loss": 0.2664,
+      "step": 1737
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9416318237536314e-05,
+      "loss": 0.3016,
+      "step": 1738
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9415982402229945e-05,
+      "loss": 0.4075,
+      "step": 1739
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9415646566923583e-05,
+      "loss": 0.4252,
+      "step": 1740
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9415310731617215e-05,
+      "loss": 0.3685,
+      "step": 1741
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.941497489631085e-05,
+      "loss": 0.3143,
+      "step": 1742
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9414639061004484e-05,
+      "loss": 0.501,
+      "step": 1743
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.941430322569812e-05,
+      "loss": 0.4115,
+      "step": 1744
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9413967390391753e-05,
+      "loss": 0.4457,
+      "step": 1745
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9413631555085388e-05,
+      "loss": 0.3694,
+      "step": 1746
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9413295719779022e-05,
+      "loss": 0.3247,
+      "step": 1747
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9412959884472654e-05,
+      "loss": 0.3358,
+      "step": 1748
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9412624049166292e-05,
+      "loss": 0.3284,
+      "step": 1749
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9412288213859923e-05,
+      "loss": 0.3067,
+      "step": 1750
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9411952378553558e-05,
+      "loss": 0.5487,
+      "step": 1751
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9411616543247192e-05,
+      "loss": 0.2142,
+      "step": 1752
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9411280707940827e-05,
+      "loss": 0.2941,
+      "step": 1753
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.941094487263446e-05,
+      "loss": 0.3808,
+      "step": 1754
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9410609037328096e-05,
+      "loss": 0.448,
+      "step": 1755
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.941027320202173e-05,
+      "loss": 0.3931,
+      "step": 1756
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9409937366715362e-05,
+      "loss": 0.4556,
+      "step": 1757
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9409601531409e-05,
+      "loss": 0.4981,
+      "step": 1758
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940926569610263e-05,
+      "loss": 0.4872,
+      "step": 1759
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9408929860796266e-05,
+      "loss": 0.5174,
+      "step": 1760
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.94085940254899e-05,
+      "loss": 0.3868,
+      "step": 1761
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9408258190183535e-05,
+      "loss": 0.418,
+      "step": 1762
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940792235487717e-05,
+      "loss": 0.4959,
+      "step": 1763
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9407586519570805e-05,
+      "loss": 0.479,
+      "step": 1764
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940725068426444e-05,
+      "loss": 0.3803,
+      "step": 1765
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940691484895807e-05,
+      "loss": 0.3574,
+      "step": 1766
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940657901365171e-05,
+      "loss": 0.4859,
+      "step": 1767
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940624317834534e-05,
+      "loss": 0.525,
+      "step": 1768
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9405907343038975e-05,
+      "loss": 0.3782,
+      "step": 1769
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940557150773261e-05,
+      "loss": 0.451,
+      "step": 1770
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9405235672426244e-05,
+      "loss": 0.319,
+      "step": 1771
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940489983711988e-05,
+      "loss": 0.3132,
+      "step": 1772
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9404564001813513e-05,
+      "loss": 0.4044,
+      "step": 1773
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9404228166507148e-05,
+      "loss": 0.4737,
+      "step": 1774
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940389233120078e-05,
+      "loss": 0.2764,
+      "step": 1775
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9403556495894417e-05,
+      "loss": 0.3939,
+      "step": 1776
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940322066058805e-05,
+      "loss": 0.4091,
+      "step": 1777
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9402884825281683e-05,
+      "loss": 0.4027,
+      "step": 1778
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9402548989975318e-05,
+      "loss": 0.29,
+      "step": 1779
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9402213154668952e-05,
+      "loss": 0.4795,
+      "step": 1780
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9401877319362587e-05,
+      "loss": 0.3798,
+      "step": 1781
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.940154148405622e-05,
+      "loss": 0.3768,
+      "step": 1782
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9401205648749856e-05,
+      "loss": 0.2474,
+      "step": 1783
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9400869813443488e-05,
+      "loss": 0.399,
+      "step": 1784
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9400533978137122e-05,
+      "loss": 0.3973,
+      "step": 1785
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9400198142830757e-05,
+      "loss": 0.2871,
+      "step": 1786
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939986230752439e-05,
+      "loss": 0.4361,
+      "step": 1787
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9399526472218026e-05,
+      "loss": 0.5045,
+      "step": 1788
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939919063691166e-05,
+      "loss": 0.403,
+      "step": 1789
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9398854801605292e-05,
+      "loss": 0.3685,
+      "step": 1790
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939851896629893e-05,
+      "loss": 0.3319,
+      "step": 1791
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939818313099256e-05,
+      "loss": 0.33,
+      "step": 1792
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9397847295686196e-05,
+      "loss": 0.2805,
+      "step": 1793
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939751146037983e-05,
+      "loss": 0.3391,
+      "step": 1794
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9397175625073465e-05,
+      "loss": 0.4201,
+      "step": 1795
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.93968397897671e-05,
+      "loss": 0.3602,
+      "step": 1796
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9396503954460735e-05,
+      "loss": 0.3932,
+      "step": 1797
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939616811915437e-05,
+      "loss": 0.3276,
+      "step": 1798
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9395832283848e-05,
+      "loss": 0.2894,
+      "step": 1799
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939549644854164e-05,
+      "loss": 0.4323,
+      "step": 1800
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939516061323527e-05,
+      "loss": 0.4919,
+      "step": 1801
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9394824777928904e-05,
+      "loss": 0.4927,
+      "step": 1802
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939448894262254e-05,
+      "loss": 0.428,
+      "step": 1803
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9394153107316174e-05,
+      "loss": 0.4554,
+      "step": 1804
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939381727200981e-05,
+      "loss": 0.3903,
+      "step": 1805
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9393481436703443e-05,
+      "loss": 0.4824,
+      "step": 1806
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9393145601397078e-05,
+      "loss": 0.2459,
+      "step": 1807
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939280976609071e-05,
+      "loss": 0.3376,
+      "step": 1808
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9392473930784347e-05,
+      "loss": 0.251,
+      "step": 1809
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9392138095477978e-05,
+      "loss": 0.2727,
+      "step": 1810
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9391802260171613e-05,
+      "loss": 0.4424,
+      "step": 1811
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9391466424865247e-05,
+      "loss": 0.1534,
+      "step": 1812
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9391130589558882e-05,
+      "loss": 0.2637,
+      "step": 1813
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9390794754252517e-05,
+      "loss": 0.2477,
+      "step": 1814
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.939045891894615e-05,
+      "loss": 0.3437,
+      "step": 1815
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9390123083639786e-05,
+      "loss": 0.3532,
+      "step": 1816
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9389787248333417e-05,
+      "loss": 0.4036,
+      "step": 1817
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9389451413027055e-05,
+      "loss": 0.451,
+      "step": 1818
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9389115577720687e-05,
+      "loss": 0.2298,
+      "step": 1819
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.938877974241432e-05,
+      "loss": 0.5844,
+      "step": 1820
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9388443907107956e-05,
+      "loss": 0.4043,
+      "step": 1821
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.938810807180159e-05,
+      "loss": 0.4531,
+      "step": 1822
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9387772236495225e-05,
+      "loss": 0.3782,
+      "step": 1823
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.938743640118886e-05,
+      "loss": 0.4936,
+      "step": 1824
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9387100565882495e-05,
+      "loss": 0.3517,
+      "step": 1825
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9386764730576126e-05,
+      "loss": 0.3268,
+      "step": 1826
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9386428895269764e-05,
+      "loss": 0.346,
+      "step": 1827
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9386093059963395e-05,
+      "loss": 0.4621,
+      "step": 1828
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.938575722465703e-05,
+      "loss": 0.3524,
+      "step": 1829
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9385421389350664e-05,
+      "loss": 0.4062,
+      "step": 1830
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.93850855540443e-05,
+      "loss": 0.4977,
+      "step": 1831
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.938474971873793e-05,
+      "loss": 0.3105,
+      "step": 1832
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9384413883431568e-05,
+      "loss": 0.4906,
+      "step": 1833
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.93840780481252e-05,
+      "loss": 0.543,
+      "step": 1834
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9383742212818834e-05,
+      "loss": 0.4306,
+      "step": 1835
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.938340637751247e-05,
+      "loss": 0.3242,
+      "step": 1836
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9383070542206103e-05,
+      "loss": 0.349,
+      "step": 1837
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9382734706899738e-05,
+      "loss": 0.4968,
+      "step": 1838
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9382398871593373e-05,
+      "loss": 0.3996,
+      "step": 1839
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9382063036287007e-05,
+      "loss": 0.4228,
+      "step": 1840
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.938172720098064e-05,
+      "loss": 0.1629,
+      "step": 1841
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9381391365674277e-05,
+      "loss": 0.4629,
+      "step": 1842
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9381055530367908e-05,
+      "loss": 0.4004,
+      "step": 1843
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9380719695061543e-05,
+      "loss": 0.4203,
+      "step": 1844
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9380383859755177e-05,
+      "loss": 0.4317,
+      "step": 1845
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9380048024448812e-05,
+      "loss": 0.3063,
+      "step": 1846
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9379712189142447e-05,
+      "loss": 0.3159,
+      "step": 1847
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.937937635383608e-05,
+      "loss": 0.4151,
+      "step": 1848
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9379040518529716e-05,
+      "loss": 0.2409,
+      "step": 1849
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9378704683223347e-05,
+      "loss": 0.3614,
+      "step": 1850
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9378368847916985e-05,
+      "loss": 0.41,
+      "step": 1851
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9378033012610616e-05,
+      "loss": 0.3276,
+      "step": 1852
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.937769717730425e-05,
+      "loss": 0.2903,
+      "step": 1853
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9377361341997886e-05,
+      "loss": 0.2077,
+      "step": 1854
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.937702550669152e-05,
+      "loss": 0.4187,
+      "step": 1855
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9376689671385155e-05,
+      "loss": 0.3532,
+      "step": 1856
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.937635383607879e-05,
+      "loss": 0.4361,
+      "step": 1857
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9376018000772424e-05,
+      "loss": 0.3936,
+      "step": 1858
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9375682165466056e-05,
+      "loss": 0.3278,
+      "step": 1859
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9375346330159694e-05,
+      "loss": 0.4319,
+      "step": 1860
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9375010494853325e-05,
+      "loss": 0.2221,
+      "step": 1861
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.937467465954696e-05,
+      "loss": 0.6665,
+      "step": 1862
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9374338824240594e-05,
+      "loss": 0.378,
+      "step": 1863
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.937400298893423e-05,
+      "loss": 0.3625,
+      "step": 1864
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9373667153627863e-05,
+      "loss": 0.4817,
+      "step": 1865
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9373331318321498e-05,
+      "loss": 0.5282,
+      "step": 1866
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9372995483015133e-05,
+      "loss": 0.3332,
+      "step": 1867
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9372659647708764e-05,
+      "loss": 0.3013,
+      "step": 1868
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9372323812402402e-05,
+      "loss": 0.2701,
+      "step": 1869
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9371987977096033e-05,
+      "loss": 0.4576,
+      "step": 1870
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9371652141789668e-05,
+      "loss": 0.2646,
+      "step": 1871
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9371316306483303e-05,
+      "loss": 0.5163,
+      "step": 1872
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9370980471176937e-05,
+      "loss": 0.3972,
+      "step": 1873
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9370644635870572e-05,
+      "loss": 0.3875,
+      "step": 1874
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9370308800564207e-05,
+      "loss": 0.3284,
+      "step": 1875
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9369972965257838e-05,
+      "loss": 0.3157,
+      "step": 1876
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9369637129951472e-05,
+      "loss": 0.3783,
+      "step": 1877
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9369301294645107e-05,
+      "loss": 0.2988,
+      "step": 1878
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9368965459338742e-05,
+      "loss": 0.4017,
+      "step": 1879
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9368629624032376e-05,
+      "loss": 0.4221,
+      "step": 1880
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.936829378872601e-05,
+      "loss": 0.2912,
+      "step": 1881
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9367957953419646e-05,
+      "loss": 0.2982,
+      "step": 1882
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9367622118113277e-05,
+      "loss": 0.4147,
+      "step": 1883
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9367286282806915e-05,
+      "loss": 0.3993,
+      "step": 1884
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9366950447500546e-05,
+      "loss": 0.2209,
+      "step": 1885
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.936661461219418e-05,
+      "loss": 0.3617,
+      "step": 1886
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9366278776887816e-05,
+      "loss": 0.3279,
+      "step": 1887
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.936594294158145e-05,
+      "loss": 0.5159,
+      "step": 1888
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9365607106275085e-05,
+      "loss": 0.4235,
+      "step": 1889
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.936527127096872e-05,
+      "loss": 0.3656,
+      "step": 1890
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9364935435662354e-05,
+      "loss": 0.2361,
+      "step": 1891
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9364599600355985e-05,
+      "loss": 0.3824,
+      "step": 1892
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9364263765049623e-05,
+      "loss": 0.4147,
+      "step": 1893
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9363927929743255e-05,
+      "loss": 0.3945,
+      "step": 1894
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.936359209443689e-05,
+      "loss": 0.564,
+      "step": 1895
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9363256259130524e-05,
+      "loss": 0.3117,
+      "step": 1896
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.936292042382416e-05,
+      "loss": 0.4342,
+      "step": 1897
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9362584588517793e-05,
+      "loss": 0.2869,
+      "step": 1898
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9362248753211425e-05,
+      "loss": 0.1621,
+      "step": 1899
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9361912917905063e-05,
+      "loss": 0.446,
+      "step": 1900
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9361577082598694e-05,
+      "loss": 0.3595,
+      "step": 1901
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.936124124729233e-05,
+      "loss": 0.4016,
+      "step": 1902
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9360905411985963e-05,
+      "loss": 0.3049,
+      "step": 1903
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9360569576679598e-05,
+      "loss": 0.5315,
+      "step": 1904
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9360233741373232e-05,
+      "loss": 0.2718,
+      "step": 1905
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9359897906066867e-05,
+      "loss": 0.4083,
+      "step": 1906
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9359562070760502e-05,
+      "loss": 0.3026,
+      "step": 1907
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9359226235454133e-05,
+      "loss": 0.3907,
+      "step": 1908
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935889040014777e-05,
+      "loss": 0.2943,
+      "step": 1909
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9358554564841402e-05,
+      "loss": 0.5138,
+      "step": 1910
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9358218729535037e-05,
+      "loss": 0.402,
+      "step": 1911
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935788289422867e-05,
+      "loss": 0.4481,
+      "step": 1912
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9357547058922306e-05,
+      "loss": 0.2904,
+      "step": 1913
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935721122361594e-05,
+      "loss": 0.4295,
+      "step": 1914
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9356875388309575e-05,
+      "loss": 0.3906,
+      "step": 1915
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935653955300321e-05,
+      "loss": 0.2133,
+      "step": 1916
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935620371769684e-05,
+      "loss": 0.4467,
+      "step": 1917
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935586788239048e-05,
+      "loss": 0.3862,
+      "step": 1918
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935553204708411e-05,
+      "loss": 0.2576,
+      "step": 1919
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9355196211777745e-05,
+      "loss": 0.2572,
+      "step": 1920
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935486037647138e-05,
+      "loss": 0.3588,
+      "step": 1921
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9354524541165015e-05,
+      "loss": 0.3122,
+      "step": 1922
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9354188705858646e-05,
+      "loss": 0.5072,
+      "step": 1923
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9353852870552284e-05,
+      "loss": 0.266,
+      "step": 1924
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9353517035245915e-05,
+      "loss": 0.4004,
+      "step": 1925
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935318119993955e-05,
+      "loss": 0.4546,
+      "step": 1926
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9352845364633184e-05,
+      "loss": 0.2948,
+      "step": 1927
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935250952932682e-05,
+      "loss": 0.3547,
+      "step": 1928
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9352173694020454e-05,
+      "loss": 0.4267,
+      "step": 1929
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.935183785871409e-05,
+      "loss": 0.4166,
+      "step": 1930
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9351502023407723e-05,
+      "loss": 0.3925,
+      "step": 1931
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9351166188101354e-05,
+      "loss": 0.3768,
+      "step": 1932
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9350830352794992e-05,
+      "loss": 0.4998,
+      "step": 1933
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9350494517488624e-05,
+      "loss": 0.3786,
+      "step": 1934
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9350158682182258e-05,
+      "loss": 0.3545,
+      "step": 1935
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9349822846875893e-05,
+      "loss": 0.3559,
+      "step": 1936
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9349487011569528e-05,
+      "loss": 0.4335,
+      "step": 1937
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9349151176263162e-05,
+      "loss": 0.515,
+      "step": 1938
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9348815340956797e-05,
+      "loss": 0.4359,
+      "step": 1939
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934847950565043e-05,
+      "loss": 0.3184,
+      "step": 1940
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9348143670344063e-05,
+      "loss": 0.4516,
+      "step": 1941
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.93478078350377e-05,
+      "loss": 0.2551,
+      "step": 1942
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9347471999731332e-05,
+      "loss": 0.2814,
+      "step": 1943
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9347136164424967e-05,
+      "loss": 0.4276,
+      "step": 1944
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.93468003291186e-05,
+      "loss": 0.3395,
+      "step": 1945
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9346464493812236e-05,
+      "loss": 0.4817,
+      "step": 1946
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934612865850587e-05,
+      "loss": 0.3693,
+      "step": 1947
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9345792823199505e-05,
+      "loss": 0.3447,
+      "step": 1948
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934545698789314e-05,
+      "loss": 0.2692,
+      "step": 1949
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934512115258677e-05,
+      "loss": 0.352,
+      "step": 1950
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934478531728041e-05,
+      "loss": 0.333,
+      "step": 1951
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934444948197404e-05,
+      "loss": 0.3446,
+      "step": 1952
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9344113646667675e-05,
+      "loss": 0.3719,
+      "step": 1953
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934377781136131e-05,
+      "loss": 0.2859,
+      "step": 1954
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9343441976054944e-05,
+      "loss": 0.4265,
+      "step": 1955
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934310614074858e-05,
+      "loss": 0.3633,
+      "step": 1956
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9342770305442214e-05,
+      "loss": 0.3264,
+      "step": 1957
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934243447013585e-05,
+      "loss": 0.3253,
+      "step": 1958
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934209863482948e-05,
+      "loss": 0.6536,
+      "step": 1959
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9341762799523118e-05,
+      "loss": 0.3814,
+      "step": 1960
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.934142696421675e-05,
+      "loss": 0.4432,
+      "step": 1961
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9341091128910384e-05,
+      "loss": 0.3779,
+      "step": 1962
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9340755293604018e-05,
+      "loss": 0.4588,
+      "step": 1963
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9340419458297653e-05,
+      "loss": 0.4915,
+      "step": 1964
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9340083622991284e-05,
+      "loss": 0.3142,
+      "step": 1965
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9339747787684922e-05,
+      "loss": 0.3894,
+      "step": 1966
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9339411952378553e-05,
+      "loss": 0.3121,
+      "step": 1967
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9339076117072188e-05,
+      "loss": 0.4145,
+      "step": 1968
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9338740281765823e-05,
+      "loss": 0.2882,
+      "step": 1969
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9338404446459457e-05,
+      "loss": 0.3403,
+      "step": 1970
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9338068611153092e-05,
+      "loss": 0.2354,
+      "step": 1971
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9337732775846727e-05,
+      "loss": 0.346,
+      "step": 1972
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.933739694054036e-05,
+      "loss": 0.3181,
+      "step": 1973
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9337061105233993e-05,
+      "loss": 0.3189,
+      "step": 1974
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.933672526992763e-05,
+      "loss": 0.3129,
+      "step": 1975
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9336389434621262e-05,
+      "loss": 0.2677,
+      "step": 1976
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9336053599314897e-05,
+      "loss": 0.351,
+      "step": 1977
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.933571776400853e-05,
+      "loss": 0.3557,
+      "step": 1978
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9335381928702166e-05,
+      "loss": 0.4175,
+      "step": 1979
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.93350460933958e-05,
+      "loss": 0.387,
+      "step": 1980
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9334710258089435e-05,
+      "loss": 0.4938,
+      "step": 1981
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.933437442278307e-05,
+      "loss": 0.2029,
+      "step": 1982
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.93340385874767e-05,
+      "loss": 0.2841,
+      "step": 1983
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.933370275217034e-05,
+      "loss": 0.4193,
+      "step": 1984
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.933336691686397e-05,
+      "loss": 0.4815,
+      "step": 1985
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9333031081557605e-05,
+      "loss": 0.3072,
+      "step": 1986
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.933269524625124e-05,
+      "loss": 0.3809,
+      "step": 1987
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9332359410944874e-05,
+      "loss": 0.1948,
+      "step": 1988
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.933202357563851e-05,
+      "loss": 0.3118,
+      "step": 1989
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9331687740332144e-05,
+      "loss": 0.4123,
+      "step": 1990
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9331351905025778e-05,
+      "loss": 0.2444,
+      "step": 1991
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.933101606971941e-05,
+      "loss": 0.3288,
+      "step": 1992
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9330680234413047e-05,
+      "loss": 0.3449,
+      "step": 1993
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.933034439910668e-05,
+      "loss": 0.2716,
+      "step": 1994
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9330008563800313e-05,
+      "loss": 0.3924,
+      "step": 1995
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9329672728493948e-05,
+      "loss": 0.2896,
+      "step": 1996
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9329336893187583e-05,
+      "loss": 0.5514,
+      "step": 1997
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9329001057881217e-05,
+      "loss": 0.4734,
+      "step": 1998
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9328665222574852e-05,
+      "loss": 0.4337,
+      "step": 1999
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9328329387268487e-05,
+      "loss": 0.2606,
+      "step": 2000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9327993551962118e-05,
+      "loss": 0.502,
+      "step": 2001
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9327657716655756e-05,
+      "loss": 0.4973,
+      "step": 2002
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9327321881349387e-05,
+      "loss": 0.4087,
+      "step": 2003
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9326986046043022e-05,
+      "loss": 0.3785,
+      "step": 2004
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9326650210736656e-05,
+      "loss": 0.412,
+      "step": 2005
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.932631437543029e-05,
+      "loss": 0.3804,
+      "step": 2006
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9325978540123926e-05,
+      "loss": 0.4091,
+      "step": 2007
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.932564270481756e-05,
+      "loss": 0.3203,
+      "step": 2008
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9325306869511195e-05,
+      "loss": 0.3475,
+      "step": 2009
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9324971034204826e-05,
+      "loss": 0.364,
+      "step": 2010
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.932463519889846e-05,
+      "loss": 0.3893,
+      "step": 2011
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9324299363592096e-05,
+      "loss": 0.2647,
+      "step": 2012
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.932396352828573e-05,
+      "loss": 0.4232,
+      "step": 2013
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9323627692979365e-05,
+      "loss": 0.4085,
+      "step": 2014
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9323291857673e-05,
+      "loss": 0.3318,
+      "step": 2015
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.932295602236663e-05,
+      "loss": 0.2864,
+      "step": 2016
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.932262018706027e-05,
+      "loss": 0.2275,
+      "step": 2017
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.93222843517539e-05,
+      "loss": 0.3681,
+      "step": 2018
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9321948516447535e-05,
+      "loss": 0.4892,
+      "step": 2019
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.932161268114117e-05,
+      "loss": 0.4015,
+      "step": 2020
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9321276845834804e-05,
+      "loss": 0.3949,
+      "step": 2021
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.932094101052844e-05,
+      "loss": 0.6562,
+      "step": 2022
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9320605175222073e-05,
+      "loss": 0.3994,
+      "step": 2023
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9320269339915708e-05,
+      "loss": 0.2845,
+      "step": 2024
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.931993350460934e-05,
+      "loss": 0.342,
+      "step": 2025
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9319597669302977e-05,
+      "loss": 0.226,
+      "step": 2026
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.931926183399661e-05,
+      "loss": 0.2287,
+      "step": 2027
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9318925998690243e-05,
+      "loss": 0.4353,
+      "step": 2028
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9318590163383878e-05,
+      "loss": 0.261,
+      "step": 2029
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9318254328077513e-05,
+      "loss": 0.4127,
+      "step": 2030
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9317918492771147e-05,
+      "loss": 0.4969,
+      "step": 2031
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9317582657464782e-05,
+      "loss": 0.2786,
+      "step": 2032
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9317246822158416e-05,
+      "loss": 0.4803,
+      "step": 2033
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9316910986852048e-05,
+      "loss": 0.3863,
+      "step": 2034
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9316575151545686e-05,
+      "loss": 0.4156,
+      "step": 2035
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9316239316239317e-05,
+      "loss": 0.3919,
+      "step": 2036
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.931590348093295e-05,
+      "loss": 0.1837,
+      "step": 2037
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9315567645626586e-05,
+      "loss": 0.2468,
+      "step": 2038
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.931523181032022e-05,
+      "loss": 0.2545,
+      "step": 2039
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9314895975013856e-05,
+      "loss": 0.3193,
+      "step": 2040
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.931456013970749e-05,
+      "loss": 0.3111,
+      "step": 2041
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9314224304401125e-05,
+      "loss": 0.7234,
+      "step": 2042
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9313888469094756e-05,
+      "loss": 0.2934,
+      "step": 2043
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9313552633788394e-05,
+      "loss": 0.4378,
+      "step": 2044
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9313216798482025e-05,
+      "loss": 0.3054,
+      "step": 2045
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.931288096317566e-05,
+      "loss": 0.3834,
+      "step": 2046
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9312545127869295e-05,
+      "loss": 0.3615,
+      "step": 2047
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.931220929256293e-05,
+      "loss": 0.3157,
+      "step": 2048
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9311873457256564e-05,
+      "loss": 0.2861,
+      "step": 2049
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.93115376219502e-05,
+      "loss": 0.455,
+      "step": 2050
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9311201786643833e-05,
+      "loss": 0.405,
+      "step": 2051
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9310865951337465e-05,
+      "loss": 0.4802,
+      "step": 2052
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9310530116031103e-05,
+      "loss": 0.2538,
+      "step": 2053
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9310194280724734e-05,
+      "loss": 0.4124,
+      "step": 2054
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.930985844541837e-05,
+      "loss": 0.2437,
+      "step": 2055
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9309522610112003e-05,
+      "loss": 0.3397,
+      "step": 2056
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9309186774805638e-05,
+      "loss": 0.2238,
+      "step": 2057
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.930885093949927e-05,
+      "loss": 0.3552,
+      "step": 2058
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9308515104192907e-05,
+      "loss": 0.29,
+      "step": 2059
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.930817926888654e-05,
+      "loss": 0.5432,
+      "step": 2060
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9307843433580173e-05,
+      "loss": 0.4889,
+      "step": 2061
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9307507598273808e-05,
+      "loss": 0.2882,
+      "step": 2062
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9307171762967442e-05,
+      "loss": 0.3389,
+      "step": 2063
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9306835927661077e-05,
+      "loss": 0.5288,
+      "step": 2064
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.930650009235471e-05,
+      "loss": 0.723,
+      "step": 2065
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9306164257048346e-05,
+      "loss": 0.4971,
+      "step": 2066
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9305828421741978e-05,
+      "loss": 0.375,
+      "step": 2067
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9305492586435616e-05,
+      "loss": 0.3632,
+      "step": 2068
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9305156751129247e-05,
+      "loss": 0.3168,
+      "step": 2069
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.930482091582288e-05,
+      "loss": 0.3099,
+      "step": 2070
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9304485080516516e-05,
+      "loss": 0.2945,
+      "step": 2071
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.930414924521015e-05,
+      "loss": 0.3566,
+      "step": 2072
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9303813409903785e-05,
+      "loss": 0.3086,
+      "step": 2073
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.930347757459742e-05,
+      "loss": 0.1802,
+      "step": 2074
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9303141739291055e-05,
+      "loss": 0.2728,
+      "step": 2075
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9302805903984686e-05,
+      "loss": 0.3213,
+      "step": 2076
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.930247006867832e-05,
+      "loss": 0.4142,
+      "step": 2077
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9302134233371955e-05,
+      "loss": 0.2241,
+      "step": 2078
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.930179839806559e-05,
+      "loss": 0.2874,
+      "step": 2079
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9301462562759225e-05,
+      "loss": 0.3198,
+      "step": 2080
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.930112672745286e-05,
+      "loss": 0.3513,
+      "step": 2081
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9300790892146494e-05,
+      "loss": 0.3834,
+      "step": 2082
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9300455056840125e-05,
+      "loss": 0.3846,
+      "step": 2083
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9300119221533763e-05,
+      "loss": 0.461,
+      "step": 2084
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9299783386227394e-05,
+      "loss": 0.4195,
+      "step": 2085
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.929944755092103e-05,
+      "loss": 0.2874,
+      "step": 2086
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9299111715614664e-05,
+      "loss": 0.386,
+      "step": 2087
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.92987758803083e-05,
+      "loss": 0.3841,
+      "step": 2088
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9298440045001933e-05,
+      "loss": 0.3895,
+      "step": 2089
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9298104209695568e-05,
+      "loss": 0.6226,
+      "step": 2090
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9297768374389202e-05,
+      "loss": 0.4139,
+      "step": 2091
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9297432539082834e-05,
+      "loss": 0.3389,
+      "step": 2092
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.929709670377647e-05,
+      "loss": 0.409,
+      "step": 2093
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9296760868470103e-05,
+      "loss": 0.3498,
+      "step": 2094
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9296425033163737e-05,
+      "loss": 0.4212,
+      "step": 2095
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9296089197857372e-05,
+      "loss": 0.2747,
+      "step": 2096
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9295753362551007e-05,
+      "loss": 0.4423,
+      "step": 2097
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.929541752724464e-05,
+      "loss": 0.3964,
+      "step": 2098
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9295081691938276e-05,
+      "loss": 0.2232,
+      "step": 2099
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9294745856631907e-05,
+      "loss": 0.5211,
+      "step": 2100
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9294410021325542e-05,
+      "loss": 0.4787,
+      "step": 2101
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9294074186019177e-05,
+      "loss": 0.1722,
+      "step": 2102
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.929373835071281e-05,
+      "loss": 0.3343,
+      "step": 2103
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9293402515406446e-05,
+      "loss": 0.2189,
+      "step": 2104
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.929306668010008e-05,
+      "loss": 0.3867,
+      "step": 2105
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9292730844793715e-05,
+      "loss": 0.4681,
+      "step": 2106
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9292395009487346e-05,
+      "loss": 0.3251,
+      "step": 2107
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9292059174180985e-05,
+      "loss": 0.4238,
+      "step": 2108
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9291723338874616e-05,
+      "loss": 0.525,
+      "step": 2109
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.929138750356825e-05,
+      "loss": 0.4456,
+      "step": 2110
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9291051668261885e-05,
+      "loss": 0.4051,
+      "step": 2111
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.929071583295552e-05,
+      "loss": 0.2106,
+      "step": 2112
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9290379997649154e-05,
+      "loss": 0.3284,
+      "step": 2113
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.929004416234279e-05,
+      "loss": 0.3575,
+      "step": 2114
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9289708327036424e-05,
+      "loss": 0.3776,
+      "step": 2115
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9289372491730055e-05,
+      "loss": 0.3448,
+      "step": 2116
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9289036656423693e-05,
+      "loss": 0.44,
+      "step": 2117
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9288700821117324e-05,
+      "loss": 0.3277,
+      "step": 2118
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.928836498581096e-05,
+      "loss": 0.2711,
+      "step": 2119
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9288029150504594e-05,
+      "loss": 0.3671,
+      "step": 2120
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9287693315198228e-05,
+      "loss": 0.3072,
+      "step": 2121
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9287357479891863e-05,
+      "loss": 0.3966,
+      "step": 2122
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9287021644585497e-05,
+      "loss": 0.2511,
+      "step": 2123
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9286685809279132e-05,
+      "loss": 0.3604,
+      "step": 2124
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9286349973972763e-05,
+      "loss": 0.1799,
+      "step": 2125
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.92860141386664e-05,
+      "loss": 0.3646,
+      "step": 2126
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9285678303360033e-05,
+      "loss": 0.3047,
+      "step": 2127
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9285342468053667e-05,
+      "loss": 0.3209,
+      "step": 2128
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9285006632747302e-05,
+      "loss": 0.5347,
+      "step": 2129
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9284670797440937e-05,
+      "loss": 0.4303,
+      "step": 2130
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.928433496213457e-05,
+      "loss": 0.3299,
+      "step": 2131
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9283999126828206e-05,
+      "loss": 0.583,
+      "step": 2132
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.928366329152184e-05,
+      "loss": 0.3376,
+      "step": 2133
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9283327456215472e-05,
+      "loss": 0.3107,
+      "step": 2134
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.928299162090911e-05,
+      "loss": 0.4269,
+      "step": 2135
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.928265578560274e-05,
+      "loss": 0.3947,
+      "step": 2136
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9282319950296376e-05,
+      "loss": 0.3821,
+      "step": 2137
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.928198411499001e-05,
+      "loss": 0.2395,
+      "step": 2138
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9281648279683645e-05,
+      "loss": 0.4239,
+      "step": 2139
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.928131244437728e-05,
+      "loss": 0.2417,
+      "step": 2140
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9280976609070914e-05,
+      "loss": 0.4199,
+      "step": 2141
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.928064077376455e-05,
+      "loss": 0.5913,
+      "step": 2142
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.928030493845818e-05,
+      "loss": 0.2783,
+      "step": 2143
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9279969103151818e-05,
+      "loss": 0.2216,
+      "step": 2144
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927963326784545e-05,
+      "loss": 0.3694,
+      "step": 2145
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9279297432539084e-05,
+      "loss": 0.2552,
+      "step": 2146
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927896159723272e-05,
+      "loss": 0.1882,
+      "step": 2147
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9278625761926353e-05,
+      "loss": 0.328,
+      "step": 2148
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9278289926619985e-05,
+      "loss": 0.3505,
+      "step": 2149
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9277954091313623e-05,
+      "loss": 0.2974,
+      "step": 2150
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9277618256007254e-05,
+      "loss": 0.3902,
+      "step": 2151
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927728242070089e-05,
+      "loss": 0.3589,
+      "step": 2152
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9276946585394523e-05,
+      "loss": 0.3152,
+      "step": 2153
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9276610750088158e-05,
+      "loss": 0.366,
+      "step": 2154
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9276274914781793e-05,
+      "loss": 0.2574,
+      "step": 2155
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9275939079475427e-05,
+      "loss": 0.3968,
+      "step": 2156
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9275603244169062e-05,
+      "loss": 0.3892,
+      "step": 2157
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9275267408862693e-05,
+      "loss": 0.3694,
+      "step": 2158
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927493157355633e-05,
+      "loss": 0.3105,
+      "step": 2159
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9274595738249962e-05,
+      "loss": 0.4281,
+      "step": 2160
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9274259902943597e-05,
+      "loss": 0.3508,
+      "step": 2161
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9273924067637232e-05,
+      "loss": 0.4022,
+      "step": 2162
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9273588232330866e-05,
+      "loss": 0.5806,
+      "step": 2163
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.92732523970245e-05,
+      "loss": 0.5239,
+      "step": 2164
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9272916561718136e-05,
+      "loss": 0.4582,
+      "step": 2165
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927258072641177e-05,
+      "loss": 0.3246,
+      "step": 2166
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.92722448911054e-05,
+      "loss": 0.4659,
+      "step": 2167
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927190905579904e-05,
+      "loss": 0.2797,
+      "step": 2168
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927157322049267e-05,
+      "loss": 0.2881,
+      "step": 2169
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9271237385186306e-05,
+      "loss": 0.2494,
+      "step": 2170
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927090154987994e-05,
+      "loss": 0.4301,
+      "step": 2171
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9270565714573575e-05,
+      "loss": 0.258,
+      "step": 2172
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.927022987926721e-05,
+      "loss": 0.5557,
+      "step": 2173
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9269894043960844e-05,
+      "loss": 0.3878,
+      "step": 2174
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926955820865448e-05,
+      "loss": 0.5305,
+      "step": 2175
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926922237334811e-05,
+      "loss": 0.3715,
+      "step": 2176
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9268886538041748e-05,
+      "loss": 0.3059,
+      "step": 2177
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926855070273538e-05,
+      "loss": 0.2261,
+      "step": 2178
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9268214867429014e-05,
+      "loss": 0.3398,
+      "step": 2179
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926787903212265e-05,
+      "loss": 0.4374,
+      "step": 2180
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9267543196816283e-05,
+      "loss": 0.5122,
+      "step": 2181
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9267207361509918e-05,
+      "loss": 0.3986,
+      "step": 2182
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9266871526203553e-05,
+      "loss": 0.3533,
+      "step": 2183
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9266535690897187e-05,
+      "loss": 0.4108,
+      "step": 2184
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926619985559082e-05,
+      "loss": 0.4427,
+      "step": 2185
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9265864020284457e-05,
+      "loss": 0.4081,
+      "step": 2186
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9265528184978088e-05,
+      "loss": 0.3764,
+      "step": 2187
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9265192349671722e-05,
+      "loss": 0.3333,
+      "step": 2188
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9264856514365357e-05,
+      "loss": 0.4801,
+      "step": 2189
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9264520679058992e-05,
+      "loss": 0.3361,
+      "step": 2190
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9264184843752623e-05,
+      "loss": 0.3036,
+      "step": 2191
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926384900844626e-05,
+      "loss": 0.1991,
+      "step": 2192
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9263513173139892e-05,
+      "loss": 0.2757,
+      "step": 2193
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9263177337833527e-05,
+      "loss": 0.4365,
+      "step": 2194
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926284150252716e-05,
+      "loss": 0.2626,
+      "step": 2195
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9262505667220796e-05,
+      "loss": 0.2664,
+      "step": 2196
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926216983191443e-05,
+      "loss": 0.2879,
+      "step": 2197
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9261833996608066e-05,
+      "loss": 0.3686,
+      "step": 2198
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.92614981613017e-05,
+      "loss": 0.3065,
+      "step": 2199
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926116232599533e-05,
+      "loss": 0.4156,
+      "step": 2200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.926082649068897e-05,
+      "loss": 0.4739,
+      "step": 2201
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.92604906553826e-05,
+      "loss": 0.2719,
+      "step": 2202
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9260154820076235e-05,
+      "loss": 0.256,
+      "step": 2203
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.925981898476987e-05,
+      "loss": 0.2299,
+      "step": 2204
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9259483149463505e-05,
+      "loss": 0.4074,
+      "step": 2205
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.925914731415714e-05,
+      "loss": 0.277,
+      "step": 2206
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9258811478850774e-05,
+      "loss": 0.3947,
+      "step": 2207
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.925847564354441e-05,
+      "loss": 0.433,
+      "step": 2208
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.925813980823804e-05,
+      "loss": 0.373,
+      "step": 2209
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9257803972931678e-05,
+      "loss": 0.3837,
+      "step": 2210
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.925746813762531e-05,
+      "loss": 0.263,
+      "step": 2211
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9257132302318944e-05,
+      "loss": 0.2717,
+      "step": 2212
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.925679646701258e-05,
+      "loss": 0.5462,
+      "step": 2213
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9256460631706213e-05,
+      "loss": 0.3818,
+      "step": 2214
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9256124796399848e-05,
+      "loss": 0.5331,
+      "step": 2215
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9255788961093482e-05,
+      "loss": 0.4815,
+      "step": 2216
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9255453125787117e-05,
+      "loss": 0.2766,
+      "step": 2217
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9255117290480748e-05,
+      "loss": 0.2715,
+      "step": 2218
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9254781455174386e-05,
+      "loss": 0.2874,
+      "step": 2219
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9254445619868018e-05,
+      "loss": 0.2402,
+      "step": 2220
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9254109784561652e-05,
+      "loss": 0.3364,
+      "step": 2221
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9253773949255287e-05,
+      "loss": 0.3362,
+      "step": 2222
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.925343811394892e-05,
+      "loss": 0.2736,
+      "step": 2223
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9253102278642556e-05,
+      "loss": 0.429,
+      "step": 2224
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.925276644333619e-05,
+      "loss": 0.579,
+      "step": 2225
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9252430608029825e-05,
+      "loss": 0.3604,
+      "step": 2226
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9252094772723457e-05,
+      "loss": 0.3199,
+      "step": 2227
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9251758937417095e-05,
+      "loss": 0.2137,
+      "step": 2228
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9251423102110726e-05,
+      "loss": 0.3954,
+      "step": 2229
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.925108726680436e-05,
+      "loss": 0.4065,
+      "step": 2230
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9250751431497995e-05,
+      "loss": 0.568,
+      "step": 2231
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.925041559619163e-05,
+      "loss": 0.4152,
+      "step": 2232
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9250079760885265e-05,
+      "loss": 0.206,
+      "step": 2233
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.92497439255789e-05,
+      "loss": 0.5756,
+      "step": 2234
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.924940809027253e-05,
+      "loss": 0.2661,
+      "step": 2235
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9249072254966165e-05,
+      "loss": 0.2138,
+      "step": 2236
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.92487364196598e-05,
+      "loss": 0.5237,
+      "step": 2237
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9248400584353434e-05,
+      "loss": 0.4141,
+      "step": 2238
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.924806474904707e-05,
+      "loss": 0.3986,
+      "step": 2239
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9247728913740704e-05,
+      "loss": 0.398,
+      "step": 2240
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.924739307843434e-05,
+      "loss": 0.3829,
+      "step": 2241
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.924705724312797e-05,
+      "loss": 0.3806,
+      "step": 2242
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9246721407821608e-05,
+      "loss": 0.3889,
+      "step": 2243
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.924638557251524e-05,
+      "loss": 0.3794,
+      "step": 2244
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9246049737208874e-05,
+      "loss": 0.3299,
+      "step": 2245
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9245713901902508e-05,
+      "loss": 0.3271,
+      "step": 2246
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9245378066596143e-05,
+      "loss": 0.321,
+      "step": 2247
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9245042231289778e-05,
+      "loss": 0.3944,
+      "step": 2248
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9244706395983412e-05,
+      "loss": 0.2993,
+      "step": 2249
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9244370560677047e-05,
+      "loss": 0.3133,
+      "step": 2250
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9244034725370678e-05,
+      "loss": 0.4275,
+      "step": 2251
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9243698890064316e-05,
+      "loss": 0.4407,
+      "step": 2252
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9243363054757947e-05,
+      "loss": 0.3878,
+      "step": 2253
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9243027219451582e-05,
+      "loss": 0.1414,
+      "step": 2254
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9242691384145217e-05,
+      "loss": 0.4696,
+      "step": 2255
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.924235554883885e-05,
+      "loss": 0.2982,
+      "step": 2256
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9242019713532486e-05,
+      "loss": 0.5864,
+      "step": 2257
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9241683878226117e-05,
+      "loss": 0.2932,
+      "step": 2258
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9241348042919755e-05,
+      "loss": 0.3274,
+      "step": 2259
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9241012207613387e-05,
+      "loss": 0.3207,
+      "step": 2260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.924067637230702e-05,
+      "loss": 0.4297,
+      "step": 2261
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9240340537000656e-05,
+      "loss": 0.5146,
+      "step": 2262
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.924000470169429e-05,
+      "loss": 0.5067,
+      "step": 2263
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9239668866387925e-05,
+      "loss": 0.2871,
+      "step": 2264
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.923933303108156e-05,
+      "loss": 0.3042,
+      "step": 2265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9238997195775194e-05,
+      "loss": 0.2773,
+      "step": 2266
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9238661360468826e-05,
+      "loss": 0.3134,
+      "step": 2267
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9238325525162464e-05,
+      "loss": 0.3917,
+      "step": 2268
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9237989689856095e-05,
+      "loss": 0.4089,
+      "step": 2269
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.923765385454973e-05,
+      "loss": 0.3488,
+      "step": 2270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9237318019243364e-05,
+      "loss": 0.5812,
+      "step": 2271
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9236982183937e-05,
+      "loss": 0.3081,
+      "step": 2272
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9236646348630634e-05,
+      "loss": 0.3541,
+      "step": 2273
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9236310513324268e-05,
+      "loss": 0.3776,
+      "step": 2274
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9235974678017903e-05,
+      "loss": 0.3888,
+      "step": 2275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9235638842711534e-05,
+      "loss": 0.2945,
+      "step": 2276
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9235303007405172e-05,
+      "loss": 0.2948,
+      "step": 2277
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9234967172098803e-05,
+      "loss": 0.368,
+      "step": 2278
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9234631336792438e-05,
+      "loss": 0.4693,
+      "step": 2279
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9234295501486073e-05,
+      "loss": 0.5459,
+      "step": 2280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9233959666179707e-05,
+      "loss": 0.332,
+      "step": 2281
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.923362383087334e-05,
+      "loss": 0.5718,
+      "step": 2282
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9233287995566977e-05,
+      "loss": 0.2386,
+      "step": 2283
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9232952160260608e-05,
+      "loss": 0.1721,
+      "step": 2284
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9232616324954243e-05,
+      "loss": 0.362,
+      "step": 2285
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9232280489647877e-05,
+      "loss": 0.3316,
+      "step": 2286
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9231944654341512e-05,
+      "loss": 0.3164,
+      "step": 2287
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9231608819035147e-05,
+      "loss": 0.1829,
+      "step": 2288
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.923127298372878e-05,
+      "loss": 0.1833,
+      "step": 2289
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9230937148422416e-05,
+      "loss": 0.5034,
+      "step": 2290
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9230601313116047e-05,
+      "loss": 0.4768,
+      "step": 2291
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9230265477809685e-05,
+      "loss": 0.3047,
+      "step": 2292
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9229929642503316e-05,
+      "loss": 0.3691,
+      "step": 2293
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.922959380719695e-05,
+      "loss": 0.2297,
+      "step": 2294
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9229257971890586e-05,
+      "loss": 0.296,
+      "step": 2295
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.922892213658422e-05,
+      "loss": 0.3863,
+      "step": 2296
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9228586301277855e-05,
+      "loss": 0.391,
+      "step": 2297
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.922825046597149e-05,
+      "loss": 0.4054,
+      "step": 2298
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9227914630665124e-05,
+      "loss": 0.2832,
+      "step": 2299
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9227578795358755e-05,
+      "loss": 0.3161,
+      "step": 2300
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9227242960052394e-05,
+      "loss": 0.2988,
+      "step": 2301
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9226907124746025e-05,
+      "loss": 0.3081,
+      "step": 2302
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.922657128943966e-05,
+      "loss": 0.3638,
+      "step": 2303
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9226235454133294e-05,
+      "loss": 0.367,
+      "step": 2304
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.922589961882693e-05,
+      "loss": 0.3993,
+      "step": 2305
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9225563783520563e-05,
+      "loss": 0.2976,
+      "step": 2306
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9225227948214198e-05,
+      "loss": 0.3856,
+      "step": 2307
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9224892112907833e-05,
+      "loss": 0.4143,
+      "step": 2308
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9224556277601464e-05,
+      "loss": 0.2935,
+      "step": 2309
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9224220442295102e-05,
+      "loss": 0.3897,
+      "step": 2310
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9223884606988733e-05,
+      "loss": 0.2362,
+      "step": 2311
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9223548771682368e-05,
+      "loss": 0.3842,
+      "step": 2312
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9223212936376003e-05,
+      "loss": 0.4608,
+      "step": 2313
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9222877101069637e-05,
+      "loss": 0.261,
+      "step": 2314
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9222541265763272e-05,
+      "loss": 0.3481,
+      "step": 2315
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9222205430456906e-05,
+      "loss": 0.4052,
+      "step": 2316
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.922186959515054e-05,
+      "loss": 0.2497,
+      "step": 2317
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9221533759844172e-05,
+      "loss": 0.4899,
+      "step": 2318
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.922119792453781e-05,
+      "loss": 0.4569,
+      "step": 2319
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.922086208923144e-05,
+      "loss": 0.4454,
+      "step": 2320
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9220526253925076e-05,
+      "loss": 0.3517,
+      "step": 2321
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.922019041861871e-05,
+      "loss": 0.2493,
+      "step": 2322
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9219854583312346e-05,
+      "loss": 0.2071,
+      "step": 2323
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.921951874800598e-05,
+      "loss": 0.2716,
+      "step": 2324
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9219182912699615e-05,
+      "loss": 0.42,
+      "step": 2325
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9218847077393246e-05,
+      "loss": 0.1606,
+      "step": 2326
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.921851124208688e-05,
+      "loss": 0.1745,
+      "step": 2327
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9218175406780515e-05,
+      "loss": 0.4456,
+      "step": 2328
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.921783957147415e-05,
+      "loss": 0.3679,
+      "step": 2329
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9217503736167785e-05,
+      "loss": 0.2743,
+      "step": 2330
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.921716790086142e-05,
+      "loss": 0.2343,
+      "step": 2331
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9216832065555054e-05,
+      "loss": 0.3514,
+      "step": 2332
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9216496230248685e-05,
+      "loss": 0.4017,
+      "step": 2333
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9216160394942323e-05,
+      "loss": 0.3725,
+      "step": 2334
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9215824559635955e-05,
+      "loss": 0.4392,
+      "step": 2335
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.921548872432959e-05,
+      "loss": 0.407,
+      "step": 2336
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9215152889023224e-05,
+      "loss": 0.287,
+      "step": 2337
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.921481705371686e-05,
+      "loss": 0.4612,
+      "step": 2338
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9214481218410493e-05,
+      "loss": 0.3778,
+      "step": 2339
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9214145383104128e-05,
+      "loss": 0.43,
+      "step": 2340
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9213809547797762e-05,
+      "loss": 0.2422,
+      "step": 2341
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9213473712491394e-05,
+      "loss": 0.2258,
+      "step": 2342
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9213137877185032e-05,
+      "loss": 0.322,
+      "step": 2343
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9212802041878663e-05,
+      "loss": 0.3197,
+      "step": 2344
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9212466206572298e-05,
+      "loss": 0.337,
+      "step": 2345
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9212130371265932e-05,
+      "loss": 0.4028,
+      "step": 2346
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9211794535959567e-05,
+      "loss": 0.2941,
+      "step": 2347
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.92114587006532e-05,
+      "loss": 0.3966,
+      "step": 2348
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9211122865346836e-05,
+      "loss": 0.2979,
+      "step": 2349
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.921078703004047e-05,
+      "loss": 0.3926,
+      "step": 2350
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9210451194734102e-05,
+      "loss": 0.4055,
+      "step": 2351
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.921011535942774e-05,
+      "loss": 0.4097,
+      "step": 2352
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920977952412137e-05,
+      "loss": 0.4606,
+      "step": 2353
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9209443688815006e-05,
+      "loss": 0.4001,
+      "step": 2354
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920910785350864e-05,
+      "loss": 0.1731,
+      "step": 2355
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9208772018202275e-05,
+      "loss": 0.3926,
+      "step": 2356
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920843618289591e-05,
+      "loss": 0.3191,
+      "step": 2357
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9208100347589545e-05,
+      "loss": 0.3778,
+      "step": 2358
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920776451228318e-05,
+      "loss": 0.3341,
+      "step": 2359
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920742867697681e-05,
+      "loss": 0.2353,
+      "step": 2360
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920709284167045e-05,
+      "loss": 0.4234,
+      "step": 2361
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920675700636408e-05,
+      "loss": 0.3847,
+      "step": 2362
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9206421171057715e-05,
+      "loss": 0.4165,
+      "step": 2363
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920608533575135e-05,
+      "loss": 0.3477,
+      "step": 2364
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9205749500444984e-05,
+      "loss": 0.3548,
+      "step": 2365
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920541366513862e-05,
+      "loss": 0.3306,
+      "step": 2366
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9205077829832253e-05,
+      "loss": 0.2536,
+      "step": 2367
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9204741994525888e-05,
+      "loss": 0.5186,
+      "step": 2368
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920440615921952e-05,
+      "loss": 0.2627,
+      "step": 2369
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9204070323913154e-05,
+      "loss": 0.4161,
+      "step": 2370
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920373448860679e-05,
+      "loss": 0.2988,
+      "step": 2371
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9203398653300423e-05,
+      "loss": 0.2936,
+      "step": 2372
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9203062817994058e-05,
+      "loss": 0.4288,
+      "step": 2373
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9202726982687692e-05,
+      "loss": 0.2521,
+      "step": 2374
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9202391147381324e-05,
+      "loss": 0.3123,
+      "step": 2375
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920205531207496e-05,
+      "loss": 0.4007,
+      "step": 2376
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9201719476768593e-05,
+      "loss": 0.3728,
+      "step": 2377
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9201383641462228e-05,
+      "loss": 0.5779,
+      "step": 2378
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9201047806155862e-05,
+      "loss": 0.3366,
+      "step": 2379
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9200711970849497e-05,
+      "loss": 0.2455,
+      "step": 2380
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.920037613554313e-05,
+      "loss": 0.4958,
+      "step": 2381
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9200040300236766e-05,
+      "loss": 0.3014,
+      "step": 2382
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.91997044649304e-05,
+      "loss": 0.2717,
+      "step": 2383
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9199368629624032e-05,
+      "loss": 0.4005,
+      "step": 2384
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919903279431767e-05,
+      "loss": 0.3241,
+      "step": 2385
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.91986969590113e-05,
+      "loss": 0.3517,
+      "step": 2386
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9198361123704936e-05,
+      "loss": 0.2503,
+      "step": 2387
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919802528839857e-05,
+      "loss": 0.3193,
+      "step": 2388
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9197689453092205e-05,
+      "loss": 0.2876,
+      "step": 2389
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919735361778584e-05,
+      "loss": 0.291,
+      "step": 2390
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9197017782479475e-05,
+      "loss": 0.4775,
+      "step": 2391
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919668194717311e-05,
+      "loss": 0.2059,
+      "step": 2392
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919634611186674e-05,
+      "loss": 0.262,
+      "step": 2393
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919601027656038e-05,
+      "loss": 0.4344,
+      "step": 2394
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919567444125401e-05,
+      "loss": 0.2961,
+      "step": 2395
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9195338605947644e-05,
+      "loss": 0.1928,
+      "step": 2396
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919500277064128e-05,
+      "loss": 0.2807,
+      "step": 2397
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9194666935334914e-05,
+      "loss": 0.3356,
+      "step": 2398
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919433110002855e-05,
+      "loss": 0.2517,
+      "step": 2399
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9193995264722183e-05,
+      "loss": 0.4268,
+      "step": 2400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9193659429415818e-05,
+      "loss": 0.3587,
+      "step": 2401
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919332359410945e-05,
+      "loss": 0.4695,
+      "step": 2402
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9192987758803087e-05,
+      "loss": 0.2101,
+      "step": 2403
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9192651923496718e-05,
+      "loss": 0.3096,
+      "step": 2404
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9192316088190353e-05,
+      "loss": 0.3648,
+      "step": 2405
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9191980252883987e-05,
+      "loss": 0.2609,
+      "step": 2406
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9191644417577622e-05,
+      "loss": 0.335,
+      "step": 2407
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9191308582271257e-05,
+      "loss": 0.2036,
+      "step": 2408
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.919097274696489e-05,
+      "loss": 0.4178,
+      "step": 2409
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9190636911658526e-05,
+      "loss": 0.4256,
+      "step": 2410
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9190301076352157e-05,
+      "loss": 0.1585,
+      "step": 2411
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9189965241045795e-05,
+      "loss": 0.3212,
+      "step": 2412
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9189629405739427e-05,
+      "loss": 0.409,
+      "step": 2413
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.918929357043306e-05,
+      "loss": 0.3706,
+      "step": 2414
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9188957735126696e-05,
+      "loss": 0.5037,
+      "step": 2415
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.918862189982033e-05,
+      "loss": 0.3206,
+      "step": 2416
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9188286064513962e-05,
+      "loss": 0.3706,
+      "step": 2417
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.91879502292076e-05,
+      "loss": 0.497,
+      "step": 2418
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.918761439390123e-05,
+      "loss": 0.5619,
+      "step": 2419
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9187278558594866e-05,
+      "loss": 0.4164,
+      "step": 2420
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.91869427232885e-05,
+      "loss": 0.2491,
+      "step": 2421
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9186606887982135e-05,
+      "loss": 0.2535,
+      "step": 2422
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.918627105267577e-05,
+      "loss": 0.2648,
+      "step": 2423
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9185935217369404e-05,
+      "loss": 0.4922,
+      "step": 2424
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.918559938206304e-05,
+      "loss": 0.5355,
+      "step": 2425
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.918526354675667e-05,
+      "loss": 0.4046,
+      "step": 2426
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9184927711450308e-05,
+      "loss": 0.3335,
+      "step": 2427
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.918459187614394e-05,
+      "loss": 0.5649,
+      "step": 2428
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9184256040837574e-05,
+      "loss": 0.2592,
+      "step": 2429
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.918392020553121e-05,
+      "loss": 0.5596,
+      "step": 2430
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9183584370224843e-05,
+      "loss": 0.3529,
+      "step": 2431
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9183248534918478e-05,
+      "loss": 0.3087,
+      "step": 2432
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9182912699612113e-05,
+      "loss": 0.4422,
+      "step": 2433
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9182576864305747e-05,
+      "loss": 0.3248,
+      "step": 2434
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.918224102899938e-05,
+      "loss": 0.3804,
+      "step": 2435
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9181905193693017e-05,
+      "loss": 0.2477,
+      "step": 2436
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9181569358386648e-05,
+      "loss": 0.4657,
+      "step": 2437
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9181233523080283e-05,
+      "loss": 0.4,
+      "step": 2438
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9180897687773917e-05,
+      "loss": 0.4486,
+      "step": 2439
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9180561852467552e-05,
+      "loss": 0.2444,
+      "step": 2440
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9180226017161187e-05,
+      "loss": 0.2487,
+      "step": 2441
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9179890181854818e-05,
+      "loss": 0.2853,
+      "step": 2442
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9179554346548456e-05,
+      "loss": 0.3379,
+      "step": 2443
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9179218511242087e-05,
+      "loss": 0.4427,
+      "step": 2444
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9178882675935722e-05,
+      "loss": 0.289,
+      "step": 2445
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9178546840629356e-05,
+      "loss": 0.4444,
+      "step": 2446
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.917821100532299e-05,
+      "loss": 0.2709,
+      "step": 2447
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9177875170016626e-05,
+      "loss": 0.2282,
+      "step": 2448
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.917753933471026e-05,
+      "loss": 0.3862,
+      "step": 2449
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9177203499403895e-05,
+      "loss": 0.3198,
+      "step": 2450
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9176867664097526e-05,
+      "loss": 0.4299,
+      "step": 2451
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9176531828791164e-05,
+      "loss": 0.2369,
+      "step": 2452
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9176195993484796e-05,
+      "loss": 0.2727,
+      "step": 2453
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.917586015817843e-05,
+      "loss": 0.2365,
+      "step": 2454
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9175524322872065e-05,
+      "loss": 0.5149,
+      "step": 2455
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.91751884875657e-05,
+      "loss": 0.5552,
+      "step": 2456
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9174852652259334e-05,
+      "loss": 0.3454,
+      "step": 2457
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.917451681695297e-05,
+      "loss": 0.4649,
+      "step": 2458
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9174180981646603e-05,
+      "loss": 0.6107,
+      "step": 2459
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9173845146340235e-05,
+      "loss": 0.3735,
+      "step": 2460
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.917350931103387e-05,
+      "loss": 0.2442,
+      "step": 2461
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9173173475727504e-05,
+      "loss": 0.5397,
+      "step": 2462
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.917283764042114e-05,
+      "loss": 0.3636,
+      "step": 2463
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9172501805114773e-05,
+      "loss": 0.3616,
+      "step": 2464
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9172165969808408e-05,
+      "loss": 0.4292,
+      "step": 2465
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.917183013450204e-05,
+      "loss": 0.3668,
+      "step": 2466
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9171494299195677e-05,
+      "loss": 0.4296,
+      "step": 2467
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.917115846388931e-05,
+      "loss": 0.3421,
+      "step": 2468
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9170822628582943e-05,
+      "loss": 0.3218,
+      "step": 2469
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9170486793276578e-05,
+      "loss": 0.4818,
+      "step": 2470
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9170150957970212e-05,
+      "loss": 0.4213,
+      "step": 2471
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9169815122663847e-05,
+      "loss": 0.5159,
+      "step": 2472
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9169479287357482e-05,
+      "loss": 0.4277,
+      "step": 2473
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9169143452051116e-05,
+      "loss": 0.3764,
+      "step": 2474
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9168807616744748e-05,
+      "loss": 0.2809,
+      "step": 2475
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9168471781438386e-05,
+      "loss": 0.309,
+      "step": 2476
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9168135946132017e-05,
+      "loss": 0.2875,
+      "step": 2477
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.916780011082565e-05,
+      "loss": 0.3813,
+      "step": 2478
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9167464275519286e-05,
+      "loss": 0.4481,
+      "step": 2479
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.916712844021292e-05,
+      "loss": 0.2905,
+      "step": 2480
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9166792604906556e-05,
+      "loss": 0.3016,
+      "step": 2481
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916645676960019e-05,
+      "loss": 0.4037,
+      "step": 2482
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9166120934293825e-05,
+      "loss": 0.2988,
+      "step": 2483
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9165785098987456e-05,
+      "loss": 0.2405,
+      "step": 2484
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9165449263681094e-05,
+      "loss": 0.2964,
+      "step": 2485
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9165113428374725e-05,
+      "loss": 0.3246,
+      "step": 2486
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916477759306836e-05,
+      "loss": 0.4519,
+      "step": 2487
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9164441757761995e-05,
+      "loss": 0.403,
+      "step": 2488
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916410592245563e-05,
+      "loss": 0.3224,
+      "step": 2489
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9163770087149264e-05,
+      "loss": 0.3759,
+      "step": 2490
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.91634342518429e-05,
+      "loss": 0.3264,
+      "step": 2491
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9163098416536533e-05,
+      "loss": 0.2195,
+      "step": 2492
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9162762581230165e-05,
+      "loss": 0.3328,
+      "step": 2493
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9162426745923803e-05,
+      "loss": 0.4696,
+      "step": 2494
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9162090910617434e-05,
+      "loss": 0.2822,
+      "step": 2495
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.916175507531107e-05,
+      "loss": 0.3854,
+      "step": 2496
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9161419240004703e-05,
+      "loss": 0.3853,
+      "step": 2497
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9161083404698338e-05,
+      "loss": 0.4902,
+      "step": 2498
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9160747569391972e-05,
+      "loss": 0.3723,
+      "step": 2499
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9160411734085607e-05,
+      "loss": 0.3804,
+      "step": 2500
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9160075898779242e-05,
+      "loss": 0.1804,
+      "step": 2501
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9159740063472873e-05,
+      "loss": 0.2315,
+      "step": 2502
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915940422816651e-05,
+      "loss": 0.2738,
+      "step": 2503
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9159068392860142e-05,
+      "loss": 0.1985,
+      "step": 2504
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9158732557553777e-05,
+      "loss": 0.3055,
+      "step": 2505
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915839672224741e-05,
+      "loss": 0.3497,
+      "step": 2506
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9158060886941046e-05,
+      "loss": 0.3372,
+      "step": 2507
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9157725051634677e-05,
+      "loss": 0.3479,
+      "step": 2508
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9157389216328315e-05,
+      "loss": 0.2296,
+      "step": 2509
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9157053381021947e-05,
+      "loss": 0.5346,
+      "step": 2510
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915671754571558e-05,
+      "loss": 0.4813,
+      "step": 2511
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9156381710409216e-05,
+      "loss": 0.319,
+      "step": 2512
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915604587510285e-05,
+      "loss": 0.2421,
+      "step": 2513
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9155710039796485e-05,
+      "loss": 0.4693,
+      "step": 2514
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915537420449012e-05,
+      "loss": 0.4023,
+      "step": 2515
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9155038369183755e-05,
+      "loss": 0.2582,
+      "step": 2516
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9154702533877386e-05,
+      "loss": 0.3663,
+      "step": 2517
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9154366698571024e-05,
+      "loss": 0.5775,
+      "step": 2518
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9154030863264655e-05,
+      "loss": 0.4316,
+      "step": 2519
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915369502795829e-05,
+      "loss": 0.3014,
+      "step": 2520
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9153359192651924e-05,
+      "loss": 0.3757,
+      "step": 2521
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915302335734556e-05,
+      "loss": 0.4459,
+      "step": 2522
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9152687522039194e-05,
+      "loss": 0.5346,
+      "step": 2523
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.915235168673283e-05,
+      "loss": 0.2842,
+      "step": 2524
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9152015851426463e-05,
+      "loss": 0.1411,
+      "step": 2525
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9151680016120094e-05,
+      "loss": 0.3348,
+      "step": 2526
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9151344180813732e-05,
+      "loss": 0.3303,
+      "step": 2527
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9151008345507364e-05,
+      "loss": 0.4145,
+      "step": 2528
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9150672510200998e-05,
+      "loss": 0.2864,
+      "step": 2529
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9150336674894633e-05,
+      "loss": 0.3584,
+      "step": 2530
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9150000839588268e-05,
+      "loss": 0.4693,
+      "step": 2531
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9149665004281902e-05,
+      "loss": 0.3073,
+      "step": 2532
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9149329168975537e-05,
+      "loss": 0.7349,
+      "step": 2533
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914899333366917e-05,
+      "loss": 0.4424,
+      "step": 2534
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9148657498362803e-05,
+      "loss": 0.4256,
+      "step": 2535
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914832166305644e-05,
+      "loss": 0.3542,
+      "step": 2536
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9147985827750072e-05,
+      "loss": 0.2231,
+      "step": 2537
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9147649992443707e-05,
+      "loss": 0.2896,
+      "step": 2538
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914731415713734e-05,
+      "loss": 0.4584,
+      "step": 2539
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9146978321830976e-05,
+      "loss": 0.4078,
+      "step": 2540
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914664248652461e-05,
+      "loss": 0.5222,
+      "step": 2541
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9146306651218245e-05,
+      "loss": 0.3365,
+      "step": 2542
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914597081591188e-05,
+      "loss": 0.3391,
+      "step": 2543
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914563498060551e-05,
+      "loss": 0.2039,
+      "step": 2544
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914529914529915e-05,
+      "loss": 0.3548,
+      "step": 2545
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914496330999278e-05,
+      "loss": 0.3247,
+      "step": 2546
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9144627474686415e-05,
+      "loss": 0.2344,
+      "step": 2547
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914429163938005e-05,
+      "loss": 0.3371,
+      "step": 2548
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9143955804073684e-05,
+      "loss": 0.4298,
+      "step": 2549
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914361996876732e-05,
+      "loss": 0.3435,
+      "step": 2550
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9143284133460954e-05,
+      "loss": 0.2867,
+      "step": 2551
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9142948298154585e-05,
+      "loss": 0.3514,
+      "step": 2552
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914261246284822e-05,
+      "loss": 0.3485,
+      "step": 2553
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9142276627541854e-05,
+      "loss": 0.1999,
+      "step": 2554
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.914194079223549e-05,
+      "loss": 0.5657,
+      "step": 2555
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9141604956929124e-05,
+      "loss": 0.3134,
+      "step": 2556
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9141269121622758e-05,
+      "loss": 0.3681,
+      "step": 2557
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9140933286316393e-05,
+      "loss": 0.3144,
+      "step": 2558
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9140597451010024e-05,
+      "loss": 0.4161,
+      "step": 2559
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9140261615703662e-05,
+      "loss": 0.2965,
+      "step": 2560
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9139925780397293e-05,
+      "loss": 0.4539,
+      "step": 2561
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9139589945090928e-05,
+      "loss": 0.3306,
+      "step": 2562
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9139254109784563e-05,
+      "loss": 0.3232,
+      "step": 2563
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9138918274478197e-05,
+      "loss": 0.3041,
+      "step": 2564
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9138582439171832e-05,
+      "loss": 0.2925,
+      "step": 2565
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9138246603865467e-05,
+      "loss": 0.2702,
+      "step": 2566
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.91379107685591e-05,
+      "loss": 0.3632,
+      "step": 2567
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9137574933252733e-05,
+      "loss": 0.4554,
+      "step": 2568
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913723909794637e-05,
+      "loss": 0.3141,
+      "step": 2569
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9136903262640002e-05,
+      "loss": 0.4891,
+      "step": 2570
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9136567427333637e-05,
+      "loss": 0.5063,
+      "step": 2571
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913623159202727e-05,
+      "loss": 0.3378,
+      "step": 2572
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9135895756720906e-05,
+      "loss": 0.2656,
+      "step": 2573
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913555992141454e-05,
+      "loss": 0.4337,
+      "step": 2574
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9135224086108175e-05,
+      "loss": 0.5193,
+      "step": 2575
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913488825080181e-05,
+      "loss": 0.3051,
+      "step": 2576
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913455241549544e-05,
+      "loss": 0.3278,
+      "step": 2577
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913421658018908e-05,
+      "loss": 0.4048,
+      "step": 2578
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913388074488271e-05,
+      "loss": 0.3803,
+      "step": 2579
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9133544909576345e-05,
+      "loss": 0.3215,
+      "step": 2580
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913320907426998e-05,
+      "loss": 0.3039,
+      "step": 2581
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9132873238963614e-05,
+      "loss": 0.26,
+      "step": 2582
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913253740365725e-05,
+      "loss": 0.2636,
+      "step": 2583
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9132201568350884e-05,
+      "loss": 0.4224,
+      "step": 2584
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9131865733044518e-05,
+      "loss": 0.3956,
+      "step": 2585
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913152989773815e-05,
+      "loss": 0.4398,
+      "step": 2586
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9131194062431787e-05,
+      "loss": 0.3622,
+      "step": 2587
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.913085822712542e-05,
+      "loss": 0.3683,
+      "step": 2588
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9130522391819053e-05,
+      "loss": 0.5504,
+      "step": 2589
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9130186556512688e-05,
+      "loss": 0.3906,
+      "step": 2590
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9129850721206323e-05,
+      "loss": 0.5318,
+      "step": 2591
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9129514885899957e-05,
+      "loss": 0.3631,
+      "step": 2592
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9129179050593592e-05,
+      "loss": 0.5164,
+      "step": 2593
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9128843215287227e-05,
+      "loss": 0.4114,
+      "step": 2594
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9128507379980858e-05,
+      "loss": 0.2344,
+      "step": 2595
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9128171544674493e-05,
+      "loss": 0.3319,
+      "step": 2596
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9127835709368127e-05,
+      "loss": 0.5157,
+      "step": 2597
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9127499874061762e-05,
+      "loss": 0.5108,
+      "step": 2598
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9127164038755396e-05,
+      "loss": 0.2944,
+      "step": 2599
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912682820344903e-05,
+      "loss": 0.3918,
+      "step": 2600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9126492368142662e-05,
+      "loss": 0.5614,
+      "step": 2601
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.91261565328363e-05,
+      "loss": 0.3102,
+      "step": 2602
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912582069752993e-05,
+      "loss": 0.4741,
+      "step": 2603
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9125484862223566e-05,
+      "loss": 0.2431,
+      "step": 2604
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.91251490269172e-05,
+      "loss": 0.2595,
+      "step": 2605
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9124813191610836e-05,
+      "loss": 0.3543,
+      "step": 2606
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912447735630447e-05,
+      "loss": 0.5218,
+      "step": 2607
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9124141520998105e-05,
+      "loss": 0.4543,
+      "step": 2608
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912380568569174e-05,
+      "loss": 0.3163,
+      "step": 2609
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912346985038537e-05,
+      "loss": 0.2888,
+      "step": 2610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912313401507901e-05,
+      "loss": 0.474,
+      "step": 2611
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912279817977264e-05,
+      "loss": 0.4868,
+      "step": 2612
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9122462344466275e-05,
+      "loss": 0.4319,
+      "step": 2613
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912212650915991e-05,
+      "loss": 0.4681,
+      "step": 2614
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9121790673853544e-05,
+      "loss": 0.3241,
+      "step": 2615
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912145483854718e-05,
+      "loss": 0.3168,
+      "step": 2616
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912111900324081e-05,
+      "loss": 0.4306,
+      "step": 2617
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9120783167934448e-05,
+      "loss": 0.3158,
+      "step": 2618
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.912044733262808e-05,
+      "loss": 0.3583,
+      "step": 2619
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9120111497321714e-05,
+      "loss": 0.4039,
+      "step": 2620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.911977566201535e-05,
+      "loss": 0.2776,
+      "step": 2621
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9119439826708983e-05,
+      "loss": 0.3857,
+      "step": 2622
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9119103991402618e-05,
+      "loss": 0.3307,
+      "step": 2623
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9118768156096253e-05,
+      "loss": 0.2349,
+      "step": 2624
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9118432320789887e-05,
+      "loss": 0.3149,
+      "step": 2625
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.911809648548352e-05,
+      "loss": 0.4841,
+      "step": 2626
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9117760650177156e-05,
+      "loss": 0.3585,
+      "step": 2627
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9117424814870788e-05,
+      "loss": 0.3303,
+      "step": 2628
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9117088979564422e-05,
+      "loss": 0.4823,
+      "step": 2629
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9116753144258057e-05,
+      "loss": 0.2493,
+      "step": 2630
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.911641730895169e-05,
+      "loss": 0.3307,
+      "step": 2631
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9116081473645326e-05,
+      "loss": 0.2693,
+      "step": 2632
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.911574563833896e-05,
+      "loss": 0.2176,
+      "step": 2633
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9115409803032596e-05,
+      "loss": 0.262,
+      "step": 2634
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9115073967726227e-05,
+      "loss": 0.4855,
+      "step": 2635
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9114738132419865e-05,
+      "loss": 0.3292,
+      "step": 2636
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9114402297113496e-05,
+      "loss": 0.3473,
+      "step": 2637
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.911406646180713e-05,
+      "loss": 0.2427,
+      "step": 2638
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9113730626500765e-05,
+      "loss": 0.4063,
+      "step": 2639
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.91133947911944e-05,
+      "loss": 0.4592,
+      "step": 2640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9113058955888035e-05,
+      "loss": 0.5798,
+      "step": 2641
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.911272312058167e-05,
+      "loss": 0.4987,
+      "step": 2642
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.91123872852753e-05,
+      "loss": 0.4201,
+      "step": 2643
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9112051449968935e-05,
+      "loss": 0.4026,
+      "step": 2644
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.911171561466257e-05,
+      "loss": 0.3321,
+      "step": 2645
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9111379779356205e-05,
+      "loss": 0.2337,
+      "step": 2646
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.911104394404984e-05,
+      "loss": 0.4446,
+      "step": 2647
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9110708108743474e-05,
+      "loss": 0.2812,
+      "step": 2648
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.911037227343711e-05,
+      "loss": 0.3136,
+      "step": 2649
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.911003643813074e-05,
+      "loss": 0.3762,
+      "step": 2650
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9109700602824378e-05,
+      "loss": 0.2971,
+      "step": 2651
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.910936476751801e-05,
+      "loss": 0.3833,
+      "step": 2652
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9109028932211644e-05,
+      "loss": 0.3191,
+      "step": 2653
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.910869309690528e-05,
+      "loss": 0.5311,
+      "step": 2654
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9108357261598913e-05,
+      "loss": 0.394,
+      "step": 2655
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9108021426292548e-05,
+      "loss": 0.307,
+      "step": 2656
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9107685590986182e-05,
+      "loss": 0.4819,
+      "step": 2657
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9107349755679817e-05,
+      "loss": 0.3446,
+      "step": 2658
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9107013920373448e-05,
+      "loss": 0.495,
+      "step": 2659
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9106678085067086e-05,
+      "loss": 0.2702,
+      "step": 2660
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9106342249760718e-05,
+      "loss": 0.5004,
+      "step": 2661
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9106006414454352e-05,
+      "loss": 0.3461,
+      "step": 2662
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9105670579147987e-05,
+      "loss": 0.5335,
+      "step": 2663
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.910533474384162e-05,
+      "loss": 0.3686,
+      "step": 2664
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9104998908535256e-05,
+      "loss": 0.3782,
+      "step": 2665
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.910466307322889e-05,
+      "loss": 0.3671,
+      "step": 2666
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9104327237922525e-05,
+      "loss": 0.2946,
+      "step": 2667
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9103991402616157e-05,
+      "loss": 0.372,
+      "step": 2668
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9103655567309795e-05,
+      "loss": 0.257,
+      "step": 2669
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9103319732003426e-05,
+      "loss": 0.4083,
+      "step": 2670
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.910298389669706e-05,
+      "loss": 0.3581,
+      "step": 2671
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9102648061390695e-05,
+      "loss": 0.2763,
+      "step": 2672
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.910231222608433e-05,
+      "loss": 0.3216,
+      "step": 2673
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9101976390777965e-05,
+      "loss": 0.2362,
+      "step": 2674
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.91016405554716e-05,
+      "loss": 0.381,
+      "step": 2675
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9101304720165234e-05,
+      "loss": 0.4501,
+      "step": 2676
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9100968884858865e-05,
+      "loss": 0.4541,
+      "step": 2677
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9100633049552503e-05,
+      "loss": 0.3295,
+      "step": 2678
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9100297214246134e-05,
+      "loss": 0.2689,
+      "step": 2679
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909996137893977e-05,
+      "loss": 0.4694,
+      "step": 2680
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9099625543633404e-05,
+      "loss": 0.2739,
+      "step": 2681
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909928970832704e-05,
+      "loss": 0.245,
+      "step": 2682
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9098953873020673e-05,
+      "loss": 0.2882,
+      "step": 2683
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9098618037714308e-05,
+      "loss": 0.3119,
+      "step": 2684
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9098282202407942e-05,
+      "loss": 0.4112,
+      "step": 2685
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9097946367101574e-05,
+      "loss": 0.4643,
+      "step": 2686
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9097610531795208e-05,
+      "loss": 0.3987,
+      "step": 2687
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9097274696488843e-05,
+      "loss": 0.4015,
+      "step": 2688
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9096938861182477e-05,
+      "loss": 0.2455,
+      "step": 2689
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9096603025876112e-05,
+      "loss": 0.4356,
+      "step": 2690
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9096267190569747e-05,
+      "loss": 0.2889,
+      "step": 2691
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9095931355263378e-05,
+      "loss": 0.2721,
+      "step": 2692
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9095595519957016e-05,
+      "loss": 0.3474,
+      "step": 2693
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9095259684650647e-05,
+      "loss": 0.4437,
+      "step": 2694
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9094923849344282e-05,
+      "loss": 0.4801,
+      "step": 2695
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9094588014037917e-05,
+      "loss": 0.3761,
+      "step": 2696
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909425217873155e-05,
+      "loss": 0.4175,
+      "step": 2697
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9093916343425186e-05,
+      "loss": 0.599,
+      "step": 2698
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909358050811882e-05,
+      "loss": 0.3225,
+      "step": 2699
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9093244672812455e-05,
+      "loss": 0.3772,
+      "step": 2700
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9092908837506086e-05,
+      "loss": 0.4133,
+      "step": 2701
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9092573002199725e-05,
+      "loss": 0.4376,
+      "step": 2702
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9092237166893356e-05,
+      "loss": 0.2549,
+      "step": 2703
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909190133158699e-05,
+      "loss": 0.3021,
+      "step": 2704
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9091565496280625e-05,
+      "loss": 0.3651,
+      "step": 2705
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909122966097426e-05,
+      "loss": 0.3347,
+      "step": 2706
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9090893825667894e-05,
+      "loss": 0.4335,
+      "step": 2707
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.909055799036153e-05,
+      "loss": 0.2799,
+      "step": 2708
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9090222155055164e-05,
+      "loss": 0.4253,
+      "step": 2709
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9089886319748795e-05,
+      "loss": 0.3686,
+      "step": 2710
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9089550484442433e-05,
+      "loss": 0.3244,
+      "step": 2711
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9089214649136064e-05,
+      "loss": 0.5053,
+      "step": 2712
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.90888788138297e-05,
+      "loss": 0.2915,
+      "step": 2713
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9088542978523333e-05,
+      "loss": 0.4559,
+      "step": 2714
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9088207143216968e-05,
+      "loss": 0.2736,
+      "step": 2715
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9087871307910603e-05,
+      "loss": 0.4036,
+      "step": 2716
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9087535472604237e-05,
+      "loss": 0.389,
+      "step": 2717
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9087199637297872e-05,
+      "loss": 0.3963,
+      "step": 2718
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9086863801991503e-05,
+      "loss": 0.3228,
+      "step": 2719
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.908652796668514e-05,
+      "loss": 0.1838,
+      "step": 2720
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9086192131378773e-05,
+      "loss": 0.4051,
+      "step": 2721
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9085856296072407e-05,
+      "loss": 0.3981,
+      "step": 2722
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9085520460766042e-05,
+      "loss": 0.4388,
+      "step": 2723
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9085184625459677e-05,
+      "loss": 0.2275,
+      "step": 2724
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.908484879015331e-05,
+      "loss": 0.3978,
+      "step": 2725
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9084512954846946e-05,
+      "loss": 0.4155,
+      "step": 2726
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.908417711954058e-05,
+      "loss": 0.4589,
+      "step": 2727
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9083841284234212e-05,
+      "loss": 0.4001,
+      "step": 2728
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.908350544892785e-05,
+      "loss": 0.2396,
+      "step": 2729
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.908316961362148e-05,
+      "loss": 0.2399,
+      "step": 2730
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9082833778315116e-05,
+      "loss": 0.3012,
+      "step": 2731
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.908249794300875e-05,
+      "loss": 0.4002,
+      "step": 2732
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9082162107702385e-05,
+      "loss": 0.2667,
+      "step": 2733
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9081826272396016e-05,
+      "loss": 0.3032,
+      "step": 2734
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9081490437089654e-05,
+      "loss": 0.397,
+      "step": 2735
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9081154601783286e-05,
+      "loss": 0.3527,
+      "step": 2736
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.908081876647692e-05,
+      "loss": 0.3738,
+      "step": 2737
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9080482931170555e-05,
+      "loss": 0.4024,
+      "step": 2738
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.908014709586419e-05,
+      "loss": 0.4723,
+      "step": 2739
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9079811260557824e-05,
+      "loss": 0.3489,
+      "step": 2740
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907947542525146e-05,
+      "loss": 0.2893,
+      "step": 2741
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9079139589945093e-05,
+      "loss": 0.3843,
+      "step": 2742
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9078803754638725e-05,
+      "loss": 0.3341,
+      "step": 2743
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9078467919332363e-05,
+      "loss": 0.4261,
+      "step": 2744
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9078132084025994e-05,
+      "loss": 0.4161,
+      "step": 2745
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907779624871963e-05,
+      "loss": 0.2983,
+      "step": 2746
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9077460413413263e-05,
+      "loss": 0.1956,
+      "step": 2747
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9077124578106898e-05,
+      "loss": 0.3579,
+      "step": 2748
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9076788742800533e-05,
+      "loss": 0.2568,
+      "step": 2749
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9076452907494167e-05,
+      "loss": 0.3093,
+      "step": 2750
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9076117072187802e-05,
+      "loss": 0.2963,
+      "step": 2751
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9075781236881433e-05,
+      "loss": 0.3643,
+      "step": 2752
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907544540157507e-05,
+      "loss": 0.5026,
+      "step": 2753
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9075109566268702e-05,
+      "loss": 0.4893,
+      "step": 2754
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9074773730962337e-05,
+      "loss": 0.3533,
+      "step": 2755
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9074437895655972e-05,
+      "loss": 0.5654,
+      "step": 2756
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9074102060349606e-05,
+      "loss": 0.3128,
+      "step": 2757
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907376622504324e-05,
+      "loss": 0.3462,
+      "step": 2758
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9073430389736876e-05,
+      "loss": 0.3741,
+      "step": 2759
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907309455443051e-05,
+      "loss": 0.3114,
+      "step": 2760
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907275871912414e-05,
+      "loss": 0.3997,
+      "step": 2761
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907242288381778e-05,
+      "loss": 0.2916,
+      "step": 2762
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907208704851141e-05,
+      "loss": 0.5075,
+      "step": 2763
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9071751213205046e-05,
+      "loss": 0.3728,
+      "step": 2764
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907141537789868e-05,
+      "loss": 0.3173,
+      "step": 2765
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9071079542592315e-05,
+      "loss": 0.4422,
+      "step": 2766
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907074370728595e-05,
+      "loss": 0.3042,
+      "step": 2767
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9070407871979584e-05,
+      "loss": 0.3837,
+      "step": 2768
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.907007203667322e-05,
+      "loss": 0.4268,
+      "step": 2769
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.906973620136685e-05,
+      "loss": 0.22,
+      "step": 2770
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9069400366060488e-05,
+      "loss": 0.3205,
+      "step": 2771
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.906906453075412e-05,
+      "loss": 0.3908,
+      "step": 2772
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9068728695447754e-05,
+      "loss": 0.4113,
+      "step": 2773
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.906839286014139e-05,
+      "loss": 0.5416,
+      "step": 2774
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9068057024835023e-05,
+      "loss": 0.3189,
+      "step": 2775
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9067721189528658e-05,
+      "loss": 0.2368,
+      "step": 2776
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9067385354222293e-05,
+      "loss": 0.6408,
+      "step": 2777
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9067049518915924e-05,
+      "loss": 0.3213,
+      "step": 2778
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.906671368360956e-05,
+      "loss": 0.158,
+      "step": 2779
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9066377848303193e-05,
+      "loss": 0.2192,
+      "step": 2780
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9066042012996828e-05,
+      "loss": 0.478,
+      "step": 2781
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9065706177690462e-05,
+      "loss": 0.3145,
+      "step": 2782
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9065370342384097e-05,
+      "loss": 0.388,
+      "step": 2783
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9065034507077732e-05,
+      "loss": 0.2014,
+      "step": 2784
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9064698671771363e-05,
+      "loss": 0.2182,
+      "step": 2785
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9064362836465e-05,
+      "loss": 0.3773,
+      "step": 2786
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9064027001158632e-05,
+      "loss": 0.1956,
+      "step": 2787
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9063691165852267e-05,
+      "loss": 0.2947,
+      "step": 2788
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.90633553305459e-05,
+      "loss": 0.5165,
+      "step": 2789
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9063019495239536e-05,
+      "loss": 0.399,
+      "step": 2790
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.906268365993317e-05,
+      "loss": 0.4562,
+      "step": 2791
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9062347824626806e-05,
+      "loss": 0.3043,
+      "step": 2792
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.906201198932044e-05,
+      "loss": 0.3043,
+      "step": 2793
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.906167615401407e-05,
+      "loss": 0.5004,
+      "step": 2794
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.906134031870771e-05,
+      "loss": 0.3191,
+      "step": 2795
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.906100448340134e-05,
+      "loss": 0.2503,
+      "step": 2796
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9060668648094975e-05,
+      "loss": 0.4286,
+      "step": 2797
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.906033281278861e-05,
+      "loss": 0.3148,
+      "step": 2798
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9059996977482245e-05,
+      "loss": 0.4579,
+      "step": 2799
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.905966114217588e-05,
+      "loss": 0.2416,
+      "step": 2800
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.905932530686951e-05,
+      "loss": 0.507,
+      "step": 2801
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.905898947156315e-05,
+      "loss": 0.4208,
+      "step": 2802
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.905865363625678e-05,
+      "loss": 0.5283,
+      "step": 2803
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9058317800950414e-05,
+      "loss": 0.3309,
+      "step": 2804
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.905798196564405e-05,
+      "loss": 0.2949,
+      "step": 2805
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9057646130337684e-05,
+      "loss": 0.4774,
+      "step": 2806
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.905731029503132e-05,
+      "loss": 0.26,
+      "step": 2807
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9056974459724953e-05,
+      "loss": 0.3947,
+      "step": 2808
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9056638624418588e-05,
+      "loss": 0.3915,
+      "step": 2809
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.905630278911222e-05,
+      "loss": 0.2462,
+      "step": 2810
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9055966953805857e-05,
+      "loss": 0.2997,
+      "step": 2811
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9055631118499488e-05,
+      "loss": 0.2745,
+      "step": 2812
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9055295283193123e-05,
+      "loss": 0.391,
+      "step": 2813
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9054959447886758e-05,
+      "loss": 0.2279,
+      "step": 2814
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9054623612580392e-05,
+      "loss": 0.5232,
+      "step": 2815
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9054287777274027e-05,
+      "loss": 0.4011,
+      "step": 2816
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.905395194196766e-05,
+      "loss": 0.5854,
+      "step": 2817
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9053616106661296e-05,
+      "loss": 0.1943,
+      "step": 2818
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9053280271354927e-05,
+      "loss": 0.5427,
+      "step": 2819
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9052944436048565e-05,
+      "loss": 0.3027,
+      "step": 2820
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9052608600742197e-05,
+      "loss": 0.3679,
+      "step": 2821
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.905227276543583e-05,
+      "loss": 0.2534,
+      "step": 2822
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9051936930129466e-05,
+      "loss": 0.3087,
+      "step": 2823
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.90516010948231e-05,
+      "loss": 0.4428,
+      "step": 2824
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9051265259516732e-05,
+      "loss": 0.4432,
+      "step": 2825
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.905092942421037e-05,
+      "loss": 0.3936,
+      "step": 2826
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9050593588904e-05,
+      "loss": 0.4218,
+      "step": 2827
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9050257753597636e-05,
+      "loss": 0.3649,
+      "step": 2828
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904992191829127e-05,
+      "loss": 0.3586,
+      "step": 2829
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9049586082984905e-05,
+      "loss": 0.435,
+      "step": 2830
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904925024767854e-05,
+      "loss": 0.1838,
+      "step": 2831
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9048914412372174e-05,
+      "loss": 0.347,
+      "step": 2832
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904857857706581e-05,
+      "loss": 0.3122,
+      "step": 2833
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904824274175944e-05,
+      "loss": 0.4394,
+      "step": 2834
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904790690645308e-05,
+      "loss": 0.3951,
+      "step": 2835
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904757107114671e-05,
+      "loss": 0.3157,
+      "step": 2836
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9047235235840344e-05,
+      "loss": 0.3048,
+      "step": 2837
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904689940053398e-05,
+      "loss": 0.4166,
+      "step": 2838
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9046563565227614e-05,
+      "loss": 0.4657,
+      "step": 2839
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9046227729921248e-05,
+      "loss": 0.3888,
+      "step": 2840
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9045891894614883e-05,
+      "loss": 0.3199,
+      "step": 2841
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9045556059308518e-05,
+      "loss": 0.3794,
+      "step": 2842
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904522022400215e-05,
+      "loss": 0.3082,
+      "step": 2843
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9044884388695787e-05,
+      "loss": 0.3366,
+      "step": 2844
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9044548553389418e-05,
+      "loss": 0.3207,
+      "step": 2845
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9044212718083053e-05,
+      "loss": 0.364,
+      "step": 2846
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9043876882776687e-05,
+      "loss": 0.4668,
+      "step": 2847
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9043541047470322e-05,
+      "loss": 0.3324,
+      "step": 2848
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9043205212163957e-05,
+      "loss": 0.4841,
+      "step": 2849
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904286937685759e-05,
+      "loss": 0.483,
+      "step": 2850
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9042533541551226e-05,
+      "loss": 0.3181,
+      "step": 2851
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9042197706244857e-05,
+      "loss": 0.1776,
+      "step": 2852
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9041861870938495e-05,
+      "loss": 0.5249,
+      "step": 2853
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9041526035632127e-05,
+      "loss": 0.4042,
+      "step": 2854
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904119020032576e-05,
+      "loss": 0.481,
+      "step": 2855
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9040854365019396e-05,
+      "loss": 0.3858,
+      "step": 2856
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.904051852971303e-05,
+      "loss": 0.4656,
+      "step": 2857
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9040182694406665e-05,
+      "loss": 0.3611,
+      "step": 2858
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.90398468591003e-05,
+      "loss": 0.333,
+      "step": 2859
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9039511023793934e-05,
+      "loss": 0.3067,
+      "step": 2860
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9039175188487566e-05,
+      "loss": 0.3578,
+      "step": 2861
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9038839353181204e-05,
+      "loss": 0.2691,
+      "step": 2862
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9038503517874835e-05,
+      "loss": 0.3071,
+      "step": 2863
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.903816768256847e-05,
+      "loss": 0.278,
+      "step": 2864
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9037831847262104e-05,
+      "loss": 0.6098,
+      "step": 2865
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.903749601195574e-05,
+      "loss": 0.3386,
+      "step": 2866
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9037160176649374e-05,
+      "loss": 0.4406,
+      "step": 2867
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9036824341343008e-05,
+      "loss": 0.2634,
+      "step": 2868
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.903648850603664e-05,
+      "loss": 0.2768,
+      "step": 2869
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9036152670730274e-05,
+      "loss": 0.3699,
+      "step": 2870
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.903581683542391e-05,
+      "loss": 0.4188,
+      "step": 2871
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9035481000117543e-05,
+      "loss": 0.4644,
+      "step": 2872
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9035145164811178e-05,
+      "loss": 0.4437,
+      "step": 2873
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9034809329504813e-05,
+      "loss": 0.3142,
+      "step": 2874
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9034473494198447e-05,
+      "loss": 0.2796,
+      "step": 2875
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.903413765889208e-05,
+      "loss": 0.3856,
+      "step": 2876
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9033801823585717e-05,
+      "loss": 0.4312,
+      "step": 2877
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9033465988279348e-05,
+      "loss": 0.3466,
+      "step": 2878
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9033130152972983e-05,
+      "loss": 0.3016,
+      "step": 2879
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9032794317666617e-05,
+      "loss": 0.2932,
+      "step": 2880
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9032458482360252e-05,
+      "loss": 0.27,
+      "step": 2881
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9032122647053886e-05,
+      "loss": 0.4285,
+      "step": 2882
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.903178681174752e-05,
+      "loss": 0.3406,
+      "step": 2883
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9031450976441156e-05,
+      "loss": 0.4334,
+      "step": 2884
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9031115141134787e-05,
+      "loss": 0.3369,
+      "step": 2885
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9030779305828425e-05,
+      "loss": 0.2156,
+      "step": 2886
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9030443470522056e-05,
+      "loss": 0.1822,
+      "step": 2887
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.903010763521569e-05,
+      "loss": 0.3764,
+      "step": 2888
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9029771799909326e-05,
+      "loss": 0.4499,
+      "step": 2889
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.902943596460296e-05,
+      "loss": 0.3001,
+      "step": 2890
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9029100129296595e-05,
+      "loss": 0.319,
+      "step": 2891
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.902876429399023e-05,
+      "loss": 0.313,
+      "step": 2892
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9028428458683864e-05,
+      "loss": 0.2728,
+      "step": 2893
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9028092623377495e-05,
+      "loss": 0.577,
+      "step": 2894
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9027756788071134e-05,
+      "loss": 0.3082,
+      "step": 2895
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9027420952764765e-05,
+      "loss": 0.2569,
+      "step": 2896
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.90270851174584e-05,
+      "loss": 0.4775,
+      "step": 2897
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9026749282152034e-05,
+      "loss": 0.4403,
+      "step": 2898
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.902641344684567e-05,
+      "loss": 0.2942,
+      "step": 2899
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9026077611539303e-05,
+      "loss": 0.2618,
+      "step": 2900
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9025741776232938e-05,
+      "loss": 0.2616,
+      "step": 2901
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9025405940926573e-05,
+      "loss": 0.2872,
+      "step": 2902
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9025070105620204e-05,
+      "loss": 0.5198,
+      "step": 2903
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9024734270313842e-05,
+      "loss": 0.3167,
+      "step": 2904
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9024398435007473e-05,
+      "loss": 0.305,
+      "step": 2905
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9024062599701108e-05,
+      "loss": 0.3438,
+      "step": 2906
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9023726764394743e-05,
+      "loss": 0.2947,
+      "step": 2907
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9023390929088377e-05,
+      "loss": 0.2439,
+      "step": 2908
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9023055093782012e-05,
+      "loss": 0.3025,
+      "step": 2909
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9022719258475646e-05,
+      "loss": 0.2034,
+      "step": 2910
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.902238342316928e-05,
+      "loss": 0.3826,
+      "step": 2911
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9022047587862912e-05,
+      "loss": 0.3733,
+      "step": 2912
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9021711752556547e-05,
+      "loss": 0.3867,
+      "step": 2913
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.902137591725018e-05,
+      "loss": 0.4605,
+      "step": 2914
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9021040081943816e-05,
+      "loss": 0.3747,
+      "step": 2915
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.902070424663745e-05,
+      "loss": 0.3938,
+      "step": 2916
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9020368411331086e-05,
+      "loss": 0.48,
+      "step": 2917
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9020032576024717e-05,
+      "loss": 0.2723,
+      "step": 2918
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9019696740718355e-05,
+      "loss": 0.3131,
+      "step": 2919
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9019360905411986e-05,
+      "loss": 0.3036,
+      "step": 2920
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.901902507010562e-05,
+      "loss": 0.3357,
+      "step": 2921
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9018689234799255e-05,
+      "loss": 0.2939,
+      "step": 2922
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.901835339949289e-05,
+      "loss": 0.3795,
+      "step": 2923
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9018017564186525e-05,
+      "loss": 0.2765,
+      "step": 2924
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.901768172888016e-05,
+      "loss": 0.2928,
+      "step": 2925
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9017345893573794e-05,
+      "loss": 0.3995,
+      "step": 2926
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9017010058267425e-05,
+      "loss": 0.3021,
+      "step": 2927
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9016674222961063e-05,
+      "loss": 0.3927,
+      "step": 2928
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9016338387654695e-05,
+      "loss": 0.203,
+      "step": 2929
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.901600255234833e-05,
+      "loss": 0.5151,
+      "step": 2930
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9015666717041964e-05,
+      "loss": 0.2067,
+      "step": 2931
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.90153308817356e-05,
+      "loss": 0.3867,
+      "step": 2932
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9014995046429233e-05,
+      "loss": 0.2264,
+      "step": 2933
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9014659211122868e-05,
+      "loss": 0.428,
+      "step": 2934
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9014323375816502e-05,
+      "loss": 0.3153,
+      "step": 2935
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9013987540510134e-05,
+      "loss": 0.3597,
+      "step": 2936
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9013651705203772e-05,
+      "loss": 0.555,
+      "step": 2937
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9013315869897403e-05,
+      "loss": 0.3946,
+      "step": 2938
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9012980034591038e-05,
+      "loss": 0.4058,
+      "step": 2939
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9012644199284672e-05,
+      "loss": 0.4401,
+      "step": 2940
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9012308363978307e-05,
+      "loss": 0.3286,
+      "step": 2941
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.901197252867194e-05,
+      "loss": 0.4528,
+      "step": 2942
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9011636693365576e-05,
+      "loss": 0.3188,
+      "step": 2943
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.901130085805921e-05,
+      "loss": 0.4647,
+      "step": 2944
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9010965022752842e-05,
+      "loss": 0.277,
+      "step": 2945
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.901062918744648e-05,
+      "loss": 0.3592,
+      "step": 2946
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.901029335214011e-05,
+      "loss": 0.4697,
+      "step": 2947
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9009957516833746e-05,
+      "loss": 0.4002,
+      "step": 2948
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.900962168152738e-05,
+      "loss": 0.356,
+      "step": 2949
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9009285846221015e-05,
+      "loss": 0.3937,
+      "step": 2950
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.900895001091465e-05,
+      "loss": 0.3507,
+      "step": 2951
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9008614175608285e-05,
+      "loss": 0.2961,
+      "step": 2952
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.900827834030192e-05,
+      "loss": 0.3101,
+      "step": 2953
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.900794250499555e-05,
+      "loss": 0.401,
+      "step": 2954
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.900760666968919e-05,
+      "loss": 0.5746,
+      "step": 2955
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.900727083438282e-05,
+      "loss": 0.3982,
+      "step": 2956
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9006934999076455e-05,
+      "loss": 0.3078,
+      "step": 2957
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.900659916377009e-05,
+      "loss": 0.2597,
+      "step": 2958
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9006263328463724e-05,
+      "loss": 0.4065,
+      "step": 2959
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9005927493157355e-05,
+      "loss": 0.1885,
+      "step": 2960
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9005591657850993e-05,
+      "loss": 0.3603,
+      "step": 2961
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9005255822544624e-05,
+      "loss": 0.3369,
+      "step": 2962
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.900491998723826e-05,
+      "loss": 0.3782,
+      "step": 2963
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9004584151931894e-05,
+      "loss": 0.1959,
+      "step": 2964
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.900424831662553e-05,
+      "loss": 0.352,
+      "step": 2965
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9003912481319163e-05,
+      "loss": 0.4538,
+      "step": 2966
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9003576646012798e-05,
+      "loss": 0.4226,
+      "step": 2967
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9003240810706432e-05,
+      "loss": 0.1849,
+      "step": 2968
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9002904975400064e-05,
+      "loss": 0.5235,
+      "step": 2969
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.90025691400937e-05,
+      "loss": 0.3496,
+      "step": 2970
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9002233304787333e-05,
+      "loss": 0.3173,
+      "step": 2971
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9001897469480967e-05,
+      "loss": 0.2534,
+      "step": 2972
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9001561634174602e-05,
+      "loss": 0.2886,
+      "step": 2973
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9001225798868237e-05,
+      "loss": 0.325,
+      "step": 2974
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.900088996356187e-05,
+      "loss": 0.3147,
+      "step": 2975
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9000554128255506e-05,
+      "loss": 0.4101,
+      "step": 2976
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.900021829294914e-05,
+      "loss": 0.3468,
+      "step": 2977
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8999882457642772e-05,
+      "loss": 0.5346,
+      "step": 2978
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8999546622336407e-05,
+      "loss": 0.422,
+      "step": 2979
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.899921078703004e-05,
+      "loss": 0.5385,
+      "step": 2980
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8998874951723676e-05,
+      "loss": 0.381,
+      "step": 2981
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.899853911641731e-05,
+      "loss": 0.2203,
+      "step": 2982
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8998203281110945e-05,
+      "loss": 0.472,
+      "step": 2983
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.899786744580458e-05,
+      "loss": 0.257,
+      "step": 2984
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.899753161049821e-05,
+      "loss": 0.3383,
+      "step": 2985
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.899719577519185e-05,
+      "loss": 0.4299,
+      "step": 2986
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.899685993988548e-05,
+      "loss": 0.4245,
+      "step": 2987
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8996524104579115e-05,
+      "loss": 0.4179,
+      "step": 2988
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.899618826927275e-05,
+      "loss": 0.3964,
+      "step": 2989
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8995852433966384e-05,
+      "loss": 0.3349,
+      "step": 2990
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.899551659866002e-05,
+      "loss": 0.3939,
+      "step": 2991
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8995180763353654e-05,
+      "loss": 0.3164,
+      "step": 2992
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.899484492804729e-05,
+      "loss": 0.3224,
+      "step": 2993
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.899450909274092e-05,
+      "loss": 0.2506,
+      "step": 2994
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8994173257434558e-05,
+      "loss": 0.194,
+      "step": 2995
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.899383742212819e-05,
+      "loss": 0.3635,
+      "step": 2996
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8993501586821824e-05,
+      "loss": 0.3419,
+      "step": 2997
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8993165751515458e-05,
+      "loss": 0.3596,
+      "step": 2998
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8992829916209093e-05,
+      "loss": 0.2488,
+      "step": 2999
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8992494080902727e-05,
+      "loss": 0.17,
+      "step": 3000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8992158245596362e-05,
+      "loss": 0.5618,
+      "step": 3001
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8991822410289997e-05,
+      "loss": 0.2337,
+      "step": 3002
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8991486574983628e-05,
+      "loss": 0.4387,
+      "step": 3003
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8991150739677263e-05,
+      "loss": 0.2334,
+      "step": 3004
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8990814904370897e-05,
+      "loss": 0.4468,
+      "step": 3005
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8990479069064532e-05,
+      "loss": 0.4586,
+      "step": 3006
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8990143233758167e-05,
+      "loss": 0.4977,
+      "step": 3007
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.89898073984518e-05,
+      "loss": 0.3379,
+      "step": 3008
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8989471563145433e-05,
+      "loss": 0.4271,
+      "step": 3009
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.898913572783907e-05,
+      "loss": 0.3656,
+      "step": 3010
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8988799892532702e-05,
+      "loss": 0.3987,
+      "step": 3011
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8988464057226336e-05,
+      "loss": 0.2165,
+      "step": 3012
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.898812822191997e-05,
+      "loss": 0.372,
+      "step": 3013
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8987792386613606e-05,
+      "loss": 0.2149,
+      "step": 3014
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.898745655130724e-05,
+      "loss": 0.2538,
+      "step": 3015
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8987120716000875e-05,
+      "loss": 0.2157,
+      "step": 3016
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.898678488069451e-05,
+      "loss": 0.4191,
+      "step": 3017
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.898644904538814e-05,
+      "loss": 0.3412,
+      "step": 3018
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.898611321008178e-05,
+      "loss": 0.312,
+      "step": 3019
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.898577737477541e-05,
+      "loss": 0.4679,
+      "step": 3020
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8985441539469045e-05,
+      "loss": 0.2353,
+      "step": 3021
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.898510570416268e-05,
+      "loss": 0.2525,
+      "step": 3022
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8984769868856314e-05,
+      "loss": 0.3356,
+      "step": 3023
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.898443403354995e-05,
+      "loss": 0.3741,
+      "step": 3024
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8984098198243583e-05,
+      "loss": 0.3286,
+      "step": 3025
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8983762362937218e-05,
+      "loss": 0.3087,
+      "step": 3026
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.898342652763085e-05,
+      "loss": 0.4053,
+      "step": 3027
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8983090692324487e-05,
+      "loss": 0.3801,
+      "step": 3028
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.898275485701812e-05,
+      "loss": 0.3498,
+      "step": 3029
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8982419021711753e-05,
+      "loss": 0.2675,
+      "step": 3030
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8982083186405388e-05,
+      "loss": 0.2952,
+      "step": 3031
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8981747351099023e-05,
+      "loss": 0.2765,
+      "step": 3032
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8981411515792657e-05,
+      "loss": 0.3959,
+      "step": 3033
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8981075680486292e-05,
+      "loss": 0.4434,
+      "step": 3034
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8980739845179927e-05,
+      "loss": 0.4256,
+      "step": 3035
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8980404009873558e-05,
+      "loss": 0.3549,
+      "step": 3036
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8980068174567196e-05,
+      "loss": 0.3671,
+      "step": 3037
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8979732339260827e-05,
+      "loss": 0.2725,
+      "step": 3038
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8979396503954462e-05,
+      "loss": 0.3264,
+      "step": 3039
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8979060668648096e-05,
+      "loss": 0.2608,
+      "step": 3040
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.897872483334173e-05,
+      "loss": 0.3705,
+      "step": 3041
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8978388998035366e-05,
+      "loss": 0.2465,
+      "step": 3042
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8978053162729e-05,
+      "loss": 0.3714,
+      "step": 3043
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8977717327422635e-05,
+      "loss": 0.328,
+      "step": 3044
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8977381492116266e-05,
+      "loss": 0.3773,
+      "step": 3045
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8977045656809904e-05,
+      "loss": 0.3409,
+      "step": 3046
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8976709821503536e-05,
+      "loss": 0.3979,
+      "step": 3047
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.897637398619717e-05,
+      "loss": 0.3305,
+      "step": 3048
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8976038150890805e-05,
+      "loss": 0.5288,
+      "step": 3049
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.897570231558444e-05,
+      "loss": 0.2551,
+      "step": 3050
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.897536648027807e-05,
+      "loss": 0.4654,
+      "step": 3051
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.897503064497171e-05,
+      "loss": 0.3876,
+      "step": 3052
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.897469480966534e-05,
+      "loss": 0.5503,
+      "step": 3053
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8974358974358975e-05,
+      "loss": 0.2235,
+      "step": 3054
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.897402313905261e-05,
+      "loss": 0.3644,
+      "step": 3055
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8973687303746244e-05,
+      "loss": 0.6073,
+      "step": 3056
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.897335146843988e-05,
+      "loss": 0.4575,
+      "step": 3057
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8973015633133513e-05,
+      "loss": 0.2911,
+      "step": 3058
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8972679797827148e-05,
+      "loss": 0.5906,
+      "step": 3059
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.897234396252078e-05,
+      "loss": 0.316,
+      "step": 3060
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8972008127214417e-05,
+      "loss": 0.3796,
+      "step": 3061
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.897167229190805e-05,
+      "loss": 0.324,
+      "step": 3062
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8971336456601683e-05,
+      "loss": 0.4264,
+      "step": 3063
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8971000621295318e-05,
+      "loss": 0.333,
+      "step": 3064
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8970664785988952e-05,
+      "loss": 0.2083,
+      "step": 3065
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8970328950682587e-05,
+      "loss": 0.4077,
+      "step": 3066
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8969993115376222e-05,
+      "loss": 0.3012,
+      "step": 3067
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8969657280069856e-05,
+      "loss": 0.402,
+      "step": 3068
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8969321444763488e-05,
+      "loss": 0.579,
+      "step": 3069
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8968985609457126e-05,
+      "loss": 0.3591,
+      "step": 3070
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8968649774150757e-05,
+      "loss": 0.2197,
+      "step": 3071
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.896831393884439e-05,
+      "loss": 0.2378,
+      "step": 3072
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8967978103538026e-05,
+      "loss": 0.3885,
+      "step": 3073
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.896764226823166e-05,
+      "loss": 0.3701,
+      "step": 3074
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.8967306432925296e-05,
+      "loss": 0.2819,
+      "step": 3075
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.896697059761893e-05,
+      "loss": 0.4297,
+      "step": 3076
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8966634762312565e-05,
+      "loss": 0.4953,
+      "step": 3077
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8966298927006196e-05,
+      "loss": 0.3026,
+      "step": 3078
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8965963091699834e-05,
+      "loss": 0.2806,
+      "step": 3079
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8965627256393465e-05,
+      "loss": 0.3166,
+      "step": 3080
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.89652914210871e-05,
+      "loss": 0.3963,
+      "step": 3081
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8964955585780735e-05,
+      "loss": 0.3264,
+      "step": 3082
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.896461975047437e-05,
+      "loss": 0.3839,
+      "step": 3083
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8964283915168004e-05,
+      "loss": 0.3329,
+      "step": 3084
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.896394807986164e-05,
+      "loss": 0.1704,
+      "step": 3085
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8963612244555273e-05,
+      "loss": 0.2569,
+      "step": 3086
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8963276409248905e-05,
+      "loss": 0.3322,
+      "step": 3087
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8962940573942543e-05,
+      "loss": 0.287,
+      "step": 3088
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8962604738636174e-05,
+      "loss": 0.2729,
+      "step": 3089
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.896226890332981e-05,
+      "loss": 0.2437,
+      "step": 3090
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8961933068023443e-05,
+      "loss": 0.2687,
+      "step": 3091
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8961597232717078e-05,
+      "loss": 0.3207,
+      "step": 3092
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8961261397410712e-05,
+      "loss": 0.3178,
+      "step": 3093
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8960925562104347e-05,
+      "loss": 0.3351,
+      "step": 3094
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8960589726797978e-05,
+      "loss": 0.2403,
+      "step": 3095
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8960253891491613e-05,
+      "loss": 0.2159,
+      "step": 3096
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8959918056185248e-05,
+      "loss": 0.4499,
+      "step": 3097
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8959582220878882e-05,
+      "loss": 0.3432,
+      "step": 3098
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8959246385572517e-05,
+      "loss": 0.4175,
+      "step": 3099
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.895891055026615e-05,
+      "loss": 0.3129,
+      "step": 3100
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8958574714959786e-05,
+      "loss": 0.3195,
+      "step": 3101
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8958238879653417e-05,
+      "loss": 0.4441,
+      "step": 3102
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8957903044347055e-05,
+      "loss": 0.3311,
+      "step": 3103
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8957567209040687e-05,
+      "loss": 0.2831,
+      "step": 3104
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.895723137373432e-05,
+      "loss": 0.4627,
+      "step": 3105
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8956895538427956e-05,
+      "loss": 0.412,
+      "step": 3106
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.895655970312159e-05,
+      "loss": 0.3657,
+      "step": 3107
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8956223867815225e-05,
+      "loss": 0.3606,
+      "step": 3108
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.895588803250886e-05,
+      "loss": 0.5905,
+      "step": 3109
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8955552197202495e-05,
+      "loss": 0.6123,
+      "step": 3110
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8955216361896126e-05,
+      "loss": 0.2845,
+      "step": 3111
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8954880526589764e-05,
+      "loss": 0.2358,
+      "step": 3112
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8954544691283395e-05,
+      "loss": 0.4086,
+      "step": 3113
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.895420885597703e-05,
+      "loss": 0.2576,
+      "step": 3114
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8953873020670664e-05,
+      "loss": 0.5835,
+      "step": 3115
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.89535371853643e-05,
+      "loss": 0.3535,
+      "step": 3116
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8953201350057934e-05,
+      "loss": 0.3445,
+      "step": 3117
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.895286551475157e-05,
+      "loss": 0.2071,
+      "step": 3118
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8952529679445203e-05,
+      "loss": 0.238,
+      "step": 3119
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8952193844138834e-05,
+      "loss": 0.4131,
+      "step": 3120
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8951858008832472e-05,
+      "loss": 0.4482,
+      "step": 3121
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8951522173526104e-05,
+      "loss": 0.3363,
+      "step": 3122
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8951186338219738e-05,
+      "loss": 0.3592,
+      "step": 3123
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8950850502913373e-05,
+      "loss": 0.3725,
+      "step": 3124
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8950514667607008e-05,
+      "loss": 0.4203,
+      "step": 3125
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8950178832300642e-05,
+      "loss": 0.3029,
+      "step": 3126
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8949842996994277e-05,
+      "loss": 0.3154,
+      "step": 3127
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.894950716168791e-05,
+      "loss": 0.4178,
+      "step": 3128
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8949171326381543e-05,
+      "loss": 0.3243,
+      "step": 3129
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.894883549107518e-05,
+      "loss": 0.3513,
+      "step": 3130
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8948499655768812e-05,
+      "loss": 0.4367,
+      "step": 3131
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8948163820462447e-05,
+      "loss": 0.3166,
+      "step": 3132
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.894782798515608e-05,
+      "loss": 0.4544,
+      "step": 3133
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8947492149849716e-05,
+      "loss": 0.4167,
+      "step": 3134
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.894715631454335e-05,
+      "loss": 0.2152,
+      "step": 3135
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8946820479236985e-05,
+      "loss": 0.3055,
+      "step": 3136
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.894648464393062e-05,
+      "loss": 0.3112,
+      "step": 3137
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.894614880862425e-05,
+      "loss": 0.2591,
+      "step": 3138
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8945812973317886e-05,
+      "loss": 0.2526,
+      "step": 3139
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.894547713801152e-05,
+      "loss": 0.3874,
+      "step": 3140
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8945141302705155e-05,
+      "loss": 0.3353,
+      "step": 3141
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.894480546739879e-05,
+      "loss": 0.2913,
+      "step": 3142
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8944469632092424e-05,
+      "loss": 0.543,
+      "step": 3143
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8944133796786056e-05,
+      "loss": 0.4232,
+      "step": 3144
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8943797961479694e-05,
+      "loss": 0.3223,
+      "step": 3145
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8943462126173325e-05,
+      "loss": 0.3647,
+      "step": 3146
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.894312629086696e-05,
+      "loss": 0.3,
+      "step": 3147
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8942790455560594e-05,
+      "loss": 0.2307,
+      "step": 3148
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.894245462025423e-05,
+      "loss": 0.2907,
+      "step": 3149
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8942118784947864e-05,
+      "loss": 0.2919,
+      "step": 3150
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8941782949641498e-05,
+      "loss": 0.3926,
+      "step": 3151
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8941447114335133e-05,
+      "loss": 0.2487,
+      "step": 3152
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8941111279028764e-05,
+      "loss": 0.2586,
+      "step": 3153
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8940775443722402e-05,
+      "loss": 0.3286,
+      "step": 3154
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8940439608416033e-05,
+      "loss": 0.4305,
+      "step": 3155
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8940103773109668e-05,
+      "loss": 0.4469,
+      "step": 3156
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8939767937803303e-05,
+      "loss": 0.3658,
+      "step": 3157
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8939432102496937e-05,
+      "loss": 0.1766,
+      "step": 3158
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8939096267190572e-05,
+      "loss": 0.3119,
+      "step": 3159
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8938760431884203e-05,
+      "loss": 0.3688,
+      "step": 3160
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893842459657784e-05,
+      "loss": 0.3846,
+      "step": 3161
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8938088761271473e-05,
+      "loss": 0.4873,
+      "step": 3162
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8937752925965107e-05,
+      "loss": 0.4678,
+      "step": 3163
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8937417090658742e-05,
+      "loss": 0.3059,
+      "step": 3164
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8937081255352377e-05,
+      "loss": 0.3016,
+      "step": 3165
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893674542004601e-05,
+      "loss": 0.4643,
+      "step": 3166
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8936409584739646e-05,
+      "loss": 0.4253,
+      "step": 3167
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893607374943328e-05,
+      "loss": 0.2531,
+      "step": 3168
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8935737914126912e-05,
+      "loss": 0.3409,
+      "step": 3169
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893540207882055e-05,
+      "loss": 0.2597,
+      "step": 3170
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893506624351418e-05,
+      "loss": 0.2628,
+      "step": 3171
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8934730408207816e-05,
+      "loss": 0.3746,
+      "step": 3172
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893439457290145e-05,
+      "loss": 0.3455,
+      "step": 3173
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8934058737595085e-05,
+      "loss": 0.5957,
+      "step": 3174
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893372290228872e-05,
+      "loss": 0.2143,
+      "step": 3175
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8933387066982354e-05,
+      "loss": 0.3621,
+      "step": 3176
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893305123167599e-05,
+      "loss": 0.3211,
+      "step": 3177
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893271539636962e-05,
+      "loss": 0.4602,
+      "step": 3178
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8932379561063258e-05,
+      "loss": 0.603,
+      "step": 3179
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893204372575689e-05,
+      "loss": 0.3238,
+      "step": 3180
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8931707890450524e-05,
+      "loss": 0.225,
+      "step": 3181
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893137205514416e-05,
+      "loss": 0.3003,
+      "step": 3182
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8931036219837793e-05,
+      "loss": 0.3612,
+      "step": 3183
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8930700384531425e-05,
+      "loss": 0.2212,
+      "step": 3184
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8930364549225063e-05,
+      "loss": 0.4589,
+      "step": 3185
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8930028713918694e-05,
+      "loss": 0.4377,
+      "step": 3186
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892969287861233e-05,
+      "loss": 0.2282,
+      "step": 3187
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8929357043305963e-05,
+      "loss": 0.3502,
+      "step": 3188
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8929021207999598e-05,
+      "loss": 0.2486,
+      "step": 3189
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8928685372693233e-05,
+      "loss": 0.3317,
+      "step": 3190
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8928349537386867e-05,
+      "loss": 0.2867,
+      "step": 3191
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8928013702080502e-05,
+      "loss": 0.2736,
+      "step": 3192
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8927677866774133e-05,
+      "loss": 0.3531,
+      "step": 3193
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892734203146777e-05,
+      "loss": 0.3399,
+      "step": 3194
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8927006196161402e-05,
+      "loss": 0.5351,
+      "step": 3195
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8926670360855037e-05,
+      "loss": 0.2993,
+      "step": 3196
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892633452554867e-05,
+      "loss": 0.2241,
+      "step": 3197
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8925998690242306e-05,
+      "loss": 0.2136,
+      "step": 3198
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892566285493594e-05,
+      "loss": 0.3366,
+      "step": 3199
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8925327019629576e-05,
+      "loss": 0.3251,
+      "step": 3200
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892499118432321e-05,
+      "loss": 0.5449,
+      "step": 3201
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892465534901684e-05,
+      "loss": 0.3933,
+      "step": 3202
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892431951371048e-05,
+      "loss": 0.4215,
+      "step": 3203
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892398367840411e-05,
+      "loss": 0.4612,
+      "step": 3204
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8923647843097745e-05,
+      "loss": 0.3923,
+      "step": 3205
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892331200779138e-05,
+      "loss": 0.4177,
+      "step": 3206
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8922976172485015e-05,
+      "loss": 0.296,
+      "step": 3207
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892264033717865e-05,
+      "loss": 0.2582,
+      "step": 3208
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8922304501872284e-05,
+      "loss": 0.4173,
+      "step": 3209
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892196866656592e-05,
+      "loss": 0.3288,
+      "step": 3210
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892163283125955e-05,
+      "loss": 0.3716,
+      "step": 3211
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8921296995953188e-05,
+      "loss": 0.4133,
+      "step": 3212
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892096116064682e-05,
+      "loss": 0.4346,
+      "step": 3213
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8920625325340454e-05,
+      "loss": 0.3982,
+      "step": 3214
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892028949003409e-05,
+      "loss": 0.2605,
+      "step": 3215
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8919953654727723e-05,
+      "loss": 0.5097,
+      "step": 3216
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8919617819421358e-05,
+      "loss": 0.2387,
+      "step": 3217
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8919281984114992e-05,
+      "loss": 0.423,
+      "step": 3218
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8918946148808627e-05,
+      "loss": 0.4215,
+      "step": 3219
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.891861031350226e-05,
+      "loss": 0.407,
+      "step": 3220
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8918274478195896e-05,
+      "loss": 0.3966,
+      "step": 3221
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8917938642889528e-05,
+      "loss": 0.2722,
+      "step": 3222
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8917602807583162e-05,
+      "loss": 0.214,
+      "step": 3223
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8917266972276797e-05,
+      "loss": 0.368,
+      "step": 3224
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.891693113697043e-05,
+      "loss": 0.3202,
+      "step": 3225
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8916595301664066e-05,
+      "loss": 0.4024,
+      "step": 3226
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.89162594663577e-05,
+      "loss": 0.3305,
+      "step": 3227
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8915923631051336e-05,
+      "loss": 0.3456,
+      "step": 3228
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8915587795744967e-05,
+      "loss": 0.1469,
+      "step": 3229
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.89152519604386e-05,
+      "loss": 0.2894,
+      "step": 3230
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8914916125132236e-05,
+      "loss": 0.3477,
+      "step": 3231
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.891458028982587e-05,
+      "loss": 0.348,
+      "step": 3232
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8914244454519505e-05,
+      "loss": 0.2982,
+      "step": 3233
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.891390861921314e-05,
+      "loss": 0.5425,
+      "step": 3234
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.891357278390677e-05,
+      "loss": 0.2705,
+      "step": 3235
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.891323694860041e-05,
+      "loss": 0.5353,
+      "step": 3236
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.891290111329404e-05,
+      "loss": 0.4448,
+      "step": 3237
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8912565277987675e-05,
+      "loss": 0.4752,
+      "step": 3238
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.891222944268131e-05,
+      "loss": 0.4266,
+      "step": 3239
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8911893607374945e-05,
+      "loss": 0.2754,
+      "step": 3240
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.891155777206858e-05,
+      "loss": 0.2894,
+      "step": 3241
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8911221936762214e-05,
+      "loss": 0.4437,
+      "step": 3242
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.891088610145585e-05,
+      "loss": 0.2873,
+      "step": 3243
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.891055026614948e-05,
+      "loss": 0.3461,
+      "step": 3244
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8910214430843118e-05,
+      "loss": 0.2602,
+      "step": 3245
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.890987859553675e-05,
+      "loss": 0.3584,
+      "step": 3246
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8909542760230384e-05,
+      "loss": 0.3291,
+      "step": 3247
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.890920692492402e-05,
+      "loss": 0.2678,
+      "step": 3248
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8908871089617653e-05,
+      "loss": 0.341,
+      "step": 3249
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8908535254311288e-05,
+      "loss": 0.457,
+      "step": 3250
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8908199419004922e-05,
+      "loss": 0.33,
+      "step": 3251
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8907863583698557e-05,
+      "loss": 0.5762,
+      "step": 3252
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8907527748392188e-05,
+      "loss": 0.4924,
+      "step": 3253
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8907191913085826e-05,
+      "loss": 0.469,
+      "step": 3254
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8906856077779458e-05,
+      "loss": 0.3803,
+      "step": 3255
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8906520242473092e-05,
+      "loss": 0.3904,
+      "step": 3256
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8906184407166727e-05,
+      "loss": 0.4432,
+      "step": 3257
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.890584857186036e-05,
+      "loss": 0.267,
+      "step": 3258
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8905512736553996e-05,
+      "loss": 0.3125,
+      "step": 3259
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.890517690124763e-05,
+      "loss": 0.3522,
+      "step": 3260
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8904841065941265e-05,
+      "loss": 0.261,
+      "step": 3261
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8904505230634897e-05,
+      "loss": 0.5055,
+      "step": 3262
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8904169395328535e-05,
+      "loss": 0.402,
+      "step": 3263
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8903833560022166e-05,
+      "loss": 0.23,
+      "step": 3264
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.89034977247158e-05,
+      "loss": 0.4653,
+      "step": 3265
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8903161889409435e-05,
+      "loss": 0.4356,
+      "step": 3266
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.890282605410307e-05,
+      "loss": 0.3442,
+      "step": 3267
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8902490218796705e-05,
+      "loss": 0.2929,
+      "step": 3268
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.890215438349034e-05,
+      "loss": 0.7387,
+      "step": 3269
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8901818548183974e-05,
+      "loss": 0.3093,
+      "step": 3270
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8901482712877605e-05,
+      "loss": 0.229,
+      "step": 3271
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8901146877571243e-05,
+      "loss": 0.3565,
+      "step": 3272
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8900811042264874e-05,
+      "loss": 0.2384,
+      "step": 3273
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.890047520695851e-05,
+      "loss": 0.3182,
+      "step": 3274
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8900139371652144e-05,
+      "loss": 0.2771,
+      "step": 3275
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.889980353634578e-05,
+      "loss": 0.4114,
+      "step": 3276
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.889946770103941e-05,
+      "loss": 0.4319,
+      "step": 3277
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8899131865733048e-05,
+      "loss": 0.2585,
+      "step": 3278
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.889879603042668e-05,
+      "loss": 0.4104,
+      "step": 3279
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8898460195120314e-05,
+      "loss": 0.4556,
+      "step": 3280
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8898124359813948e-05,
+      "loss": 0.4558,
+      "step": 3281
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8897788524507583e-05,
+      "loss": 0.4796,
+      "step": 3282
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8897452689201217e-05,
+      "loss": 0.4423,
+      "step": 3283
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8897116853894852e-05,
+      "loss": 0.2243,
+      "step": 3284
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8896781018588487e-05,
+      "loss": 0.3996,
+      "step": 3285
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8896445183282118e-05,
+      "loss": 0.428,
+      "step": 3286
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8896109347975756e-05,
+      "loss": 0.3363,
+      "step": 3287
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8895773512669387e-05,
+      "loss": 0.3755,
+      "step": 3288
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8895437677363022e-05,
+      "loss": 0.4004,
+      "step": 3289
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8895101842056657e-05,
+      "loss": 0.3811,
+      "step": 3290
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.889476600675029e-05,
+      "loss": 0.4086,
+      "step": 3291
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8894430171443926e-05,
+      "loss": 0.3324,
+      "step": 3292
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.889409433613756e-05,
+      "loss": 0.3789,
+      "step": 3293
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8893758500831195e-05,
+      "loss": 0.402,
+      "step": 3294
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8893422665524826e-05,
+      "loss": 0.3082,
+      "step": 3295
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8893086830218464e-05,
+      "loss": 0.2751,
+      "step": 3296
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8892750994912096e-05,
+      "loss": 0.6242,
+      "step": 3297
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.889241515960573e-05,
+      "loss": 0.2028,
+      "step": 3298
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8892079324299365e-05,
+      "loss": 0.3844,
+      "step": 3299
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8891743488993e-05,
+      "loss": 0.3085,
+      "step": 3300
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8891407653686634e-05,
+      "loss": 0.3353,
+      "step": 3301
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.889107181838027e-05,
+      "loss": 0.3188,
+      "step": 3302
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8890735983073904e-05,
+      "loss": 0.264,
+      "step": 3303
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8890400147767535e-05,
+      "loss": 0.415,
+      "step": 3304
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8890064312461173e-05,
+      "loss": 0.443,
+      "step": 3305
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8889728477154804e-05,
+      "loss": 0.272,
+      "step": 3306
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.888939264184844e-05,
+      "loss": 0.3691,
+      "step": 3307
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8889056806542073e-05,
+      "loss": 0.4125,
+      "step": 3308
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8888720971235708e-05,
+      "loss": 0.3273,
+      "step": 3309
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8888385135929343e-05,
+      "loss": 0.4005,
+      "step": 3310
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8888049300622977e-05,
+      "loss": 0.4337,
+      "step": 3311
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8887713465316612e-05,
+      "loss": 0.5139,
+      "step": 3312
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8887377630010243e-05,
+      "loss": 0.3144,
+      "step": 3313
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.888704179470388e-05,
+      "loss": 0.2771,
+      "step": 3314
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8886705959397513e-05,
+      "loss": 0.3022,
+      "step": 3315
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8886370124091147e-05,
+      "loss": 0.4089,
+      "step": 3316
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8886034288784782e-05,
+      "loss": 0.4237,
+      "step": 3317
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8885698453478417e-05,
+      "loss": 0.5184,
+      "step": 3318
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8885362618172048e-05,
+      "loss": 0.2424,
+      "step": 3319
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8885026782865686e-05,
+      "loss": 0.5523,
+      "step": 3320
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8884690947559317e-05,
+      "loss": 0.2614,
+      "step": 3321
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8884355112252952e-05,
+      "loss": 0.283,
+      "step": 3322
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8884019276946586e-05,
+      "loss": 0.2437,
+      "step": 3323
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.888368344164022e-05,
+      "loss": 0.455,
+      "step": 3324
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8883347606333856e-05,
+      "loss": 0.4541,
+      "step": 3325
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.888301177102749e-05,
+      "loss": 0.4179,
+      "step": 3326
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8882675935721125e-05,
+      "loss": 0.4092,
+      "step": 3327
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8882340100414756e-05,
+      "loss": 0.465,
+      "step": 3328
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8882004265108394e-05,
+      "loss": 0.3666,
+      "step": 3329
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8881668429802026e-05,
+      "loss": 0.2901,
+      "step": 3330
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.888133259449566e-05,
+      "loss": 0.5366,
+      "step": 3331
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8880996759189295e-05,
+      "loss": 0.3965,
+      "step": 3332
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.888066092388293e-05,
+      "loss": 0.2157,
+      "step": 3333
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8880325088576564e-05,
+      "loss": 0.5824,
+      "step": 3334
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.88799892532702e-05,
+      "loss": 0.2415,
+      "step": 3335
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8879653417963833e-05,
+      "loss": 0.3283,
+      "step": 3336
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8879317582657465e-05,
+      "loss": 0.3748,
+      "step": 3337
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8878981747351103e-05,
+      "loss": 0.5015,
+      "step": 3338
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8878645912044734e-05,
+      "loss": 0.3709,
+      "step": 3339
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887831007673837e-05,
+      "loss": 0.249,
+      "step": 3340
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8877974241432003e-05,
+      "loss": 0.2887,
+      "step": 3341
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8877638406125638e-05,
+      "loss": 0.3153,
+      "step": 3342
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8877302570819273e-05,
+      "loss": 0.5061,
+      "step": 3343
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8876966735512904e-05,
+      "loss": 0.4804,
+      "step": 3344
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8876630900206542e-05,
+      "loss": 0.4443,
+      "step": 3345
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8876295064900173e-05,
+      "loss": 0.4024,
+      "step": 3346
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8875959229593808e-05,
+      "loss": 0.4403,
+      "step": 3347
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8875623394287442e-05,
+      "loss": 0.2824,
+      "step": 3348
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8875287558981077e-05,
+      "loss": 0.3381,
+      "step": 3349
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8874951723674712e-05,
+      "loss": 0.2989,
+      "step": 3350
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8874615888368346e-05,
+      "loss": 0.3895,
+      "step": 3351
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887428005306198e-05,
+      "loss": 0.4289,
+      "step": 3352
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8873944217755612e-05,
+      "loss": 0.3961,
+      "step": 3353
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887360838244925e-05,
+      "loss": 0.5153,
+      "step": 3354
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887327254714288e-05,
+      "loss": 0.4838,
+      "step": 3355
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8872936711836516e-05,
+      "loss": 0.3852,
+      "step": 3356
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887260087653015e-05,
+      "loss": 0.2685,
+      "step": 3357
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8872265041223786e-05,
+      "loss": 0.4787,
+      "step": 3358
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887192920591742e-05,
+      "loss": 0.1853,
+      "step": 3359
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8871593370611055e-05,
+      "loss": 0.3098,
+      "step": 3360
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887125753530469e-05,
+      "loss": 0.2718,
+      "step": 3361
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887092169999832e-05,
+      "loss": 0.2562,
+      "step": 3362
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887058586469196e-05,
+      "loss": 0.292,
+      "step": 3363
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887025002938559e-05,
+      "loss": 0.3858,
+      "step": 3364
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8869914194079225e-05,
+      "loss": 0.3339,
+      "step": 3365
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.886957835877286e-05,
+      "loss": 0.2609,
+      "step": 3366
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8869242523466494e-05,
+      "loss": 0.32,
+      "step": 3367
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8868906688160125e-05,
+      "loss": 0.2419,
+      "step": 3368
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8868570852853763e-05,
+      "loss": 0.431,
+      "step": 3369
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8868235017547395e-05,
+      "loss": 0.2107,
+      "step": 3370
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.886789918224103e-05,
+      "loss": 0.2629,
+      "step": 3371
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8867563346934664e-05,
+      "loss": 0.3284,
+      "step": 3372
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.88672275116283e-05,
+      "loss": 0.3937,
+      "step": 3373
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8866891676321933e-05,
+      "loss": 0.329,
+      "step": 3374
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8866555841015568e-05,
+      "loss": 0.2703,
+      "step": 3375
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8866220005709202e-05,
+      "loss": 0.4357,
+      "step": 3376
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8865884170402834e-05,
+      "loss": 0.355,
+      "step": 3377
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8865548335096472e-05,
+      "loss": 0.3482,
+      "step": 3378
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8865212499790103e-05,
+      "loss": 0.3688,
+      "step": 3379
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8864876664483738e-05,
+      "loss": 0.2713,
+      "step": 3380
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8864540829177372e-05,
+      "loss": 0.3549,
+      "step": 3381
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8864204993871007e-05,
+      "loss": 0.3732,
+      "step": 3382
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.886386915856464e-05,
+      "loss": 0.3321,
+      "step": 3383
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8863533323258276e-05,
+      "loss": 0.3047,
+      "step": 3384
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.886319748795191e-05,
+      "loss": 0.37,
+      "step": 3385
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8862861652645542e-05,
+      "loss": 0.324,
+      "step": 3386
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.886252581733918e-05,
+      "loss": 0.4287,
+      "step": 3387
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.886218998203281e-05,
+      "loss": 0.2559,
+      "step": 3388
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8861854146726446e-05,
+      "loss": 0.4016,
+      "step": 3389
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.886151831142008e-05,
+      "loss": 0.417,
+      "step": 3390
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8861182476113715e-05,
+      "loss": 0.5317,
+      "step": 3391
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.886084664080735e-05,
+      "loss": 0.5255,
+      "step": 3392
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8860510805500985e-05,
+      "loss": 0.3809,
+      "step": 3393
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.886017497019462e-05,
+      "loss": 0.2767,
+      "step": 3394
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.885983913488825e-05,
+      "loss": 0.4961,
+      "step": 3395
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.885950329958189e-05,
+      "loss": 0.3236,
+      "step": 3396
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.885916746427552e-05,
+      "loss": 0.4623,
+      "step": 3397
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8858831628969154e-05,
+      "loss": 0.3514,
+      "step": 3398
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.885849579366279e-05,
+      "loss": 0.4044,
+      "step": 3399
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8858159958356424e-05,
+      "loss": 0.4386,
+      "step": 3400
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.885782412305006e-05,
+      "loss": 0.3758,
+      "step": 3401
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8857488287743693e-05,
+      "loss": 0.4521,
+      "step": 3402
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8857152452437328e-05,
+      "loss": 0.4506,
+      "step": 3403
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.885681661713096e-05,
+      "loss": 0.418,
+      "step": 3404
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8856480781824597e-05,
+      "loss": 0.4174,
+      "step": 3405
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8856144946518228e-05,
+      "loss": 0.2316,
+      "step": 3406
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8855809111211863e-05,
+      "loss": 0.3433,
+      "step": 3407
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8855473275905498e-05,
+      "loss": 0.4072,
+      "step": 3408
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8855137440599132e-05,
+      "loss": 0.2893,
+      "step": 3409
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8854801605292763e-05,
+      "loss": 0.2098,
+      "step": 3410
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.88544657699864e-05,
+      "loss": 0.2213,
+      "step": 3411
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8854129934680033e-05,
+      "loss": 0.1903,
+      "step": 3412
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8853794099373667e-05,
+      "loss": 0.2447,
+      "step": 3413
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8853458264067302e-05,
+      "loss": 0.4148,
+      "step": 3414
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8853122428760937e-05,
+      "loss": 0.2736,
+      "step": 3415
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.885278659345457e-05,
+      "loss": 0.2958,
+      "step": 3416
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8852450758148206e-05,
+      "loss": 0.316,
+      "step": 3417
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.885211492284184e-05,
+      "loss": 0.4246,
+      "step": 3418
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8851779087535472e-05,
+      "loss": 0.6284,
+      "step": 3419
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.885144325222911e-05,
+      "loss": 0.453,
+      "step": 3420
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.885110741692274e-05,
+      "loss": 0.4579,
+      "step": 3421
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8850771581616376e-05,
+      "loss": 0.349,
+      "step": 3422
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.885043574631001e-05,
+      "loss": 0.354,
+      "step": 3423
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8850099911003645e-05,
+      "loss": 0.4249,
+      "step": 3424
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.884976407569728e-05,
+      "loss": 0.4457,
+      "step": 3425
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8849428240390914e-05,
+      "loss": 0.4088,
+      "step": 3426
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.884909240508455e-05,
+      "loss": 0.4236,
+      "step": 3427
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.884875656977818e-05,
+      "loss": 0.367,
+      "step": 3428
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.884842073447182e-05,
+      "loss": 0.3456,
+      "step": 3429
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.884808489916545e-05,
+      "loss": 0.3983,
+      "step": 3430
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8847749063859084e-05,
+      "loss": 0.4843,
+      "step": 3431
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.884741322855272e-05,
+      "loss": 0.3315,
+      "step": 3432
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8847077393246354e-05,
+      "loss": 0.3854,
+      "step": 3433
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8846741557939988e-05,
+      "loss": 0.5295,
+      "step": 3434
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8846405722633623e-05,
+      "loss": 0.4261,
+      "step": 3435
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8846069887327258e-05,
+      "loss": 0.2284,
+      "step": 3436
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.884573405202089e-05,
+      "loss": 0.4642,
+      "step": 3437
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8845398216714527e-05,
+      "loss": 0.3934,
+      "step": 3438
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8845062381408158e-05,
+      "loss": 0.3698,
+      "step": 3439
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8844726546101793e-05,
+      "loss": 0.3854,
+      "step": 3440
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8844390710795427e-05,
+      "loss": 0.433,
+      "step": 3441
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8844054875489062e-05,
+      "loss": 0.3083,
+      "step": 3442
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8843719040182697e-05,
+      "loss": 0.3323,
+      "step": 3443
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.884338320487633e-05,
+      "loss": 0.3937,
+      "step": 3444
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8843047369569966e-05,
+      "loss": 0.4599,
+      "step": 3445
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8842711534263597e-05,
+      "loss": 0.3545,
+      "step": 3446
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8842375698957235e-05,
+      "loss": 0.4082,
+      "step": 3447
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8842039863650867e-05,
+      "loss": 0.5037,
+      "step": 3448
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.88417040283445e-05,
+      "loss": 0.336,
+      "step": 3449
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8841368193038136e-05,
+      "loss": 0.3994,
+      "step": 3450
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.884103235773177e-05,
+      "loss": 0.2979,
+      "step": 3451
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8840696522425405e-05,
+      "loss": 0.4419,
+      "step": 3452
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.884036068711904e-05,
+      "loss": 0.311,
+      "step": 3453
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.884002485181267e-05,
+      "loss": 0.3169,
+      "step": 3454
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8839689016506306e-05,
+      "loss": 0.5368,
+      "step": 3455
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.883935318119994e-05,
+      "loss": 0.2895,
+      "step": 3456
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8839017345893575e-05,
+      "loss": 0.3512,
+      "step": 3457
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.883868151058721e-05,
+      "loss": 0.2456,
+      "step": 3458
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8838345675280844e-05,
+      "loss": 0.2819,
+      "step": 3459
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.883800983997448e-05,
+      "loss": 0.2367,
+      "step": 3460
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.883767400466811e-05,
+      "loss": 0.3785,
+      "step": 3461
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8837338169361748e-05,
+      "loss": 0.3285,
+      "step": 3462
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.883700233405538e-05,
+      "loss": 0.3059,
+      "step": 3463
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8836666498749014e-05,
+      "loss": 0.4481,
+      "step": 3464
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.883633066344265e-05,
+      "loss": 0.1674,
+      "step": 3465
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8835994828136283e-05,
+      "loss": 0.3025,
+      "step": 3466
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8835658992829918e-05,
+      "loss": 0.4407,
+      "step": 3467
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8835323157523553e-05,
+      "loss": 0.4118,
+      "step": 3468
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8834987322217187e-05,
+      "loss": 0.3146,
+      "step": 3469
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.883465148691082e-05,
+      "loss": 0.3105,
+      "step": 3470
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8834315651604457e-05,
+      "loss": 0.2951,
+      "step": 3471
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8833979816298088e-05,
+      "loss": 0.4772,
+      "step": 3472
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8833643980991723e-05,
+      "loss": 0.345,
+      "step": 3473
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8833308145685357e-05,
+      "loss": 0.4702,
+      "step": 3474
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8832972310378992e-05,
+      "loss": 0.562,
+      "step": 3475
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8832636475072626e-05,
+      "loss": 0.2968,
+      "step": 3476
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883230063976626e-05,
+      "loss": 0.3517,
+      "step": 3477
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8831964804459896e-05,
+      "loss": 0.3745,
+      "step": 3478
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8831628969153527e-05,
+      "loss": 0.2798,
+      "step": 3479
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8831293133847165e-05,
+      "loss": 0.4143,
+      "step": 3480
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8830957298540796e-05,
+      "loss": 0.4561,
+      "step": 3481
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883062146323443e-05,
+      "loss": 0.2398,
+      "step": 3482
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8830285627928066e-05,
+      "loss": 0.3397,
+      "step": 3483
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88299497926217e-05,
+      "loss": 0.3024,
+      "step": 3484
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8829613957315335e-05,
+      "loss": 0.2421,
+      "step": 3485
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882927812200897e-05,
+      "loss": 0.2961,
+      "step": 3486
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8828942286702604e-05,
+      "loss": 0.2024,
+      "step": 3487
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8828606451396235e-05,
+      "loss": 0.483,
+      "step": 3488
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8828270616089874e-05,
+      "loss": 0.3094,
+      "step": 3489
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8827934780783505e-05,
+      "loss": 0.3517,
+      "step": 3490
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882759894547714e-05,
+      "loss": 0.6248,
+      "step": 3491
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8827263110170774e-05,
+      "loss": 0.5534,
+      "step": 3492
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882692727486441e-05,
+      "loss": 0.2488,
+      "step": 3493
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8826591439558043e-05,
+      "loss": 0.3004,
+      "step": 3494
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8826255604251678e-05,
+      "loss": 0.4224,
+      "step": 3495
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8825919768945313e-05,
+      "loss": 0.3133,
+      "step": 3496
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8825583933638944e-05,
+      "loss": 0.2901,
+      "step": 3497
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8825248098332582e-05,
+      "loss": 0.3447,
+      "step": 3498
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8824912263026213e-05,
+      "loss": 0.3327,
+      "step": 3499
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8824576427719848e-05,
+      "loss": 0.3608,
+      "step": 3500
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8824240592413483e-05,
+      "loss": 0.3479,
+      "step": 3501
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8823904757107117e-05,
+      "loss": 0.3808,
+      "step": 3502
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882356892180075e-05,
+      "loss": 0.3281,
+      "step": 3503
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8823233086494386e-05,
+      "loss": 0.5353,
+      "step": 3504
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8822897251188018e-05,
+      "loss": 0.307,
+      "step": 3505
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8822561415881652e-05,
+      "loss": 0.3671,
+      "step": 3506
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8822225580575287e-05,
+      "loss": 0.5478,
+      "step": 3507
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882188974526892e-05,
+      "loss": 0.2644,
+      "step": 3508
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8821553909962556e-05,
+      "loss": 0.3217,
+      "step": 3509
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882121807465619e-05,
+      "loss": 0.3897,
+      "step": 3510
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8820882239349826e-05,
+      "loss": 0.35,
+      "step": 3511
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8820546404043457e-05,
+      "loss": 0.3124,
+      "step": 3512
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8820210568737095e-05,
+      "loss": 0.3527,
+      "step": 3513
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8819874733430726e-05,
+      "loss": 0.3432,
+      "step": 3514
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881953889812436e-05,
+      "loss": 0.2683,
+      "step": 3515
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8819203062817995e-05,
+      "loss": 0.4321,
+      "step": 3516
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881886722751163e-05,
+      "loss": 0.4416,
+      "step": 3517
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8818531392205265e-05,
+      "loss": 0.4174,
+      "step": 3518
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88181955568989e-05,
+      "loss": 0.3398,
+      "step": 3519
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8817859721592534e-05,
+      "loss": 0.5794,
+      "step": 3520
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8817523886286165e-05,
+      "loss": 0.3717,
+      "step": 3521
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88171880509798e-05,
+      "loss": 0.3109,
+      "step": 3522
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8816852215673435e-05,
+      "loss": 0.4044,
+      "step": 3523
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881651638036707e-05,
+      "loss": 0.3236,
+      "step": 3524
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8816180545060704e-05,
+      "loss": 0.2613,
+      "step": 3525
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881584470975434e-05,
+      "loss": 0.3624,
+      "step": 3526
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8815508874447973e-05,
+      "loss": 0.3052,
+      "step": 3527
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8815173039141604e-05,
+      "loss": 0.3196,
+      "step": 3528
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8814837203835242e-05,
+      "loss": 0.5854,
+      "step": 3529
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8814501368528874e-05,
+      "loss": 0.3266,
+      "step": 3530
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881416553322251e-05,
+      "loss": 0.3061,
+      "step": 3531
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8813829697916143e-05,
+      "loss": 0.1735,
+      "step": 3532
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8813493862609778e-05,
+      "loss": 0.311,
+      "step": 3533
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8813158027303412e-05,
+      "loss": 0.2016,
+      "step": 3534
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8812822191997047e-05,
+      "loss": 0.4094,
+      "step": 3535
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881248635669068e-05,
+      "loss": 0.4102,
+      "step": 3536
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8812150521384313e-05,
+      "loss": 0.2427,
+      "step": 3537
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881181468607795e-05,
+      "loss": 0.3884,
+      "step": 3538
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8811478850771582e-05,
+      "loss": 0.2947,
+      "step": 3539
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8811143015465217e-05,
+      "loss": 0.2182,
+      "step": 3540
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881080718015885e-05,
+      "loss": 0.4193,
+      "step": 3541
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8810471344852486e-05,
+      "loss": 0.3784,
+      "step": 3542
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881013550954612e-05,
+      "loss": 0.3026,
+      "step": 3543
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8809799674239755e-05,
+      "loss": 0.3689,
+      "step": 3544
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8809463838933387e-05,
+      "loss": 0.2803,
+      "step": 3545
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880912800362702e-05,
+      "loss": 0.3781,
+      "step": 3546
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8808792168320656e-05,
+      "loss": 0.3441,
+      "step": 3547
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880845633301429e-05,
+      "loss": 0.4313,
+      "step": 3548
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8808120497707925e-05,
+      "loss": 0.2279,
+      "step": 3549
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880778466240156e-05,
+      "loss": 0.5023,
+      "step": 3550
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8807448827095195e-05,
+      "loss": 0.2605,
+      "step": 3551
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8807112991788826e-05,
+      "loss": 0.4273,
+      "step": 3552
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8806777156482464e-05,
+      "loss": 0.5161,
+      "step": 3553
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8806441321176095e-05,
+      "loss": 0.3,
+      "step": 3554
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880610548586973e-05,
+      "loss": 0.3661,
+      "step": 3555
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8805769650563364e-05,
+      "loss": 0.3951,
+      "step": 3556
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8805433815257e-05,
+      "loss": 0.514,
+      "step": 3557
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8805097979950634e-05,
+      "loss": 0.3609,
+      "step": 3558
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880476214464427e-05,
+      "loss": 0.465,
+      "step": 3559
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8804426309337903e-05,
+      "loss": 0.4848,
+      "step": 3560
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8804090474031534e-05,
+      "loss": 0.3799,
+      "step": 3561
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8803754638725172e-05,
+      "loss": 0.5257,
+      "step": 3562
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8803418803418804e-05,
+      "loss": 0.5234,
+      "step": 3563
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8803082968112438e-05,
+      "loss": 0.2729,
+      "step": 3564
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8802747132806073e-05,
+      "loss": 0.2413,
+      "step": 3565
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8802411297499707e-05,
+      "loss": 0.394,
+      "step": 3566
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8802075462193342e-05,
+      "loss": 0.2566,
+      "step": 3567
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8801739626886977e-05,
+      "loss": 0.4714,
+      "step": 3568
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880140379158061e-05,
+      "loss": 0.2646,
+      "step": 3569
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8801067956274243e-05,
+      "loss": 0.3202,
+      "step": 3570
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880073212096788e-05,
+      "loss": 0.356,
+      "step": 3571
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8800396285661512e-05,
+      "loss": 0.2668,
+      "step": 3572
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8800060450355147e-05,
+      "loss": 0.3851,
+      "step": 3573
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879972461504878e-05,
+      "loss": 0.496,
+      "step": 3574
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8799388779742416e-05,
+      "loss": 0.2711,
+      "step": 3575
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879905294443605e-05,
+      "loss": 0.3927,
+      "step": 3576
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8798717109129685e-05,
+      "loss": 0.3216,
+      "step": 3577
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879838127382332e-05,
+      "loss": 0.4978,
+      "step": 3578
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879804543851695e-05,
+      "loss": 0.4094,
+      "step": 3579
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879770960321059e-05,
+      "loss": 0.183,
+      "step": 3580
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879737376790422e-05,
+      "loss": 0.3097,
+      "step": 3581
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8797037932597855e-05,
+      "loss": 0.2872,
+      "step": 3582
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879670209729149e-05,
+      "loss": 0.5185,
+      "step": 3583
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8796366261985124e-05,
+      "loss": 0.4284,
+      "step": 3584
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879603042667876e-05,
+      "loss": 0.2785,
+      "step": 3585
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8795694591372394e-05,
+      "loss": 0.3845,
+      "step": 3586
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8795358756066028e-05,
+      "loss": 0.3032,
+      "step": 3587
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879502292075966e-05,
+      "loss": 0.4931,
+      "step": 3588
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8794687085453294e-05,
+      "loss": 0.547,
+      "step": 3589
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879435125014693e-05,
+      "loss": 0.4934,
+      "step": 3590
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8794015414840564e-05,
+      "loss": 0.3391,
+      "step": 3591
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8793679579534198e-05,
+      "loss": 0.3157,
+      "step": 3592
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8793343744227833e-05,
+      "loss": 0.3197,
+      "step": 3593
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8793007908921464e-05,
+      "loss": 0.381,
+      "step": 3594
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8792672073615102e-05,
+      "loss": 0.2343,
+      "step": 3595
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8792336238308733e-05,
+      "loss": 0.2991,
+      "step": 3596
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8792000403002368e-05,
+      "loss": 0.3468,
+      "step": 3597
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8791664567696003e-05,
+      "loss": 0.3802,
+      "step": 3598
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8791328732389637e-05,
+      "loss": 0.3173,
+      "step": 3599
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8790992897083272e-05,
+      "loss": 0.4021,
+      "step": 3600
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8790657061776907e-05,
+      "loss": 0.3356,
+      "step": 3601
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879032122647054e-05,
+      "loss": 0.4497,
+      "step": 3602
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8789985391164172e-05,
+      "loss": 0.3594,
+      "step": 3603
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878964955585781e-05,
+      "loss": 0.3478,
+      "step": 3604
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8789313720551442e-05,
+      "loss": 0.3796,
+      "step": 3605
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8788977885245076e-05,
+      "loss": 0.3671,
+      "step": 3606
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878864204993871e-05,
+      "loss": 0.2629,
+      "step": 3607
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8788306214632346e-05,
+      "loss": 0.4649,
+      "step": 3608
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878797037932598e-05,
+      "loss": 0.3211,
+      "step": 3609
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8787634544019615e-05,
+      "loss": 0.4096,
+      "step": 3610
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878729870871325e-05,
+      "loss": 0.3743,
+      "step": 3611
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878696287340688e-05,
+      "loss": 0.2278,
+      "step": 3612
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878662703810052e-05,
+      "loss": 0.298,
+      "step": 3613
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878629120279415e-05,
+      "loss": 0.4294,
+      "step": 3614
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8785955367487785e-05,
+      "loss": 0.2978,
+      "step": 3615
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878561953218142e-05,
+      "loss": 0.4944,
+      "step": 3616
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8785283696875054e-05,
+      "loss": 0.427,
+      "step": 3617
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878494786156869e-05,
+      "loss": 0.4845,
+      "step": 3618
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8784612026262323e-05,
+      "loss": 0.3228,
+      "step": 3619
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8784276190955958e-05,
+      "loss": 0.3535,
+      "step": 3620
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878394035564959e-05,
+      "loss": 0.2604,
+      "step": 3621
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8783604520343227e-05,
+      "loss": 0.3771,
+      "step": 3622
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878326868503686e-05,
+      "loss": 0.3605,
+      "step": 3623
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8782932849730493e-05,
+      "loss": 0.3476,
+      "step": 3624
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8782597014424128e-05,
+      "loss": 0.3465,
+      "step": 3625
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8782261179117763e-05,
+      "loss": 0.2195,
+      "step": 3626
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8781925343811397e-05,
+      "loss": 0.263,
+      "step": 3627
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8781589508505032e-05,
+      "loss": 0.5048,
+      "step": 3628
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8781253673198667e-05,
+      "loss": 0.4035,
+      "step": 3629
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8780917837892298e-05,
+      "loss": 0.326,
+      "step": 3630
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8780582002585936e-05,
+      "loss": 0.3071,
+      "step": 3631
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8780246167279567e-05,
+      "loss": 0.4044,
+      "step": 3632
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8779910331973202e-05,
+      "loss": 0.4083,
+      "step": 3633
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8779574496666836e-05,
+      "loss": 0.342,
+      "step": 3634
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877923866136047e-05,
+      "loss": 0.2284,
+      "step": 3635
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8778902826054102e-05,
+      "loss": 0.5167,
+      "step": 3636
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877856699074774e-05,
+      "loss": 0.3493,
+      "step": 3637
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877823115544137e-05,
+      "loss": 0.4482,
+      "step": 3638
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8777895320135006e-05,
+      "loss": 0.4834,
+      "step": 3639
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877755948482864e-05,
+      "loss": 0.2828,
+      "step": 3640
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8777223649522276e-05,
+      "loss": 0.3198,
+      "step": 3641
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877688781421591e-05,
+      "loss": 0.4059,
+      "step": 3642
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8776551978909545e-05,
+      "loss": 0.3301,
+      "step": 3643
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877621614360318e-05,
+      "loss": 0.3911,
+      "step": 3644
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877588030829681e-05,
+      "loss": 0.3315,
+      "step": 3645
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877554447299045e-05,
+      "loss": 0.3947,
+      "step": 3646
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877520863768408e-05,
+      "loss": 0.3334,
+      "step": 3647
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8774872802377715e-05,
+      "loss": 0.3259,
+      "step": 3648
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877453696707135e-05,
+      "loss": 0.434,
+      "step": 3649
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8774201131764984e-05,
+      "loss": 0.363,
+      "step": 3650
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877386529645862e-05,
+      "loss": 0.3558,
+      "step": 3651
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8773529461152253e-05,
+      "loss": 0.301,
+      "step": 3652
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8773193625845888e-05,
+      "loss": 0.2227,
+      "step": 3653
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877285779053952e-05,
+      "loss": 0.5683,
+      "step": 3654
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8772521955233157e-05,
+      "loss": 0.4589,
+      "step": 3655
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877218611992679e-05,
+      "loss": 0.3195,
+      "step": 3656
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8771850284620423e-05,
+      "loss": 0.3838,
+      "step": 3657
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8771514449314058e-05,
+      "loss": 0.4259,
+      "step": 3658
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8771178614007692e-05,
+      "loss": 0.2257,
+      "step": 3659
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8770842778701327e-05,
+      "loss": 0.3408,
+      "step": 3660
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8770506943394962e-05,
+      "loss": 0.409,
+      "step": 3661
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8770171108088596e-05,
+      "loss": 0.4846,
+      "step": 3662
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8769835272782228e-05,
+      "loss": 0.3223,
+      "step": 3663
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8769499437475866e-05,
+      "loss": 0.3137,
+      "step": 3664
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8769163602169497e-05,
+      "loss": 0.4856,
+      "step": 3665
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.876882776686313e-05,
+      "loss": 0.3441,
+      "step": 3666
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8768491931556766e-05,
+      "loss": 0.1972,
+      "step": 3667
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.87681560962504e-05,
+      "loss": 0.1971,
+      "step": 3668
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8767820260944036e-05,
+      "loss": 0.3744,
+      "step": 3669
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.876748442563767e-05,
+      "loss": 0.2166,
+      "step": 3670
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8767148590331305e-05,
+      "loss": 0.1724,
+      "step": 3671
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8766812755024936e-05,
+      "loss": 0.4103,
+      "step": 3672
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8766476919718574e-05,
+      "loss": 0.302,
+      "step": 3673
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8766141084412205e-05,
+      "loss": 0.2814,
+      "step": 3674
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876580524910584e-05,
+      "loss": 0.3389,
+      "step": 3675
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8765469413799475e-05,
+      "loss": 0.4471,
+      "step": 3676
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876513357849311e-05,
+      "loss": 0.3143,
+      "step": 3677
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8764797743186744e-05,
+      "loss": 0.3744,
+      "step": 3678
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876446190788038e-05,
+      "loss": 0.3459,
+      "step": 3679
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876412607257401e-05,
+      "loss": 0.4629,
+      "step": 3680
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8763790237267644e-05,
+      "loss": 0.6438,
+      "step": 3681
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876345440196128e-05,
+      "loss": 0.2842,
+      "step": 3682
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8763118566654914e-05,
+      "loss": 0.208,
+      "step": 3683
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876278273134855e-05,
+      "loss": 0.3572,
+      "step": 3684
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8762446896042183e-05,
+      "loss": 0.2265,
+      "step": 3685
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8762111060735818e-05,
+      "loss": 0.2808,
+      "step": 3686
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876177522542945e-05,
+      "loss": 0.4397,
+      "step": 3687
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8761439390123087e-05,
+      "loss": 0.4912,
+      "step": 3688
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8761103554816718e-05,
+      "loss": 0.2878,
+      "step": 3689
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8760767719510353e-05,
+      "loss": 0.2804,
+      "step": 3690
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8760431884203988e-05,
+      "loss": 0.4116,
+      "step": 3691
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8760096048897622e-05,
+      "loss": 0.4989,
+      "step": 3692
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8759760213591257e-05,
+      "loss": 0.3551,
+      "step": 3693
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875942437828489e-05,
+      "loss": 0.2676,
+      "step": 3694
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8759088542978526e-05,
+      "loss": 0.3507,
+      "step": 3695
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8758752707672157e-05,
+      "loss": 0.2533,
+      "step": 3696
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8758416872365795e-05,
+      "loss": 0.2494,
+      "step": 3697
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8758081037059427e-05,
+      "loss": 0.5251,
+      "step": 3698
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875774520175306e-05,
+      "loss": 0.3867,
+      "step": 3699
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8757409366446696e-05,
+      "loss": 0.3555,
+      "step": 3700
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875707353114033e-05,
+      "loss": 0.3822,
+      "step": 3701
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8756737695833965e-05,
+      "loss": 0.3016,
+      "step": 3702
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8756401860527597e-05,
+      "loss": 0.4395,
+      "step": 3703
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8756066025221235e-05,
+      "loss": 0.2726,
+      "step": 3704
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8755730189914866e-05,
+      "loss": 0.3665,
+      "step": 3705
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87553943546085e-05,
+      "loss": 0.229,
+      "step": 3706
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8755058519302135e-05,
+      "loss": 0.5357,
+      "step": 3707
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875472268399577e-05,
+      "loss": 0.4313,
+      "step": 3708
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8754386848689404e-05,
+      "loss": 0.1369,
+      "step": 3709
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875405101338304e-05,
+      "loss": 0.3679,
+      "step": 3710
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8753715178076674e-05,
+      "loss": 0.4627,
+      "step": 3711
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8753379342770305e-05,
+      "loss": 0.4541,
+      "step": 3712
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8753043507463943e-05,
+      "loss": 0.3847,
+      "step": 3713
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8752707672157574e-05,
+      "loss": 0.333,
+      "step": 3714
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875237183685121e-05,
+      "loss": 0.2284,
+      "step": 3715
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8752036001544844e-05,
+      "loss": 0.3765,
+      "step": 3716
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8751700166238478e-05,
+      "loss": 0.364,
+      "step": 3717
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8751364330932113e-05,
+      "loss": 0.2976,
+      "step": 3718
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8751028495625748e-05,
+      "loss": 0.4409,
+      "step": 3719
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8750692660319382e-05,
+      "loss": 0.322,
+      "step": 3720
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8750356825013013e-05,
+      "loss": 0.3811,
+      "step": 3721
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875002098970665e-05,
+      "loss": 0.2458,
+      "step": 3722
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8749685154400283e-05,
+      "loss": 0.3598,
+      "step": 3723
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8749349319093917e-05,
+      "loss": 0.4685,
+      "step": 3724
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8749013483787552e-05,
+      "loss": 0.4746,
+      "step": 3725
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8748677648481187e-05,
+      "loss": 0.286,
+      "step": 3726
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8748341813174818e-05,
+      "loss": 0.3707,
+      "step": 3727
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8748005977868456e-05,
+      "loss": 0.1974,
+      "step": 3728
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8747670142562087e-05,
+      "loss": 0.3591,
+      "step": 3729
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8747334307255722e-05,
+      "loss": 0.3474,
+      "step": 3730
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8746998471949357e-05,
+      "loss": 0.299,
+      "step": 3731
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874666263664299e-05,
+      "loss": 0.267,
+      "step": 3732
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8746326801336626e-05,
+      "loss": 0.4166,
+      "step": 3733
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874599096603026e-05,
+      "loss": 0.3458,
+      "step": 3734
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8745655130723895e-05,
+      "loss": 0.225,
+      "step": 3735
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8745319295417526e-05,
+      "loss": 0.4385,
+      "step": 3736
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8744983460111164e-05,
+      "loss": 0.518,
+      "step": 3737
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8744647624804796e-05,
+      "loss": 0.2381,
+      "step": 3738
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874431178949843e-05,
+      "loss": 0.3513,
+      "step": 3739
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8743975954192065e-05,
+      "loss": 0.3371,
+      "step": 3740
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87436401188857e-05,
+      "loss": 0.2305,
+      "step": 3741
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8743304283579334e-05,
+      "loss": 0.2921,
+      "step": 3742
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874296844827297e-05,
+      "loss": 0.3883,
+      "step": 3743
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8742632612966604e-05,
+      "loss": 0.3512,
+      "step": 3744
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8742296777660235e-05,
+      "loss": 0.3824,
+      "step": 3745
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8741960942353873e-05,
+      "loss": 0.5473,
+      "step": 3746
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8741625107047504e-05,
+      "loss": 0.1663,
+      "step": 3747
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874128927174114e-05,
+      "loss": 0.4319,
+      "step": 3748
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8740953436434773e-05,
+      "loss": 0.4348,
+      "step": 3749
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8740617601128408e-05,
+      "loss": 0.2989,
+      "step": 3750
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8740281765822043e-05,
+      "loss": 0.3705,
+      "step": 3751
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8739945930515677e-05,
+      "loss": 0.3123,
+      "step": 3752
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8739610095209312e-05,
+      "loss": 0.4804,
+      "step": 3753
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8739274259902943e-05,
+      "loss": 0.3536,
+      "step": 3754
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873893842459658e-05,
+      "loss": 0.2767,
+      "step": 3755
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8738602589290213e-05,
+      "loss": 0.3471,
+      "step": 3756
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8738266753983847e-05,
+      "loss": 0.3296,
+      "step": 3757
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8737930918677482e-05,
+      "loss": 0.2436,
+      "step": 3758
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8737595083371116e-05,
+      "loss": 0.3312,
+      "step": 3759
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873725924806475e-05,
+      "loss": 0.4558,
+      "step": 3760
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8736923412758386e-05,
+      "loss": 0.4257,
+      "step": 3761
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873658757745202e-05,
+      "loss": 0.3868,
+      "step": 3762
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8736251742145652e-05,
+      "loss": 0.4074,
+      "step": 3763
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873591590683929e-05,
+      "loss": 0.3765,
+      "step": 3764
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873558007153292e-05,
+      "loss": 0.3313,
+      "step": 3765
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8735244236226556e-05,
+      "loss": 0.4075,
+      "step": 3766
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873490840092019e-05,
+      "loss": 0.5735,
+      "step": 3767
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8734572565613825e-05,
+      "loss": 0.4229,
+      "step": 3768
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873423673030746e-05,
+      "loss": 0.2471,
+      "step": 3769
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8733900895001094e-05,
+      "loss": 0.2281,
+      "step": 3770
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8733565059694725e-05,
+      "loss": 0.3553,
+      "step": 3771
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873322922438836e-05,
+      "loss": 0.3291,
+      "step": 3772
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8732893389081995e-05,
+      "loss": 0.4089,
+      "step": 3773
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873255755377563e-05,
+      "loss": 0.2979,
+      "step": 3774
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8732221718469264e-05,
+      "loss": 0.2937,
+      "step": 3775
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87318858831629e-05,
+      "loss": 0.3834,
+      "step": 3776
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8731550047856533e-05,
+      "loss": 0.4854,
+      "step": 3777
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8731214212550165e-05,
+      "loss": 0.3245,
+      "step": 3778
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8730878377243803e-05,
+      "loss": 0.2851,
+      "step": 3779
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8730542541937434e-05,
+      "loss": 0.5614,
+      "step": 3780
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873020670663107e-05,
+      "loss": 0.3783,
+      "step": 3781
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8729870871324703e-05,
+      "loss": 0.3815,
+      "step": 3782
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8729535036018338e-05,
+      "loss": 0.3471,
+      "step": 3783
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8729199200711973e-05,
+      "loss": 0.3226,
+      "step": 3784
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8728863365405607e-05,
+      "loss": 0.3217,
+      "step": 3785
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8728527530099242e-05,
+      "loss": 0.3173,
+      "step": 3786
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8728191694792873e-05,
+      "loss": 0.4737,
+      "step": 3787
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872785585948651e-05,
+      "loss": 0.3191,
+      "step": 3788
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8727520024180142e-05,
+      "loss": 0.4555,
+      "step": 3789
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8727184188873777e-05,
+      "loss": 0.3039,
+      "step": 3790
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872684835356741e-05,
+      "loss": 0.3245,
+      "step": 3791
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8726512518261046e-05,
+      "loss": 0.2547,
+      "step": 3792
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872617668295468e-05,
+      "loss": 0.2927,
+      "step": 3793
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8725840847648316e-05,
+      "loss": 0.4457,
+      "step": 3794
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872550501234195e-05,
+      "loss": 0.3662,
+      "step": 3795
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872516917703558e-05,
+      "loss": 0.4862,
+      "step": 3796
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872483334172922e-05,
+      "loss": 0.226,
+      "step": 3797
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872449750642285e-05,
+      "loss": 0.368,
+      "step": 3798
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8724161671116485e-05,
+      "loss": 0.4792,
+      "step": 3799
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872382583581012e-05,
+      "loss": 0.3349,
+      "step": 3800
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8723490000503755e-05,
+      "loss": 0.293,
+      "step": 3801
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872315416519739e-05,
+      "loss": 0.2615,
+      "step": 3802
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8722818329891024e-05,
+      "loss": 0.2538,
+      "step": 3803
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872248249458466e-05,
+      "loss": 0.4199,
+      "step": 3804
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872214665927829e-05,
+      "loss": 0.3116,
+      "step": 3805
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8721810823971928e-05,
+      "loss": 0.6321,
+      "step": 3806
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872147498866556e-05,
+      "loss": 0.2636,
+      "step": 3807
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8721139153359194e-05,
+      "loss": 0.2122,
+      "step": 3808
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872080331805283e-05,
+      "loss": 0.2067,
+      "step": 3809
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8720467482746463e-05,
+      "loss": 0.438,
+      "step": 3810
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8720131647440098e-05,
+      "loss": 0.2794,
+      "step": 3811
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8719795812133732e-05,
+      "loss": 0.4982,
+      "step": 3812
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8719459976827367e-05,
+      "loss": 0.1729,
+      "step": 3813
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8719124141521e-05,
+      "loss": 0.2185,
+      "step": 3814
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8718788306214633e-05,
+      "loss": 0.3039,
+      "step": 3815
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8718452470908268e-05,
+      "loss": 0.4137,
+      "step": 3816
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8718116635601902e-05,
+      "loss": 0.4283,
+      "step": 3817
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8717780800295537e-05,
+      "loss": 0.3321,
+      "step": 3818
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871744496498917e-05,
+      "loss": 0.2256,
+      "step": 3819
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8717109129682803e-05,
+      "loss": 0.4411,
+      "step": 3820
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871677329437644e-05,
+      "loss": 0.4249,
+      "step": 3821
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8716437459070072e-05,
+      "loss": 0.4013,
+      "step": 3822
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8716101623763707e-05,
+      "loss": 0.3782,
+      "step": 3823
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871576578845734e-05,
+      "loss": 0.3247,
+      "step": 3824
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8715429953150976e-05,
+      "loss": 0.4711,
+      "step": 3825
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871509411784461e-05,
+      "loss": 0.3083,
+      "step": 3826
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8714758282538245e-05,
+      "loss": 0.339,
+      "step": 3827
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871442244723188e-05,
+      "loss": 0.3235,
+      "step": 3828
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871408661192551e-05,
+      "loss": 0.4118,
+      "step": 3829
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871375077661915e-05,
+      "loss": 0.2893,
+      "step": 3830
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871341494131278e-05,
+      "loss": 0.2405,
+      "step": 3831
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8713079106006415e-05,
+      "loss": 0.3854,
+      "step": 3832
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871274327070005e-05,
+      "loss": 0.2427,
+      "step": 3833
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8712407435393685e-05,
+      "loss": 0.2911,
+      "step": 3834
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871207160008732e-05,
+      "loss": 0.1946,
+      "step": 3835
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8711735764780954e-05,
+      "loss": 0.2565,
+      "step": 3836
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871139992947459e-05,
+      "loss": 0.2863,
+      "step": 3837
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871106409416822e-05,
+      "loss": 0.4553,
+      "step": 3838
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8710728258861858e-05,
+      "loss": 0.2777,
+      "step": 3839
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871039242355549e-05,
+      "loss": 0.3688,
+      "step": 3840
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8710056588249124e-05,
+      "loss": 0.4406,
+      "step": 3841
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870972075294276e-05,
+      "loss": 0.3761,
+      "step": 3842
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8709384917636393e-05,
+      "loss": 0.2215,
+      "step": 3843
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8709049082330028e-05,
+      "loss": 0.3105,
+      "step": 3844
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8708713247023662e-05,
+      "loss": 0.4485,
+      "step": 3845
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8708377411717297e-05,
+      "loss": 0.2424,
+      "step": 3846
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8708041576410928e-05,
+      "loss": 0.3543,
+      "step": 3847
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8707705741104566e-05,
+      "loss": 0.2917,
+      "step": 3848
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8707369905798197e-05,
+      "loss": 0.2562,
+      "step": 3849
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8707034070491832e-05,
+      "loss": 0.2386,
+      "step": 3850
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8706698235185467e-05,
+      "loss": 0.3478,
+      "step": 3851
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87063623998791e-05,
+      "loss": 0.2684,
+      "step": 3852
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8706026564572736e-05,
+      "loss": 0.2106,
+      "step": 3853
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870569072926637e-05,
+      "loss": 0.4307,
+      "step": 3854
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8705354893960005e-05,
+      "loss": 0.5751,
+      "step": 3855
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8705019058653637e-05,
+      "loss": 0.391,
+      "step": 3856
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8704683223347275e-05,
+      "loss": 0.3274,
+      "step": 3857
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8704347388040906e-05,
+      "loss": 0.2894,
+      "step": 3858
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870401155273454e-05,
+      "loss": 0.4185,
+      "step": 3859
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8703675717428175e-05,
+      "loss": 0.5448,
+      "step": 3860
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870333988212181e-05,
+      "loss": 0.3116,
+      "step": 3861
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870300404681544e-05,
+      "loss": 0.3753,
+      "step": 3862
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870266821150908e-05,
+      "loss": 0.3898,
+      "step": 3863
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870233237620271e-05,
+      "loss": 0.2415,
+      "step": 3864
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8701996540896345e-05,
+      "loss": 0.3455,
+      "step": 3865
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870166070558998e-05,
+      "loss": 0.2476,
+      "step": 3866
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8701324870283614e-05,
+      "loss": 0.3954,
+      "step": 3867
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870098903497725e-05,
+      "loss": 0.3559,
+      "step": 3868
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8700653199670884e-05,
+      "loss": 0.2375,
+      "step": 3869
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870031736436452e-05,
+      "loss": 0.4556,
+      "step": 3870
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.869998152905815e-05,
+      "loss": 0.2168,
+      "step": 3871
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8699645693751788e-05,
+      "loss": 0.3131,
+      "step": 3872
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.869930985844542e-05,
+      "loss": 0.3063,
+      "step": 3873
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8698974023139054e-05,
+      "loss": 0.3144,
+      "step": 3874
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8698638187832688e-05,
+      "loss": 0.4045,
+      "step": 3875
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8698302352526323e-05,
+      "loss": 0.3849,
+      "step": 3876
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8697966517219957e-05,
+      "loss": 0.4083,
+      "step": 3877
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8697630681913592e-05,
+      "loss": 0.2578,
+      "step": 3878
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8697294846607227e-05,
+      "loss": 0.4281,
+      "step": 3879
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8696959011300858e-05,
+      "loss": 0.3396,
+      "step": 3880
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8696623175994493e-05,
+      "loss": 0.2767,
+      "step": 3881
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8696287340688127e-05,
+      "loss": 0.4123,
+      "step": 3882
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8695951505381762e-05,
+      "loss": 0.4535,
+      "step": 3883
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8695615670075397e-05,
+      "loss": 0.4216,
+      "step": 3884
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.869527983476903e-05,
+      "loss": 0.3175,
+      "step": 3885
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8694943999462666e-05,
+      "loss": 0.1938,
+      "step": 3886
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8694608164156297e-05,
+      "loss": 0.4559,
+      "step": 3887
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8694272328849935e-05,
+      "loss": 0.4526,
+      "step": 3888
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8693936493543566e-05,
+      "loss": 0.3711,
+      "step": 3889
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.86936006582372e-05,
+      "loss": 0.4324,
+      "step": 3890
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8693264822930836e-05,
+      "loss": 0.2867,
+      "step": 3891
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.869292898762447e-05,
+      "loss": 0.2977,
+      "step": 3892
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8692593152318105e-05,
+      "loss": 0.3192,
+      "step": 3893
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.869225731701174e-05,
+      "loss": 0.3266,
+      "step": 3894
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8691921481705374e-05,
+      "loss": 0.3915,
+      "step": 3895
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8691585646399006e-05,
+      "loss": 0.2815,
+      "step": 3896
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8691249811092644e-05,
+      "loss": 0.3221,
+      "step": 3897
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8690913975786275e-05,
+      "loss": 0.4443,
+      "step": 3898
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.869057814047991e-05,
+      "loss": 0.3636,
+      "step": 3899
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8690242305173544e-05,
+      "loss": 0.4319,
+      "step": 3900
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.868990646986718e-05,
+      "loss": 0.4011,
+      "step": 3901
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8689570634560813e-05,
+      "loss": 0.1457,
+      "step": 3902
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8689234799254448e-05,
+      "loss": 0.2962,
+      "step": 3903
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8688898963948083e-05,
+      "loss": 0.3725,
+      "step": 3904
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8688563128641714e-05,
+      "loss": 0.4212,
+      "step": 3905
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.868822729333535e-05,
+      "loss": 0.4943,
+      "step": 3906
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8687891458028983e-05,
+      "loss": 0.3709,
+      "step": 3907
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8687555622722618e-05,
+      "loss": 0.2977,
+      "step": 3908
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8687219787416253e-05,
+      "loss": 0.3899,
+      "step": 3909
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8686883952109887e-05,
+      "loss": 0.261,
+      "step": 3910
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.868654811680352e-05,
+      "loss": 0.5184,
+      "step": 3911
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8686212281497157e-05,
+      "loss": 0.258,
+      "step": 3912
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8685876446190788e-05,
+      "loss": 0.3117,
+      "step": 3913
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8685540610884422e-05,
+      "loss": 0.502,
+      "step": 3914
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8685204775578057e-05,
+      "loss": 0.3891,
+      "step": 3915
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8684868940271692e-05,
+      "loss": 0.3755,
+      "step": 3916
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8684533104965326e-05,
+      "loss": 0.2466,
+      "step": 3917
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.868419726965896e-05,
+      "loss": 0.3609,
+      "step": 3918
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8683861434352596e-05,
+      "loss": 0.3345,
+      "step": 3919
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8683525599046227e-05,
+      "loss": 0.466,
+      "step": 3920
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8683189763739865e-05,
+      "loss": 0.4632,
+      "step": 3921
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8682853928433496e-05,
+      "loss": 0.3643,
+      "step": 3922
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.868251809312713e-05,
+      "loss": 0.2286,
+      "step": 3923
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8682182257820766e-05,
+      "loss": 0.3485,
+      "step": 3924
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.86818464225144e-05,
+      "loss": 0.3513,
+      "step": 3925
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8681510587208035e-05,
+      "loss": 0.2775,
+      "step": 3926
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.868117475190167e-05,
+      "loss": 0.1978,
+      "step": 3927
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8680838916595304e-05,
+      "loss": 0.4673,
+      "step": 3928
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8680503081288935e-05,
+      "loss": 0.2675,
+      "step": 3929
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8680167245982573e-05,
+      "loss": 0.2853,
+      "step": 3930
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8679831410676205e-05,
+      "loss": 0.5152,
+      "step": 3931
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.867949557536984e-05,
+      "loss": 0.3082,
+      "step": 3932
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8679159740063474e-05,
+      "loss": 0.2205,
+      "step": 3933
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.867882390475711e-05,
+      "loss": 0.4801,
+      "step": 3934
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8678488069450743e-05,
+      "loss": 0.4266,
+      "step": 3935
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8678152234144378e-05,
+      "loss": 0.5741,
+      "step": 3936
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8677816398838013e-05,
+      "loss": 0.4861,
+      "step": 3937
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8677480563531644e-05,
+      "loss": 0.2585,
+      "step": 3938
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8677144728225282e-05,
+      "loss": 0.2325,
+      "step": 3939
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8676808892918913e-05,
+      "loss": 0.2786,
+      "step": 3940
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8676473057612548e-05,
+      "loss": 0.393,
+      "step": 3941
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8676137222306182e-05,
+      "loss": 0.3937,
+      "step": 3942
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8675801386999817e-05,
+      "loss": 0.3566,
+      "step": 3943
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8675465551693452e-05,
+      "loss": 0.4461,
+      "step": 3944
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8675129716387086e-05,
+      "loss": 0.2894,
+      "step": 3945
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.867479388108072e-05,
+      "loss": 0.5766,
+      "step": 3946
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8674458045774352e-05,
+      "loss": 0.3186,
+      "step": 3947
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.867412221046799e-05,
+      "loss": 0.3314,
+      "step": 3948
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.867378637516162e-05,
+      "loss": 0.3359,
+      "step": 3949
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8673450539855256e-05,
+      "loss": 0.4581,
+      "step": 3950
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.867311470454889e-05,
+      "loss": 0.4607,
+      "step": 3951
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8672778869242526e-05,
+      "loss": 0.1669,
+      "step": 3952
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8672443033936157e-05,
+      "loss": 0.258,
+      "step": 3953
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8672107198629795e-05,
+      "loss": 0.2542,
+      "step": 3954
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8671771363323426e-05,
+      "loss": 0.2764,
+      "step": 3955
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.867143552801706e-05,
+      "loss": 0.3559,
+      "step": 3956
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8671099692710695e-05,
+      "loss": 0.4327,
+      "step": 3957
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.867076385740433e-05,
+      "loss": 0.3822,
+      "step": 3958
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8670428022097965e-05,
+      "loss": 0.268,
+      "step": 3959
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.86700921867916e-05,
+      "loss": 0.5328,
+      "step": 3960
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8669756351485234e-05,
+      "loss": 0.3111,
+      "step": 3961
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8669420516178865e-05,
+      "loss": 0.3914,
+      "step": 3962
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8669084680872503e-05,
+      "loss": 0.2623,
+      "step": 3963
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8668748845566135e-05,
+      "loss": 0.4415,
+      "step": 3964
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.866841301025977e-05,
+      "loss": 0.2642,
+      "step": 3965
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8668077174953404e-05,
+      "loss": 0.3884,
+      "step": 3966
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.866774133964704e-05,
+      "loss": 0.3477,
+      "step": 3967
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8667405504340673e-05,
+      "loss": 0.2758,
+      "step": 3968
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8667069669034308e-05,
+      "loss": 0.4278,
+      "step": 3969
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8666733833727942e-05,
+      "loss": 0.3134,
+      "step": 3970
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8666397998421574e-05,
+      "loss": 0.4476,
+      "step": 3971
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.866606216311521e-05,
+      "loss": 0.3936,
+      "step": 3972
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8665726327808843e-05,
+      "loss": 0.3244,
+      "step": 3973
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8665390492502478e-05,
+      "loss": 0.3147,
+      "step": 3974
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8665054657196112e-05,
+      "loss": 0.2648,
+      "step": 3975
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8664718821889747e-05,
+      "loss": 0.3236,
+      "step": 3976
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.866438298658338e-05,
+      "loss": 0.2687,
+      "step": 3977
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8664047151277016e-05,
+      "loss": 0.2903,
+      "step": 3978
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.866371131597065e-05,
+      "loss": 0.2775,
+      "step": 3979
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8663375480664282e-05,
+      "loss": 0.333,
+      "step": 3980
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.866303964535792e-05,
+      "loss": 0.3427,
+      "step": 3981
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.866270381005155e-05,
+      "loss": 0.3675,
+      "step": 3982
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8662367974745186e-05,
+      "loss": 0.4253,
+      "step": 3983
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.866203213943882e-05,
+      "loss": 0.2884,
+      "step": 3984
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8661696304132455e-05,
+      "loss": 0.3347,
+      "step": 3985
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.866136046882609e-05,
+      "loss": 0.3305,
+      "step": 3986
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8661024633519725e-05,
+      "loss": 0.471,
+      "step": 3987
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.866068879821336e-05,
+      "loss": 0.2902,
+      "step": 3988
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.866035296290699e-05,
+      "loss": 0.4457,
+      "step": 3989
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.866001712760063e-05,
+      "loss": 0.2449,
+      "step": 3990
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.865968129229426e-05,
+      "loss": 0.2932,
+      "step": 3991
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8659345456987894e-05,
+      "loss": 0.3372,
+      "step": 3992
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.865900962168153e-05,
+      "loss": 0.4555,
+      "step": 3993
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8658673786375164e-05,
+      "loss": 0.3963,
+      "step": 3994
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.86583379510688e-05,
+      "loss": 0.3765,
+      "step": 3995
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8658002115762433e-05,
+      "loss": 0.3029,
+      "step": 3996
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8657666280456064e-05,
+      "loss": 0.4218,
+      "step": 3997
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.86573304451497e-05,
+      "loss": 0.2965,
+      "step": 3998
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8656994609843334e-05,
+      "loss": 0.337,
+      "step": 3999
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8656658774536968e-05,
+      "loss": 0.37,
+      "step": 4000
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8656322939230603e-05,
+      "loss": 0.3102,
+      "step": 4001
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8655987103924238e-05,
+      "loss": 0.3016,
+      "step": 4002
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8655651268617872e-05,
+      "loss": 0.5893,
+      "step": 4003
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8655315433311503e-05,
+      "loss": 0.5967,
+      "step": 4004
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.865497959800514e-05,
+      "loss": 0.4526,
+      "step": 4005
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8654643762698773e-05,
+      "loss": 0.3671,
+      "step": 4006
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8654307927392407e-05,
+      "loss": 0.3297,
+      "step": 4007
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8653972092086042e-05,
+      "loss": 0.504,
+      "step": 4008
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8653636256779677e-05,
+      "loss": 0.3944,
+      "step": 4009
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.865330042147331e-05,
+      "loss": 0.372,
+      "step": 4010
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8652964586166946e-05,
+      "loss": 0.3434,
+      "step": 4011
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.865262875086058e-05,
+      "loss": 0.1983,
+      "step": 4012
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8652292915554212e-05,
+      "loss": 0.4302,
+      "step": 4013
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.865195708024785e-05,
+      "loss": 0.2862,
+      "step": 4014
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.865162124494148e-05,
+      "loss": 0.4393,
+      "step": 4015
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8651285409635116e-05,
+      "loss": 0.3977,
+      "step": 4016
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.865094957432875e-05,
+      "loss": 0.3517,
+      "step": 4017
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8650613739022385e-05,
+      "loss": 0.3619,
+      "step": 4018
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.865027790371602e-05,
+      "loss": 0.4467,
+      "step": 4019
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8649942068409654e-05,
+      "loss": 0.2468,
+      "step": 4020
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.864960623310329e-05,
+      "loss": 0.265,
+      "step": 4021
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.864927039779692e-05,
+      "loss": 0.1818,
+      "step": 4022
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.864893456249056e-05,
+      "loss": 0.4475,
+      "step": 4023
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.864859872718419e-05,
+      "loss": 0.1899,
+      "step": 4024
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8648262891877824e-05,
+      "loss": 0.2589,
+      "step": 4025
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.864792705657146e-05,
+      "loss": 0.4422,
+      "step": 4026
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8647591221265094e-05,
+      "loss": 0.4477,
+      "step": 4027
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8647255385958728e-05,
+      "loss": 0.3818,
+      "step": 4028
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8646919550652363e-05,
+      "loss": 0.4821,
+      "step": 4029
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8646583715345998e-05,
+      "loss": 0.4123,
+      "step": 4030
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.864624788003963e-05,
+      "loss": 0.3512,
+      "step": 4031
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8645912044733267e-05,
+      "loss": 0.3016,
+      "step": 4032
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8645576209426898e-05,
+      "loss": 0.314,
+      "step": 4033
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8645240374120533e-05,
+      "loss": 0.4566,
+      "step": 4034
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8644904538814167e-05,
+      "loss": 0.4456,
+      "step": 4035
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8644568703507802e-05,
+      "loss": 0.1818,
+      "step": 4036
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8644232868201437e-05,
+      "loss": 0.4136,
+      "step": 4037
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.864389703289507e-05,
+      "loss": 0.4659,
+      "step": 4038
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8643561197588706e-05,
+      "loss": 0.372,
+      "step": 4039
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8643225362282337e-05,
+      "loss": 0.3367,
+      "step": 4040
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8642889526975972e-05,
+      "loss": 0.3197,
+      "step": 4041
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8642553691669607e-05,
+      "loss": 0.3131,
+      "step": 4042
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.864221785636324e-05,
+      "loss": 0.3342,
+      "step": 4043
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8641882021056876e-05,
+      "loss": 0.2772,
+      "step": 4044
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.864154618575051e-05,
+      "loss": 0.3409,
+      "step": 4045
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8641210350444142e-05,
+      "loss": 0.4996,
+      "step": 4046
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.864087451513778e-05,
+      "loss": 0.2196,
+      "step": 4047
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.864053867983141e-05,
+      "loss": 0.3498,
+      "step": 4048
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8640202844525046e-05,
+      "loss": 0.4392,
+      "step": 4049
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.863986700921868e-05,
+      "loss": 0.2274,
+      "step": 4050
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8639531173912315e-05,
+      "loss": 0.526,
+      "step": 4051
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.863919533860595e-05,
+      "loss": 0.4758,
+      "step": 4052
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8638859503299584e-05,
+      "loss": 0.3745,
+      "step": 4053
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.863852366799322e-05,
+      "loss": 0.4974,
+      "step": 4054
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.863818783268685e-05,
+      "loss": 0.4126,
+      "step": 4055
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8637851997380488e-05,
+      "loss": 0.4409,
+      "step": 4056
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.863751616207412e-05,
+      "loss": 0.2315,
+      "step": 4057
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8637180326767754e-05,
+      "loss": 0.2841,
+      "step": 4058
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.863684449146139e-05,
+      "loss": 0.4349,
+      "step": 4059
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8636508656155023e-05,
+      "loss": 0.3728,
+      "step": 4060
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8636172820848658e-05,
+      "loss": 0.2883,
+      "step": 4061
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.863583698554229e-05,
+      "loss": 0.4839,
+      "step": 4062
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8635501150235927e-05,
+      "loss": 0.2636,
+      "step": 4063
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.863516531492956e-05,
+      "loss": 0.3405,
+      "step": 4064
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8634829479623193e-05,
+      "loss": 0.2754,
+      "step": 4065
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8634493644316828e-05,
+      "loss": 0.402,
+      "step": 4066
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8634157809010463e-05,
+      "loss": 0.1565,
+      "step": 4067
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8633821973704097e-05,
+      "loss": 0.2875,
+      "step": 4068
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8633486138397732e-05,
+      "loss": 0.2531,
+      "step": 4069
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8633150303091366e-05,
+      "loss": 0.3623,
+      "step": 4070
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8632814467784998e-05,
+      "loss": 0.2492,
+      "step": 4071
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8632478632478636e-05,
+      "loss": 0.2431,
+      "step": 4072
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8632142797172267e-05,
+      "loss": 0.3579,
+      "step": 4073
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.86318069618659e-05,
+      "loss": 0.4286,
+      "step": 4074
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8631471126559536e-05,
+      "loss": 0.2921,
+      "step": 4075
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.863113529125317e-05,
+      "loss": 0.5181,
+      "step": 4076
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8630799455946806e-05,
+      "loss": 0.255,
+      "step": 4077
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.863046362064044e-05,
+      "loss": 0.3518,
+      "step": 4078
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8630127785334075e-05,
+      "loss": 0.3753,
+      "step": 4079
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8629791950027706e-05,
+      "loss": 0.2769,
+      "step": 4080
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8629456114721344e-05,
+      "loss": 0.3425,
+      "step": 4081
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8629120279414975e-05,
+      "loss": 0.388,
+      "step": 4082
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.862878444410861e-05,
+      "loss": 0.3899,
+      "step": 4083
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8628448608802245e-05,
+      "loss": 0.5784,
+      "step": 4084
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.862811277349588e-05,
+      "loss": 0.257,
+      "step": 4085
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8627776938189514e-05,
+      "loss": 0.3703,
+      "step": 4086
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.862744110288315e-05,
+      "loss": 0.2198,
+      "step": 4087
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.862710526757678e-05,
+      "loss": 0.3776,
+      "step": 4088
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8626769432270415e-05,
+      "loss": 0.3527,
+      "step": 4089
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.862643359696405e-05,
+      "loss": 0.3953,
+      "step": 4090
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8626097761657684e-05,
+      "loss": 0.4342,
+      "step": 4091
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.862576192635132e-05,
+      "loss": 0.3936,
+      "step": 4092
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8625426091044953e-05,
+      "loss": 0.4065,
+      "step": 4093
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8625090255738588e-05,
+      "loss": 0.4415,
+      "step": 4094
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.862475442043222e-05,
+      "loss": 0.4598,
+      "step": 4095
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8624418585125857e-05,
+      "loss": 0.4485,
+      "step": 4096
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.862408274981949e-05,
+      "loss": 0.5902,
+      "step": 4097
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8623746914513123e-05,
+      "loss": 0.3452,
+      "step": 4098
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8623411079206758e-05,
+      "loss": 0.4569,
+      "step": 4099
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8623075243900392e-05,
+      "loss": 0.3969,
+      "step": 4100
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8622739408594027e-05,
+      "loss": 0.308,
+      "step": 4101
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.862240357328766e-05,
+      "loss": 0.3305,
+      "step": 4102
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8622067737981296e-05,
+      "loss": 0.5556,
+      "step": 4103
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8621731902674928e-05,
+      "loss": 0.4851,
+      "step": 4104
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8621396067368566e-05,
+      "loss": 0.4107,
+      "step": 4105
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8621060232062197e-05,
+      "loss": 0.2121,
+      "step": 4106
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.862072439675583e-05,
+      "loss": 0.3831,
+      "step": 4107
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8620388561449466e-05,
+      "loss": 0.3179,
+      "step": 4108
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.86200527261431e-05,
+      "loss": 0.3526,
+      "step": 4109
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8619716890836735e-05,
+      "loss": 0.2965,
+      "step": 4110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.861938105553037e-05,
+      "loss": 0.3888,
+      "step": 4111
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8619045220224005e-05,
+      "loss": 0.27,
+      "step": 4112
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8618709384917636e-05,
+      "loss": 0.47,
+      "step": 4113
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8618373549611274e-05,
+      "loss": 0.2843,
+      "step": 4114
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8618037714304905e-05,
+      "loss": 0.2514,
+      "step": 4115
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.861770187899854e-05,
+      "loss": 0.2686,
+      "step": 4116
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8617366043692175e-05,
+      "loss": 0.2973,
+      "step": 4117
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.861703020838581e-05,
+      "loss": 0.3403,
+      "step": 4118
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8616694373079444e-05,
+      "loss": 0.5055,
+      "step": 4119
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.861635853777308e-05,
+      "loss": 0.311,
+      "step": 4120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8616022702466713e-05,
+      "loss": 0.2293,
+      "step": 4121
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8615686867160344e-05,
+      "loss": 0.3731,
+      "step": 4122
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8615351031853982e-05,
+      "loss": 0.2119,
+      "step": 4123
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8615015196547614e-05,
+      "loss": 0.3353,
+      "step": 4124
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.861467936124125e-05,
+      "loss": 0.2741,
+      "step": 4125
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8614343525934883e-05,
+      "loss": 0.3741,
+      "step": 4126
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8614007690628518e-05,
+      "loss": 0.3561,
+      "step": 4127
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8613671855322152e-05,
+      "loss": 0.3711,
+      "step": 4128
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8613336020015787e-05,
+      "loss": 0.3004,
+      "step": 4129
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.861300018470942e-05,
+      "loss": 0.5317,
+      "step": 4130
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8612664349403053e-05,
+      "loss": 0.202,
+      "step": 4131
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8612328514096688e-05,
+      "loss": 0.1945,
+      "step": 4132
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8611992678790322e-05,
+      "loss": 0.4314,
+      "step": 4133
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8611656843483957e-05,
+      "loss": 0.3813,
+      "step": 4134
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.861132100817759e-05,
+      "loss": 0.2587,
+      "step": 4135
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8610985172871226e-05,
+      "loss": 0.4916,
+      "step": 4136
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8610649337564857e-05,
+      "loss": 0.469,
+      "step": 4137
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8610313502258495e-05,
+      "loss": 0.3239,
+      "step": 4138
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8609977666952127e-05,
+      "loss": 0.3383,
+      "step": 4139
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.860964183164576e-05,
+      "loss": 0.2281,
+      "step": 4140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8609305996339396e-05,
+      "loss": 0.356,
+      "step": 4141
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.860897016103303e-05,
+      "loss": 0.3681,
+      "step": 4142
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8608634325726665e-05,
+      "loss": 0.4439,
+      "step": 4143
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.86082984904203e-05,
+      "loss": 0.4012,
+      "step": 4144
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8607962655113935e-05,
+      "loss": 0.442,
+      "step": 4145
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8607626819807566e-05,
+      "loss": 0.4256,
+      "step": 4146
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8607290984501204e-05,
+      "loss": 0.4364,
+      "step": 4147
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8606955149194835e-05,
+      "loss": 0.4251,
+      "step": 4148
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.860661931388847e-05,
+      "loss": 0.352,
+      "step": 4149
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8606283478582104e-05,
+      "loss": 0.3957,
+      "step": 4150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.860594764327574e-05,
+      "loss": 0.3345,
+      "step": 4151
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8605611807969374e-05,
+      "loss": 0.2879,
+      "step": 4152
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.860527597266301e-05,
+      "loss": 0.3166,
+      "step": 4153
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8604940137356643e-05,
+      "loss": 0.4347,
+      "step": 4154
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8604604302050274e-05,
+      "loss": 0.3415,
+      "step": 4155
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8604268466743912e-05,
+      "loss": 0.5198,
+      "step": 4156
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8603932631437544e-05,
+      "loss": 0.4821,
+      "step": 4157
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8603596796131178e-05,
+      "loss": 0.3939,
+      "step": 4158
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8603260960824813e-05,
+      "loss": 0.2078,
+      "step": 4159
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8602925125518447e-05,
+      "loss": 0.4712,
+      "step": 4160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8602589290212082e-05,
+      "loss": 0.3505,
+      "step": 4161
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8602253454905717e-05,
+      "loss": 0.4091,
+      "step": 4162
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.860191761959935e-05,
+      "loss": 0.5962,
+      "step": 4163
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8601581784292983e-05,
+      "loss": 0.2893,
+      "step": 4164
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.860124594898662e-05,
+      "loss": 0.3988,
+      "step": 4165
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8600910113680252e-05,
+      "loss": 0.3307,
+      "step": 4166
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8600574278373887e-05,
+      "loss": 0.4629,
+      "step": 4167
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.860023844306752e-05,
+      "loss": 0.337,
+      "step": 4168
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8599902607761156e-05,
+      "loss": 0.3345,
+      "step": 4169
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.859956677245479e-05,
+      "loss": 0.5505,
+      "step": 4170
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8599230937148425e-05,
+      "loss": 0.2547,
+      "step": 4171
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.859889510184206e-05,
+      "loss": 0.3191,
+      "step": 4172
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.859855926653569e-05,
+      "loss": 0.2697,
+      "step": 4173
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.859822343122933e-05,
+      "loss": 0.2697,
+      "step": 4174
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.859788759592296e-05,
+      "loss": 0.2646,
+      "step": 4175
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8597551760616595e-05,
+      "loss": 0.4752,
+      "step": 4176
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.859721592531023e-05,
+      "loss": 0.4764,
+      "step": 4177
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8596880090003864e-05,
+      "loss": 0.5493,
+      "step": 4178
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8596544254697496e-05,
+      "loss": 0.349,
+      "step": 4179
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8596208419391134e-05,
+      "loss": 0.2783,
+      "step": 4180
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8595872584084765e-05,
+      "loss": 0.2454,
+      "step": 4181
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.85955367487784e-05,
+      "loss": 0.3296,
+      "step": 4182
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8595200913472034e-05,
+      "loss": 0.4965,
+      "step": 4183
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.859486507816567e-05,
+      "loss": 0.5626,
+      "step": 4184
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8594529242859303e-05,
+      "loss": 0.2854,
+      "step": 4185
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8594193407552938e-05,
+      "loss": 0.2053,
+      "step": 4186
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8593857572246573e-05,
+      "loss": 0.4484,
+      "step": 4187
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8593521736940204e-05,
+      "loss": 0.5246,
+      "step": 4188
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8593185901633842e-05,
+      "loss": 0.3058,
+      "step": 4189
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8592850066327473e-05,
+      "loss": 0.2545,
+      "step": 4190
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8592514231021108e-05,
+      "loss": 0.373,
+      "step": 4191
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8592178395714743e-05,
+      "loss": 0.3271,
+      "step": 4192
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8591842560408377e-05,
+      "loss": 0.3423,
+      "step": 4193
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8591506725102012e-05,
+      "loss": 0.3923,
+      "step": 4194
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8591170889795647e-05,
+      "loss": 0.308,
+      "step": 4195
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.859083505448928e-05,
+      "loss": 0.4464,
+      "step": 4196
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8590499219182912e-05,
+      "loss": 0.2729,
+      "step": 4197
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.859016338387655e-05,
+      "loss": 0.326,
+      "step": 4198
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8589827548570182e-05,
+      "loss": 0.5279,
+      "step": 4199
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8589491713263816e-05,
+      "loss": 0.2498,
+      "step": 4200
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.858915587795745e-05,
+      "loss": 0.2852,
+      "step": 4201
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8588820042651086e-05,
+      "loss": 0.1833,
+      "step": 4202
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.858848420734472e-05,
+      "loss": 0.2976,
+      "step": 4203
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8588148372038355e-05,
+      "loss": 0.3058,
+      "step": 4204
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.858781253673199e-05,
+      "loss": 0.462,
+      "step": 4205
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.858747670142562e-05,
+      "loss": 0.3611,
+      "step": 4206
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.858714086611926e-05,
+      "loss": 0.3443,
+      "step": 4207
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.858680503081289e-05,
+      "loss": 0.2642,
+      "step": 4208
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8586469195506525e-05,
+      "loss": 0.2717,
+      "step": 4209
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.858613336020016e-05,
+      "loss": 0.3515,
+      "step": 4210
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8585797524893794e-05,
+      "loss": 0.2567,
+      "step": 4211
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.858546168958743e-05,
+      "loss": 0.3907,
+      "step": 4212
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8585125854281063e-05,
+      "loss": 0.1833,
+      "step": 4213
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8584790018974698e-05,
+      "loss": 0.1672,
+      "step": 4214
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.858445418366833e-05,
+      "loss": 0.3051,
+      "step": 4215
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8584118348361967e-05,
+      "loss": 0.3588,
+      "step": 4216
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.85837825130556e-05,
+      "loss": 0.3634,
+      "step": 4217
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8583446677749233e-05,
+      "loss": 0.3765,
+      "step": 4218
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8583110842442868e-05,
+      "loss": 0.3049,
+      "step": 4219
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8582775007136503e-05,
+      "loss": 0.1462,
+      "step": 4220
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8582439171830137e-05,
+      "loss": 0.2377,
+      "step": 4221
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8582103336523772e-05,
+      "loss": 0.32,
+      "step": 4222
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8581767501217403e-05,
+      "loss": 0.3151,
+      "step": 4223
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8581431665911038e-05,
+      "loss": 0.3275,
+      "step": 4224
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8581095830604672e-05,
+      "loss": 0.1249,
+      "step": 4225
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8580759995298307e-05,
+      "loss": 0.3156,
+      "step": 4226
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8580424159991942e-05,
+      "loss": 0.4777,
+      "step": 4227
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8580088324685576e-05,
+      "loss": 0.4138,
+      "step": 4228
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.857975248937921e-05,
+      "loss": 0.3893,
+      "step": 4229
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8579416654072842e-05,
+      "loss": 0.3877,
+      "step": 4230
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.857908081876648e-05,
+      "loss": 0.2071,
+      "step": 4231
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.857874498346011e-05,
+      "loss": 0.3678,
+      "step": 4232
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8578409148153746e-05,
+      "loss": 0.3316,
+      "step": 4233
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.857807331284738e-05,
+      "loss": 0.3281,
+      "step": 4234
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8577737477541016e-05,
+      "loss": 0.1969,
+      "step": 4235
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.857740164223465e-05,
+      "loss": 0.3132,
+      "step": 4236
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8577065806928285e-05,
+      "loss": 0.3037,
+      "step": 4237
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.857672997162192e-05,
+      "loss": 0.4938,
+      "step": 4238
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.857639413631555e-05,
+      "loss": 0.4354,
+      "step": 4239
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.857605830100919e-05,
+      "loss": 0.4327,
+      "step": 4240
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.857572246570282e-05,
+      "loss": 0.3681,
+      "step": 4241
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8575386630396455e-05,
+      "loss": 0.3391,
+      "step": 4242
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.857505079509009e-05,
+      "loss": 0.5028,
+      "step": 4243
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8574714959783724e-05,
+      "loss": 0.3023,
+      "step": 4244
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.857437912447736e-05,
+      "loss": 0.3842,
+      "step": 4245
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.857404328917099e-05,
+      "loss": 0.3745,
+      "step": 4246
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8573707453864628e-05,
+      "loss": 0.4788,
+      "step": 4247
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.857337161855826e-05,
+      "loss": 0.4534,
+      "step": 4248
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8573035783251894e-05,
+      "loss": 0.3812,
+      "step": 4249
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.857269994794553e-05,
+      "loss": 0.5049,
+      "step": 4250
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8572364112639163e-05,
+      "loss": 0.4146,
+      "step": 4251
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8572028277332798e-05,
+      "loss": 0.3791,
+      "step": 4252
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8571692442026432e-05,
+      "loss": 0.3038,
+      "step": 4253
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8571356606720067e-05,
+      "loss": 0.2537,
+      "step": 4254
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.85710207714137e-05,
+      "loss": 0.3839,
+      "step": 4255
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8570684936107336e-05,
+      "loss": 0.3133,
+      "step": 4256
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8570349100800968e-05,
+      "loss": 0.3181,
+      "step": 4257
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8570013265494602e-05,
+      "loss": 0.6156,
+      "step": 4258
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8569677430188237e-05,
+      "loss": 0.3118,
+      "step": 4259
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.856934159488187e-05,
+      "loss": 0.3098,
+      "step": 4260
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8569005759575506e-05,
+      "loss": 0.4142,
+      "step": 4261
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.856866992426914e-05,
+      "loss": 0.3193,
+      "step": 4262
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8568334088962775e-05,
+      "loss": 0.2903,
+      "step": 4263
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8567998253656407e-05,
+      "loss": 0.4532,
+      "step": 4264
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8567662418350045e-05,
+      "loss": 0.3774,
+      "step": 4265
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8567326583043676e-05,
+      "loss": 0.2183,
+      "step": 4266
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.856699074773731e-05,
+      "loss": 0.179,
+      "step": 4267
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8566654912430945e-05,
+      "loss": 0.3147,
+      "step": 4268
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.856631907712458e-05,
+      "loss": 0.1936,
+      "step": 4269
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.856598324181821e-05,
+      "loss": 0.4132,
+      "step": 4270
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.856564740651185e-05,
+      "loss": 0.1522,
+      "step": 4271
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.856531157120548e-05,
+      "loss": 0.398,
+      "step": 4272
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8564975735899115e-05,
+      "loss": 0.3816,
+      "step": 4273
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.856463990059275e-05,
+      "loss": 0.3633,
+      "step": 4274
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8564304065286384e-05,
+      "loss": 0.3839,
+      "step": 4275
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.856396822998002e-05,
+      "loss": 0.3244,
+      "step": 4276
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8563632394673654e-05,
+      "loss": 0.5482,
+      "step": 4277
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.856329655936729e-05,
+      "loss": 0.3821,
+      "step": 4278
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.856296072406092e-05,
+      "loss": 0.4688,
+      "step": 4279
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8562624888754558e-05,
+      "loss": 0.4072,
+      "step": 4280
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.856228905344819e-05,
+      "loss": 0.3202,
+      "step": 4281
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8561953218141824e-05,
+      "loss": 0.3723,
+      "step": 4282
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8561617382835458e-05,
+      "loss": 0.5589,
+      "step": 4283
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8561281547529093e-05,
+      "loss": 0.3762,
+      "step": 4284
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8560945712222728e-05,
+      "loss": 0.1934,
+      "step": 4285
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8560609876916362e-05,
+      "loss": 0.2248,
+      "step": 4286
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8560274041609997e-05,
+      "loss": 0.2207,
+      "step": 4287
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8559938206303628e-05,
+      "loss": 0.2414,
+      "step": 4288
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8559602370997266e-05,
+      "loss": 0.4807,
+      "step": 4289
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8559266535690897e-05,
+      "loss": 0.3524,
+      "step": 4290
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8558930700384532e-05,
+      "loss": 0.3394,
+      "step": 4291
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8558594865078167e-05,
+      "loss": 0.2248,
+      "step": 4292
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.85582590297718e-05,
+      "loss": 0.4117,
+      "step": 4293
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8557923194465436e-05,
+      "loss": 0.4275,
+      "step": 4294
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.855758735915907e-05,
+      "loss": 0.5107,
+      "step": 4295
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8557251523852705e-05,
+      "loss": 0.2041,
+      "step": 4296
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8556915688546337e-05,
+      "loss": 0.4602,
+      "step": 4297
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8556579853239975e-05,
+      "loss": 0.5204,
+      "step": 4298
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8556244017933606e-05,
+      "loss": 0.34,
+      "step": 4299
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.855590818262724e-05,
+      "loss": 0.2491,
+      "step": 4300
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8555572347320875e-05,
+      "loss": 0.34,
+      "step": 4301
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.855523651201451e-05,
+      "loss": 0.3425,
+      "step": 4302
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8554900676708144e-05,
+      "loss": 0.2864,
+      "step": 4303
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.855456484140178e-05,
+      "loss": 0.3793,
+      "step": 4304
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8554229006095414e-05,
+      "loss": 0.4015,
+      "step": 4305
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8553893170789045e-05,
+      "loss": 0.3594,
+      "step": 4306
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8553557335482683e-05,
+      "loss": 0.2623,
+      "step": 4307
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8553221500176314e-05,
+      "loss": 0.3388,
+      "step": 4308
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.855288566486995e-05,
+      "loss": 0.4804,
+      "step": 4309
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8552549829563584e-05,
+      "loss": 0.3751,
+      "step": 4310
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8552213994257218e-05,
+      "loss": 0.4362,
+      "step": 4311
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8551878158950853e-05,
+      "loss": 0.4188,
+      "step": 4312
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8551542323644488e-05,
+      "loss": 0.2728,
+      "step": 4313
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.855120648833812e-05,
+      "loss": 0.3447,
+      "step": 4314
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8550870653031753e-05,
+      "loss": 0.2424,
+      "step": 4315
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8550534817725388e-05,
+      "loss": 0.3638,
+      "step": 4316
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8550198982419023e-05,
+      "loss": 0.4283,
+      "step": 4317
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8549863147112657e-05,
+      "loss": 0.4692,
+      "step": 4318
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8549527311806292e-05,
+      "loss": 0.3997,
+      "step": 4319
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8549191476499927e-05,
+      "loss": 0.343,
+      "step": 4320
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8548855641193558e-05,
+      "loss": 0.2843,
+      "step": 4321
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8548519805887196e-05,
+      "loss": 0.2631,
+      "step": 4322
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8548183970580827e-05,
+      "loss": 0.3364,
+      "step": 4323
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8547848135274462e-05,
+      "loss": 0.3931,
+      "step": 4324
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8547512299968097e-05,
+      "loss": 0.2917,
+      "step": 4325
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.854717646466173e-05,
+      "loss": 0.4274,
+      "step": 4326
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8546840629355366e-05,
+      "loss": 0.4,
+      "step": 4327
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8546504794049e-05,
+      "loss": 0.3443,
+      "step": 4328
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8546168958742635e-05,
+      "loss": 0.4661,
+      "step": 4329
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8545833123436266e-05,
+      "loss": 0.3363,
+      "step": 4330
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8545497288129904e-05,
+      "loss": 0.3505,
+      "step": 4331
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8545161452823536e-05,
+      "loss": 0.4544,
+      "step": 4332
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.854482561751717e-05,
+      "loss": 0.2635,
+      "step": 4333
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8544489782210805e-05,
+      "loss": 0.5083,
+      "step": 4334
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.854415394690444e-05,
+      "loss": 0.4662,
+      "step": 4335
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8543818111598074e-05,
+      "loss": 0.3612,
+      "step": 4336
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.854348227629171e-05,
+      "loss": 0.3016,
+      "step": 4337
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8543146440985344e-05,
+      "loss": 0.3263,
+      "step": 4338
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8542810605678975e-05,
+      "loss": 0.3828,
+      "step": 4339
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8542474770372613e-05,
+      "loss": 0.3885,
+      "step": 4340
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8542138935066244e-05,
+      "loss": 0.3976,
+      "step": 4341
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.854180309975988e-05,
+      "loss": 0.3496,
+      "step": 4342
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8541467264453513e-05,
+      "loss": 0.4046,
+      "step": 4343
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8541131429147148e-05,
+      "loss": 0.3181,
+      "step": 4344
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8540795593840783e-05,
+      "loss": 0.292,
+      "step": 4345
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8540459758534417e-05,
+      "loss": 0.282,
+      "step": 4346
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8540123923228052e-05,
+      "loss": 0.2822,
+      "step": 4347
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8539788087921683e-05,
+      "loss": 0.214,
+      "step": 4348
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.853945225261532e-05,
+      "loss": 0.371,
+      "step": 4349
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8539116417308953e-05,
+      "loss": 0.4164,
+      "step": 4350
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8538780582002587e-05,
+      "loss": 0.3317,
+      "step": 4351
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8538444746696222e-05,
+      "loss": 0.3185,
+      "step": 4352
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8538108911389856e-05,
+      "loss": 0.2808,
+      "step": 4353
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.853777307608349e-05,
+      "loss": 0.3352,
+      "step": 4354
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8537437240777126e-05,
+      "loss": 0.2645,
+      "step": 4355
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.853710140547076e-05,
+      "loss": 0.4576,
+      "step": 4356
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.853676557016439e-05,
+      "loss": 0.4093,
+      "step": 4357
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8536429734858026e-05,
+      "loss": 0.199,
+      "step": 4358
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.853609389955166e-05,
+      "loss": 0.2261,
+      "step": 4359
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8535758064245296e-05,
+      "loss": 0.2213,
+      "step": 4360
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.853542222893893e-05,
+      "loss": 0.1846,
+      "step": 4361
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8535086393632565e-05,
+      "loss": 0.1891,
+      "step": 4362
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8534750558326196e-05,
+      "loss": 0.408,
+      "step": 4363
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8534414723019834e-05,
+      "loss": 0.1465,
+      "step": 4364
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8534078887713465e-05,
+      "loss": 0.347,
+      "step": 4365
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.85337430524071e-05,
+      "loss": 0.2386,
+      "step": 4366
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8533407217100735e-05,
+      "loss": 0.4996,
+      "step": 4367
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.853307138179437e-05,
+      "loss": 0.4773,
+      "step": 4368
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8532735546488004e-05,
+      "loss": 0.3614,
+      "step": 4369
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.853239971118164e-05,
+      "loss": 0.2963,
+      "step": 4370
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8532063875875273e-05,
+      "loss": 0.6387,
+      "step": 4371
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8531728040568905e-05,
+      "loss": 0.4698,
+      "step": 4372
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8531392205262543e-05,
+      "loss": 0.4198,
+      "step": 4373
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8531056369956174e-05,
+      "loss": 0.457,
+      "step": 4374
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.853072053464981e-05,
+      "loss": 0.446,
+      "step": 4375
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8530384699343443e-05,
+      "loss": 0.2803,
+      "step": 4376
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8530048864037078e-05,
+      "loss": 0.3308,
+      "step": 4377
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8529713028730713e-05,
+      "loss": 0.412,
+      "step": 4378
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8529377193424347e-05,
+      "loss": 0.3342,
+      "step": 4379
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8529041358117982e-05,
+      "loss": 0.2652,
+      "step": 4380
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8528705522811613e-05,
+      "loss": 0.3504,
+      "step": 4381
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.852836968750525e-05,
+      "loss": 0.2535,
+      "step": 4382
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8528033852198882e-05,
+      "loss": 0.4492,
+      "step": 4383
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8527698016892517e-05,
+      "loss": 0.2521,
+      "step": 4384
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.852736218158615e-05,
+      "loss": 0.6083,
+      "step": 4385
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8527026346279786e-05,
+      "loss": 0.4254,
+      "step": 4386
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.852669051097342e-05,
+      "loss": 0.3102,
+      "step": 4387
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8526354675667056e-05,
+      "loss": 0.3629,
+      "step": 4388
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.852601884036069e-05,
+      "loss": 0.293,
+      "step": 4389
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.852568300505432e-05,
+      "loss": 0.2297,
+      "step": 4390
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.852534716974796e-05,
+      "loss": 0.3097,
+      "step": 4391
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.852501133444159e-05,
+      "loss": 0.2817,
+      "step": 4392
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8524675499135225e-05,
+      "loss": 0.4514,
+      "step": 4393
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.852433966382886e-05,
+      "loss": 0.2154,
+      "step": 4394
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8524003828522495e-05,
+      "loss": 0.4147,
+      "step": 4395
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.852366799321613e-05,
+      "loss": 0.3864,
+      "step": 4396
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8523332157909764e-05,
+      "loss": 0.3501,
+      "step": 4397
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.85229963226034e-05,
+      "loss": 0.4512,
+      "step": 4398
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.852266048729703e-05,
+      "loss": 0.2962,
+      "step": 4399
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8522324651990668e-05,
+      "loss": 0.4528,
+      "step": 4400
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.85219888166843e-05,
+      "loss": 0.3193,
+      "step": 4401
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8521652981377934e-05,
+      "loss": 0.375,
+      "step": 4402
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.852131714607157e-05,
+      "loss": 0.4447,
+      "step": 4403
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8520981310765203e-05,
+      "loss": 0.3228,
+      "step": 4404
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8520645475458834e-05,
+      "loss": 0.4187,
+      "step": 4405
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8520309640152472e-05,
+      "loss": 0.28,
+      "step": 4406
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8519973804846104e-05,
+      "loss": 0.3293,
+      "step": 4407
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.851963796953974e-05,
+      "loss": 0.316,
+      "step": 4408
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8519302134233373e-05,
+      "loss": 0.2237,
+      "step": 4409
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8518966298927008e-05,
+      "loss": 0.3614,
+      "step": 4410
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8518630463620642e-05,
+      "loss": 0.4782,
+      "step": 4411
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8518294628314277e-05,
+      "loss": 0.5816,
+      "step": 4412
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.851795879300791e-05,
+      "loss": 0.323,
+      "step": 4413
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8517622957701543e-05,
+      "loss": 0.3451,
+      "step": 4414
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.851728712239518e-05,
+      "loss": 0.3037,
+      "step": 4415
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8516951287088812e-05,
+      "loss": 0.2946,
+      "step": 4416
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8516615451782447e-05,
+      "loss": 0.398,
+      "step": 4417
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.851627961647608e-05,
+      "loss": 0.2093,
+      "step": 4418
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8515943781169716e-05,
+      "loss": 0.2432,
+      "step": 4419
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.851560794586335e-05,
+      "loss": 0.482,
+      "step": 4420
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8515272110556985e-05,
+      "loss": 0.2622,
+      "step": 4421
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.851493627525062e-05,
+      "loss": 0.477,
+      "step": 4422
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.851460043994425e-05,
+      "loss": 0.297,
+      "step": 4423
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8514264604637886e-05,
+      "loss": 0.319,
+      "step": 4424
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.851392876933152e-05,
+      "loss": 0.3539,
+      "step": 4425
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8513592934025155e-05,
+      "loss": 0.5408,
+      "step": 4426
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.851325709871879e-05,
+      "loss": 0.369,
+      "step": 4427
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8512921263412425e-05,
+      "loss": 0.2139,
+      "step": 4428
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.851258542810606e-05,
+      "loss": 0.3684,
+      "step": 4429
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.851224959279969e-05,
+      "loss": 0.2665,
+      "step": 4430
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.851191375749333e-05,
+      "loss": 0.3207,
+      "step": 4431
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.851157792218696e-05,
+      "loss": 0.4054,
+      "step": 4432
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8511242086880594e-05,
+      "loss": 0.395,
+      "step": 4433
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.851090625157423e-05,
+      "loss": 0.378,
+      "step": 4434
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8510570416267864e-05,
+      "loss": 0.4327,
+      "step": 4435
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.85102345809615e-05,
+      "loss": 0.3075,
+      "step": 4436
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8509898745655133e-05,
+      "loss": 0.2816,
+      "step": 4437
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8509562910348768e-05,
+      "loss": 0.4279,
+      "step": 4438
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.85092270750424e-05,
+      "loss": 0.3074,
+      "step": 4439
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8508891239736037e-05,
+      "loss": 0.29,
+      "step": 4440
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8508555404429668e-05,
+      "loss": 0.4159,
+      "step": 4441
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8508219569123303e-05,
+      "loss": 0.2421,
+      "step": 4442
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8507883733816937e-05,
+      "loss": 0.2461,
+      "step": 4443
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8507547898510572e-05,
+      "loss": 0.4358,
+      "step": 4444
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8507212063204207e-05,
+      "loss": 0.2266,
+      "step": 4445
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.850687622789784e-05,
+      "loss": 0.3497,
+      "step": 4446
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8506540392591476e-05,
+      "loss": 0.4257,
+      "step": 4447
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8506204557285107e-05,
+      "loss": 0.2094,
+      "step": 4448
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8505868721978742e-05,
+      "loss": 0.3451,
+      "step": 4449
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8505532886672377e-05,
+      "loss": 0.4626,
+      "step": 4450
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.850519705136601e-05,
+      "loss": 0.368,
+      "step": 4451
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8504861216059646e-05,
+      "loss": 0.4708,
+      "step": 4452
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.850452538075328e-05,
+      "loss": 0.3822,
+      "step": 4453
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8504189545446912e-05,
+      "loss": 0.4516,
+      "step": 4454
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.850385371014055e-05,
+      "loss": 0.3813,
+      "step": 4455
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.850351787483418e-05,
+      "loss": 0.2836,
+      "step": 4456
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8503182039527816e-05,
+      "loss": 0.2326,
+      "step": 4457
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.850284620422145e-05,
+      "loss": 0.2774,
+      "step": 4458
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8502510368915085e-05,
+      "loss": 0.3882,
+      "step": 4459
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.850217453360872e-05,
+      "loss": 0.204,
+      "step": 4460
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8501838698302354e-05,
+      "loss": 0.3364,
+      "step": 4461
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.850150286299599e-05,
+      "loss": 0.3618,
+      "step": 4462
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.850116702768962e-05,
+      "loss": 0.45,
+      "step": 4463
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8500831192383258e-05,
+      "loss": 0.306,
+      "step": 4464
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.850049535707689e-05,
+      "loss": 0.6313,
+      "step": 4465
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8500159521770524e-05,
+      "loss": 0.5404,
+      "step": 4466
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.849982368646416e-05,
+      "loss": 0.2144,
+      "step": 4467
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8499487851157794e-05,
+      "loss": 0.4833,
+      "step": 4468
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8499152015851428e-05,
+      "loss": 0.4281,
+      "step": 4469
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8498816180545063e-05,
+      "loss": 0.4988,
+      "step": 4470
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8498480345238697e-05,
+      "loss": 0.3059,
+      "step": 4471
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.849814450993233e-05,
+      "loss": 0.4333,
+      "step": 4472
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8497808674625967e-05,
+      "loss": 0.4481,
+      "step": 4473
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8497472839319598e-05,
+      "loss": 0.4186,
+      "step": 4474
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8497137004013233e-05,
+      "loss": 0.3209,
+      "step": 4475
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8496801168706867e-05,
+      "loss": 0.4845,
+      "step": 4476
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8496465333400502e-05,
+      "loss": 0.3338,
+      "step": 4477
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8496129498094137e-05,
+      "loss": 0.5526,
+      "step": 4478
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.849579366278777e-05,
+      "loss": 0.4298,
+      "step": 4479
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8495457827481406e-05,
+      "loss": 0.3386,
+      "step": 4480
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8495121992175037e-05,
+      "loss": 0.2572,
+      "step": 4481
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8494786156868675e-05,
+      "loss": 0.488,
+      "step": 4482
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8494450321562306e-05,
+      "loss": 0.4273,
+      "step": 4483
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.849411448625594e-05,
+      "loss": 0.4342,
+      "step": 4484
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8493778650949576e-05,
+      "loss": 0.4525,
+      "step": 4485
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.849344281564321e-05,
+      "loss": 0.2672,
+      "step": 4486
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8493106980336845e-05,
+      "loss": 0.4141,
+      "step": 4487
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.849277114503048e-05,
+      "loss": 0.3083,
+      "step": 4488
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8492435309724114e-05,
+      "loss": 0.4478,
+      "step": 4489
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8492099474417746e-05,
+      "loss": 0.5535,
+      "step": 4490
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8491763639111384e-05,
+      "loss": 0.2727,
+      "step": 4491
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8491427803805015e-05,
+      "loss": 0.3561,
+      "step": 4492
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.849109196849865e-05,
+      "loss": 0.2568,
+      "step": 4493
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8490756133192284e-05,
+      "loss": 0.247,
+      "step": 4494
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.849042029788592e-05,
+      "loss": 0.305,
+      "step": 4495
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.849008446257955e-05,
+      "loss": 0.5408,
+      "step": 4496
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8489748627273188e-05,
+      "loss": 0.3903,
+      "step": 4497
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.848941279196682e-05,
+      "loss": 0.3279,
+      "step": 4498
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8489076956660454e-05,
+      "loss": 0.2188,
+      "step": 4499
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.848874112135409e-05,
+      "loss": 0.2258,
+      "step": 4500
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8488405286047723e-05,
+      "loss": 0.3672,
+      "step": 4501
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8488069450741358e-05,
+      "loss": 0.2578,
+      "step": 4502
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8487733615434993e-05,
+      "loss": 0.4308,
+      "step": 4503
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8487397780128627e-05,
+      "loss": 0.323,
+      "step": 4504
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.848706194482226e-05,
+      "loss": 0.3,
+      "step": 4505
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8486726109515897e-05,
+      "loss": 0.2846,
+      "step": 4506
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8486390274209528e-05,
+      "loss": 0.4106,
+      "step": 4507
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8486054438903162e-05,
+      "loss": 0.2186,
+      "step": 4508
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8485718603596797e-05,
+      "loss": 0.4414,
+      "step": 4509
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8485382768290432e-05,
+      "loss": 0.3329,
+      "step": 4510
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8485046932984066e-05,
+      "loss": 0.3656,
+      "step": 4511
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.84847110976777e-05,
+      "loss": 0.3081,
+      "step": 4512
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8484375262371336e-05,
+      "loss": 0.2862,
+      "step": 4513
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8484039427064967e-05,
+      "loss": 0.4197,
+      "step": 4514
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8483703591758605e-05,
+      "loss": 0.3501,
+      "step": 4515
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8483367756452236e-05,
+      "loss": 0.3617,
+      "step": 4516
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.848303192114587e-05,
+      "loss": 0.5134,
+      "step": 4517
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8482696085839506e-05,
+      "loss": 0.3524,
+      "step": 4518
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.848236025053314e-05,
+      "loss": 0.2526,
+      "step": 4519
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8482024415226775e-05,
+      "loss": 0.4694,
+      "step": 4520
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.848168857992041e-05,
+      "loss": 0.3803,
+      "step": 4521
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8481352744614044e-05,
+      "loss": 0.3593,
+      "step": 4522
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8481016909307675e-05,
+      "loss": 0.4842,
+      "step": 4523
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8480681074001313e-05,
+      "loss": 0.3764,
+      "step": 4524
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8480345238694945e-05,
+      "loss": 0.357,
+      "step": 4525
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.848000940338858e-05,
+      "loss": 0.3672,
+      "step": 4526
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8479673568082214e-05,
+      "loss": 0.4604,
+      "step": 4527
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.847933773277585e-05,
+      "loss": 0.1873,
+      "step": 4528
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8479001897469483e-05,
+      "loss": 0.3852,
+      "step": 4529
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8478666062163118e-05,
+      "loss": 0.2778,
+      "step": 4530
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8478330226856753e-05,
+      "loss": 0.4163,
+      "step": 4531
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8477994391550384e-05,
+      "loss": 0.5058,
+      "step": 4532
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8477658556244022e-05,
+      "loss": 0.2261,
+      "step": 4533
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8477322720937653e-05,
+      "loss": 0.2808,
+      "step": 4534
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8476986885631288e-05,
+      "loss": 0.3203,
+      "step": 4535
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8476651050324922e-05,
+      "loss": 0.518,
+      "step": 4536
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8476315215018557e-05,
+      "loss": 0.2217,
+      "step": 4537
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8475979379712192e-05,
+      "loss": 0.6193,
+      "step": 4538
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8475643544405826e-05,
+      "loss": 0.419,
+      "step": 4539
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8475307709099458e-05,
+      "loss": 0.4677,
+      "step": 4540
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8474971873793092e-05,
+      "loss": 0.2895,
+      "step": 4541
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8474636038486727e-05,
+      "loss": 0.2769,
+      "step": 4542
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.847430020318036e-05,
+      "loss": 0.5288,
+      "step": 4543
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8473964367873996e-05,
+      "loss": 0.4319,
+      "step": 4544
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.847362853256763e-05,
+      "loss": 0.1907,
+      "step": 4545
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8473292697261266e-05,
+      "loss": 0.3573,
+      "step": 4546
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8472956861954897e-05,
+      "loss": 0.3305,
+      "step": 4547
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8472621026648535e-05,
+      "loss": 0.2061,
+      "step": 4548
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8472285191342166e-05,
+      "loss": 0.5469,
+      "step": 4549
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.84719493560358e-05,
+      "loss": 0.5909,
+      "step": 4550
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8471613520729435e-05,
+      "loss": 0.3341,
+      "step": 4551
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.847127768542307e-05,
+      "loss": 0.3337,
+      "step": 4552
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8470941850116705e-05,
+      "loss": 0.2452,
+      "step": 4553
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.847060601481034e-05,
+      "loss": 0.4416,
+      "step": 4554
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8470270179503974e-05,
+      "loss": 0.5985,
+      "step": 4555
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8469934344197605e-05,
+      "loss": 0.4815,
+      "step": 4556
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8469598508891243e-05,
+      "loss": 0.4495,
+      "step": 4557
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8469262673584874e-05,
+      "loss": 0.4616,
+      "step": 4558
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.846892683827851e-05,
+      "loss": 0.4183,
+      "step": 4559
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8468591002972144e-05,
+      "loss": 0.4419,
+      "step": 4560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.846825516766578e-05,
+      "loss": 0.4173,
+      "step": 4561
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8467919332359413e-05,
+      "loss": 0.409,
+      "step": 4562
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8467583497053048e-05,
+      "loss": 0.2295,
+      "step": 4563
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8467247661746682e-05,
+      "loss": 0.3897,
+      "step": 4564
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8466911826440314e-05,
+      "loss": 0.2893,
+      "step": 4565
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.846657599113395e-05,
+      "loss": 0.2987,
+      "step": 4566
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8466240155827583e-05,
+      "loss": 0.3347,
+      "step": 4567
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8465904320521218e-05,
+      "loss": 0.4362,
+      "step": 4568
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8465568485214852e-05,
+      "loss": 0.4097,
+      "step": 4569
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8465232649908487e-05,
+      "loss": 0.344,
+      "step": 4570
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.846489681460212e-05,
+      "loss": 0.2467,
+      "step": 4571
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8464560979295756e-05,
+      "loss": 0.3674,
+      "step": 4572
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.846422514398939e-05,
+      "loss": 0.1637,
+      "step": 4573
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8463889308683022e-05,
+      "loss": 0.3252,
+      "step": 4574
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.846355347337666e-05,
+      "loss": 0.3389,
+      "step": 4575
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.846321763807029e-05,
+      "loss": 0.3756,
+      "step": 4576
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8462881802763926e-05,
+      "loss": 0.5689,
+      "step": 4577
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.846254596745756e-05,
+      "loss": 0.167,
+      "step": 4578
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8462210132151195e-05,
+      "loss": 0.4016,
+      "step": 4579
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.846187429684483e-05,
+      "loss": 0.3384,
+      "step": 4580
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8461538461538465e-05,
+      "loss": 0.3165,
+      "step": 4581
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.84612026262321e-05,
+      "loss": 0.1602,
+      "step": 4582
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.846086679092573e-05,
+      "loss": 0.2471,
+      "step": 4583
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8460530955619365e-05,
+      "loss": 0.4095,
+      "step": 4584
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8460195120313e-05,
+      "loss": 0.2094,
+      "step": 4585
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8459859285006634e-05,
+      "loss": 0.277,
+      "step": 4586
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.845952344970027e-05,
+      "loss": 0.2352,
+      "step": 4587
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8459187614393904e-05,
+      "loss": 0.2048,
+      "step": 4588
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8458851779087535e-05,
+      "loss": 0.3609,
+      "step": 4589
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8458515943781173e-05,
+      "loss": 0.1978,
+      "step": 4590
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8458180108474804e-05,
+      "loss": 0.5348,
+      "step": 4591
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.845784427316844e-05,
+      "loss": 0.4287,
+      "step": 4592
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8457508437862074e-05,
+      "loss": 0.3482,
+      "step": 4593
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8457172602555708e-05,
+      "loss": 0.3801,
+      "step": 4594
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8456836767249343e-05,
+      "loss": 0.2186,
+      "step": 4595
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8456500931942978e-05,
+      "loss": 0.3554,
+      "step": 4596
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8456165096636612e-05,
+      "loss": 0.4136,
+      "step": 4597
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8455829261330243e-05,
+      "loss": 0.2401,
+      "step": 4598
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.845549342602388e-05,
+      "loss": 0.4888,
+      "step": 4599
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8455157590717513e-05,
+      "loss": 0.4323,
+      "step": 4600
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8454821755411147e-05,
+      "loss": 0.4427,
+      "step": 4601
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8454485920104782e-05,
+      "loss": 0.3233,
+      "step": 4602
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8454150084798417e-05,
+      "loss": 0.289,
+      "step": 4603
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.845381424949205e-05,
+      "loss": 0.368,
+      "step": 4604
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8453478414185683e-05,
+      "loss": 0.3905,
+      "step": 4605
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.845314257887932e-05,
+      "loss": 0.5332,
+      "step": 4606
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8452806743572952e-05,
+      "loss": 0.2298,
+      "step": 4607
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8452470908266587e-05,
+      "loss": 0.3156,
+      "step": 4608
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.845213507296022e-05,
+      "loss": 0.4741,
+      "step": 4609
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8451799237653856e-05,
+      "loss": 0.3223,
+      "step": 4610
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.845146340234749e-05,
+      "loss": 0.4085,
+      "step": 4611
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8451127567041125e-05,
+      "loss": 0.4389,
+      "step": 4612
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.845079173173476e-05,
+      "loss": 0.2376,
+      "step": 4613
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.845045589642839e-05,
+      "loss": 0.331,
+      "step": 4614
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.845012006112203e-05,
+      "loss": 0.2817,
+      "step": 4615
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.844978422581566e-05,
+      "loss": 0.3301,
+      "step": 4616
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8449448390509295e-05,
+      "loss": 0.3552,
+      "step": 4617
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.844911255520293e-05,
+      "loss": 0.3399,
+      "step": 4618
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8448776719896564e-05,
+      "loss": 0.4379,
+      "step": 4619
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.84484408845902e-05,
+      "loss": 0.5192,
+      "step": 4620
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8448105049283834e-05,
+      "loss": 0.492,
+      "step": 4621
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8447769213977468e-05,
+      "loss": 0.4759,
+      "step": 4622
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.84474333786711e-05,
+      "loss": 0.4234,
+      "step": 4623
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8447097543364738e-05,
+      "loss": 0.3079,
+      "step": 4624
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.844676170805837e-05,
+      "loss": 0.2023,
+      "step": 4625
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8446425872752003e-05,
+      "loss": 0.3415,
+      "step": 4626
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8446090037445638e-05,
+      "loss": 0.2205,
+      "step": 4627
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8445754202139273e-05,
+      "loss": 0.2491,
+      "step": 4628
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8445418366832904e-05,
+      "loss": 0.3742,
+      "step": 4629
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8445082531526542e-05,
+      "loss": 0.3972,
+      "step": 4630
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8444746696220173e-05,
+      "loss": 0.4206,
+      "step": 4631
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8444410860913808e-05,
+      "loss": 0.3975,
+      "step": 4632
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8444075025607443e-05,
+      "loss": 0.3029,
+      "step": 4633
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8443739190301077e-05,
+      "loss": 0.3848,
+      "step": 4634
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8443403354994712e-05,
+      "loss": 0.2517,
+      "step": 4635
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8443067519688347e-05,
+      "loss": 0.2815,
+      "step": 4636
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.844273168438198e-05,
+      "loss": 0.2305,
+      "step": 4637
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8442395849075612e-05,
+      "loss": 0.3797,
+      "step": 4638
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.844206001376925e-05,
+      "loss": 0.3889,
+      "step": 4639
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8441724178462882e-05,
+      "loss": 0.3309,
+      "step": 4640
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8441388343156516e-05,
+      "loss": 0.3635,
+      "step": 4641
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.844105250785015e-05,
+      "loss": 0.312,
+      "step": 4642
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8440716672543786e-05,
+      "loss": 0.3199,
+      "step": 4643
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.844038083723742e-05,
+      "loss": 0.2598,
+      "step": 4644
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8440045001931055e-05,
+      "loss": 0.2597,
+      "step": 4645
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.843970916662469e-05,
+      "loss": 0.3659,
+      "step": 4646
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.843937333131832e-05,
+      "loss": 0.4052,
+      "step": 4647
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.843903749601196e-05,
+      "loss": 0.3128,
+      "step": 4648
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.843870166070559e-05,
+      "loss": 0.3508,
+      "step": 4649
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8438365825399225e-05,
+      "loss": 0.3462,
+      "step": 4650
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.843802999009286e-05,
+      "loss": 0.306,
+      "step": 4651
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8437694154786494e-05,
+      "loss": 0.3382,
+      "step": 4652
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.843735831948013e-05,
+      "loss": 0.2523,
+      "step": 4653
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8437022484173763e-05,
+      "loss": 0.3867,
+      "step": 4654
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8436686648867398e-05,
+      "loss": 0.3242,
+      "step": 4655
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.843635081356103e-05,
+      "loss": 0.5995,
+      "step": 4656
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8436014978254667e-05,
+      "loss": 0.3076,
+      "step": 4657
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.84356791429483e-05,
+      "loss": 0.3398,
+      "step": 4658
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8435343307641933e-05,
+      "loss": 0.3888,
+      "step": 4659
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8435007472335568e-05,
+      "loss": 0.325,
+      "step": 4660
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8434671637029203e-05,
+      "loss": 0.3806,
+      "step": 4661
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8434335801722837e-05,
+      "loss": 0.2153,
+      "step": 4662
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8433999966416472e-05,
+      "loss": 0.1948,
+      "step": 4663
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8433664131110106e-05,
+      "loss": 0.1607,
+      "step": 4664
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8433328295803738e-05,
+      "loss": 0.3201,
+      "step": 4665
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8432992460497376e-05,
+      "loss": 0.3037,
+      "step": 4666
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8432656625191007e-05,
+      "loss": 0.4382,
+      "step": 4667
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.843232078988464e-05,
+      "loss": 0.2935,
+      "step": 4668
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8431984954578276e-05,
+      "loss": 0.391,
+      "step": 4669
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.843164911927191e-05,
+      "loss": 0.2816,
+      "step": 4670
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8431313283965546e-05,
+      "loss": 0.2165,
+      "step": 4671
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.843097744865918e-05,
+      "loss": 0.3859,
+      "step": 4672
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8430641613352815e-05,
+      "loss": 0.5582,
+      "step": 4673
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8430305778046446e-05,
+      "loss": 0.3767,
+      "step": 4674
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.842996994274008e-05,
+      "loss": 0.3221,
+      "step": 4675
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8429634107433715e-05,
+      "loss": 0.3053,
+      "step": 4676
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.842929827212735e-05,
+      "loss": 0.1902,
+      "step": 4677
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8428962436820985e-05,
+      "loss": 0.3305,
+      "step": 4678
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.842862660151462e-05,
+      "loss": 0.3301,
+      "step": 4679
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.842829076620825e-05,
+      "loss": 0.2927,
+      "step": 4680
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.842795493090189e-05,
+      "loss": 0.3658,
+      "step": 4681
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.842761909559552e-05,
+      "loss": 0.481,
+      "step": 4682
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8427283260289155e-05,
+      "loss": 0.2599,
+      "step": 4683
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.842694742498279e-05,
+      "loss": 0.418,
+      "step": 4684
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8426611589676424e-05,
+      "loss": 0.3074,
+      "step": 4685
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.842627575437006e-05,
+      "loss": 0.51,
+      "step": 4686
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8425939919063693e-05,
+      "loss": 0.4498,
+      "step": 4687
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8425604083757328e-05,
+      "loss": 0.3668,
+      "step": 4688
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.842526824845096e-05,
+      "loss": 0.3119,
+      "step": 4689
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8424932413144597e-05,
+      "loss": 0.5871,
+      "step": 4690
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.842459657783823e-05,
+      "loss": 0.2571,
+      "step": 4691
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8424260742531863e-05,
+      "loss": 0.3062,
+      "step": 4692
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8423924907225498e-05,
+      "loss": 0.3762,
+      "step": 4693
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8423589071919132e-05,
+      "loss": 0.3613,
+      "step": 4694
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8423253236612767e-05,
+      "loss": 0.3429,
+      "step": 4695
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.84229174013064e-05,
+      "loss": 0.3475,
+      "step": 4696
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8422581566000036e-05,
+      "loss": 0.2638,
+      "step": 4697
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8422245730693668e-05,
+      "loss": 0.236,
+      "step": 4698
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8421909895387306e-05,
+      "loss": 0.2784,
+      "step": 4699
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8421574060080937e-05,
+      "loss": 0.3938,
+      "step": 4700
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.842123822477457e-05,
+      "loss": 0.4008,
+      "step": 4701
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8420902389468206e-05,
+      "loss": 0.3705,
+      "step": 4702
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.842056655416184e-05,
+      "loss": 0.3908,
+      "step": 4703
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8420230718855475e-05,
+      "loss": 0.4492,
+      "step": 4704
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.841989488354911e-05,
+      "loss": 0.3765,
+      "step": 4705
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8419559048242745e-05,
+      "loss": 0.1795,
+      "step": 4706
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8419223212936376e-05,
+      "loss": 0.2911,
+      "step": 4707
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8418887377630014e-05,
+      "loss": 0.4034,
+      "step": 4708
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8418551542323645e-05,
+      "loss": 0.3678,
+      "step": 4709
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.841821570701728e-05,
+      "loss": 0.2449,
+      "step": 4710
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8417879871710915e-05,
+      "loss": 0.4334,
+      "step": 4711
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.841754403640455e-05,
+      "loss": 0.3272,
+      "step": 4712
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8417208201098184e-05,
+      "loss": 0.4764,
+      "step": 4713
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.841687236579182e-05,
+      "loss": 0.3365,
+      "step": 4714
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8416536530485453e-05,
+      "loss": 0.4638,
+      "step": 4715
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8416200695179084e-05,
+      "loss": 0.4549,
+      "step": 4716
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8415864859872722e-05,
+      "loss": 0.3442,
+      "step": 4717
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8415529024566354e-05,
+      "loss": 0.3054,
+      "step": 4718
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.841519318925999e-05,
+      "loss": 0.2313,
+      "step": 4719
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8414857353953623e-05,
+      "loss": 0.3491,
+      "step": 4720
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8414521518647258e-05,
+      "loss": 0.4778,
+      "step": 4721
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.841418568334089e-05,
+      "loss": 0.3034,
+      "step": 4722
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8413849848034527e-05,
+      "loss": 0.3315,
+      "step": 4723
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8413514012728158e-05,
+      "loss": 0.3698,
+      "step": 4724
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8413178177421793e-05,
+      "loss": 0.2605,
+      "step": 4725
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8412842342115427e-05,
+      "loss": 0.3115,
+      "step": 4726
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8412506506809062e-05,
+      "loss": 0.4568,
+      "step": 4727
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8412170671502697e-05,
+      "loss": 0.2709,
+      "step": 4728
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.841183483619633e-05,
+      "loss": 0.6211,
+      "step": 4729
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8411499000889966e-05,
+      "loss": 0.4071,
+      "step": 4730
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8411163165583597e-05,
+      "loss": 0.2628,
+      "step": 4731
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8410827330277235e-05,
+      "loss": 0.2557,
+      "step": 4732
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8410491494970867e-05,
+      "loss": 0.45,
+      "step": 4733
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.84101556596645e-05,
+      "loss": 0.3442,
+      "step": 4734
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8409819824358136e-05,
+      "loss": 0.3104,
+      "step": 4735
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.840948398905177e-05,
+      "loss": 0.412,
+      "step": 4736
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8409148153745405e-05,
+      "loss": 0.2797,
+      "step": 4737
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.840881231843904e-05,
+      "loss": 0.2324,
+      "step": 4738
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8408476483132675e-05,
+      "loss": 0.2551,
+      "step": 4739
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8408140647826306e-05,
+      "loss": 0.3454,
+      "step": 4740
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8407804812519944e-05,
+      "loss": 0.3149,
+      "step": 4741
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8407468977213575e-05,
+      "loss": 0.1637,
+      "step": 4742
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.840713314190721e-05,
+      "loss": 0.2681,
+      "step": 4743
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8406797306600844e-05,
+      "loss": 0.2601,
+      "step": 4744
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.840646147129448e-05,
+      "loss": 0.409,
+      "step": 4745
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8406125635988114e-05,
+      "loss": 0.3056,
+      "step": 4746
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.840578980068175e-05,
+      "loss": 0.4672,
+      "step": 4747
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8405453965375383e-05,
+      "loss": 0.3792,
+      "step": 4748
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8405118130069014e-05,
+      "loss": 0.2844,
+      "step": 4749
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8404782294762652e-05,
+      "loss": 0.3751,
+      "step": 4750
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8404446459456284e-05,
+      "loss": 0.4171,
+      "step": 4751
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8404110624149918e-05,
+      "loss": 0.6126,
+      "step": 4752
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8403774788843553e-05,
+      "loss": 0.3487,
+      "step": 4753
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8403438953537187e-05,
+      "loss": 0.4291,
+      "step": 4754
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8403103118230822e-05,
+      "loss": 0.4587,
+      "step": 4755
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8402767282924457e-05,
+      "loss": 0.4004,
+      "step": 4756
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.840243144761809e-05,
+      "loss": 0.3095,
+      "step": 4757
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8402095612311723e-05,
+      "loss": 0.3201,
+      "step": 4758
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.840175977700536e-05,
+      "loss": 0.2271,
+      "step": 4759
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8401423941698992e-05,
+      "loss": 0.2385,
+      "step": 4760
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8401088106392627e-05,
+      "loss": 0.325,
+      "step": 4761
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.840075227108626e-05,
+      "loss": 0.3628,
+      "step": 4762
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8400416435779896e-05,
+      "loss": 0.1964,
+      "step": 4763
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8400080600473527e-05,
+      "loss": 0.3749,
+      "step": 4764
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8399744765167165e-05,
+      "loss": 0.5147,
+      "step": 4765
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8399408929860796e-05,
+      "loss": 0.2678,
+      "step": 4766
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.839907309455443e-05,
+      "loss": 0.3753,
+      "step": 4767
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8398737259248066e-05,
+      "loss": 0.4898,
+      "step": 4768
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.83984014239417e-05,
+      "loss": 0.3987,
+      "step": 4769
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8398065588635335e-05,
+      "loss": 0.3019,
+      "step": 4770
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.839772975332897e-05,
+      "loss": 0.3145,
+      "step": 4771
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8397393918022604e-05,
+      "loss": 0.2569,
+      "step": 4772
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8397058082716236e-05,
+      "loss": 0.3246,
+      "step": 4773
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8396722247409874e-05,
+      "loss": 0.2067,
+      "step": 4774
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8396386412103505e-05,
+      "loss": 0.2275,
+      "step": 4775
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.839605057679714e-05,
+      "loss": 0.4156,
+      "step": 4776
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8395714741490774e-05,
+      "loss": 0.4995,
+      "step": 4777
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.839537890618441e-05,
+      "loss": 0.2166,
+      "step": 4778
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8395043070878043e-05,
+      "loss": 0.2482,
+      "step": 4779
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8394707235571678e-05,
+      "loss": 0.3183,
+      "step": 4780
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8394371400265313e-05,
+      "loss": 0.3046,
+      "step": 4781
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8394035564958944e-05,
+      "loss": 0.268,
+      "step": 4782
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.839369972965258e-05,
+      "loss": 0.3938,
+      "step": 4783
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8393363894346213e-05,
+      "loss": 0.3972,
+      "step": 4784
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8393028059039848e-05,
+      "loss": 0.2774,
+      "step": 4785
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8392692223733483e-05,
+      "loss": 0.2124,
+      "step": 4786
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8392356388427117e-05,
+      "loss": 0.2096,
+      "step": 4787
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8392020553120752e-05,
+      "loss": 0.2965,
+      "step": 4788
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8391684717814383e-05,
+      "loss": 0.2627,
+      "step": 4789
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.839134888250802e-05,
+      "loss": 0.3384,
+      "step": 4790
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8391013047201652e-05,
+      "loss": 0.2734,
+      "step": 4791
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8390677211895287e-05,
+      "loss": 0.4825,
+      "step": 4792
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8390341376588922e-05,
+      "loss": 0.3455,
+      "step": 4793
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8390005541282556e-05,
+      "loss": 0.4297,
+      "step": 4794
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.838966970597619e-05,
+      "loss": 0.2851,
+      "step": 4795
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8389333870669826e-05,
+      "loss": 0.3118,
+      "step": 4796
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.838899803536346e-05,
+      "loss": 0.5067,
+      "step": 4797
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.838866220005709e-05,
+      "loss": 0.1496,
+      "step": 4798
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.838832636475073e-05,
+      "loss": 0.3097,
+      "step": 4799
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.838799052944436e-05,
+      "loss": 0.2783,
+      "step": 4800
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8387654694137996e-05,
+      "loss": 0.3081,
+      "step": 4801
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.838731885883163e-05,
+      "loss": 0.2953,
+      "step": 4802
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8386983023525265e-05,
+      "loss": 0.4104,
+      "step": 4803
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.83866471882189e-05,
+      "loss": 0.2813,
+      "step": 4804
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8386311352912534e-05,
+      "loss": 0.2714,
+      "step": 4805
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.838597551760617e-05,
+      "loss": 0.3352,
+      "step": 4806
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.83856396822998e-05,
+      "loss": 0.4651,
+      "step": 4807
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8385303846993438e-05,
+      "loss": 0.2189,
+      "step": 4808
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.838496801168707e-05,
+      "loss": 0.4059,
+      "step": 4809
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8384632176380704e-05,
+      "loss": 0.354,
+      "step": 4810
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.838429634107434e-05,
+      "loss": 0.3864,
+      "step": 4811
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8383960505767973e-05,
+      "loss": 0.3641,
+      "step": 4812
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8383624670461605e-05,
+      "loss": 0.2119,
+      "step": 4813
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8383288835155243e-05,
+      "loss": 0.2996,
+      "step": 4814
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8382952999848874e-05,
+      "loss": 0.2199,
+      "step": 4815
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.838261716454251e-05,
+      "loss": 0.2383,
+      "step": 4816
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8382281329236143e-05,
+      "loss": 0.2054,
+      "step": 4817
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8381945493929778e-05,
+      "loss": 0.3772,
+      "step": 4818
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8381609658623412e-05,
+      "loss": 0.331,
+      "step": 4819
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8381273823317047e-05,
+      "loss": 0.437,
+      "step": 4820
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8380937988010682e-05,
+      "loss": 0.3782,
+      "step": 4821
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8380602152704313e-05,
+      "loss": 0.335,
+      "step": 4822
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.838026631739795e-05,
+      "loss": 0.3561,
+      "step": 4823
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8379930482091582e-05,
+      "loss": 0.3372,
+      "step": 4824
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8379594646785217e-05,
+      "loss": 0.5358,
+      "step": 4825
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.837925881147885e-05,
+      "loss": 0.3785,
+      "step": 4826
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8378922976172486e-05,
+      "loss": 0.3049,
+      "step": 4827
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.837858714086612e-05,
+      "loss": 0.2526,
+      "step": 4828
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8378251305559756e-05,
+      "loss": 0.224,
+      "step": 4829
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.837791547025339e-05,
+      "loss": 0.242,
+      "step": 4830
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.837757963494702e-05,
+      "loss": 0.2264,
+      "step": 4831
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.837724379964066e-05,
+      "loss": 0.227,
+      "step": 4832
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.837690796433429e-05,
+      "loss": 0.2043,
+      "step": 4833
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8376572129027925e-05,
+      "loss": 0.4439,
+      "step": 4834
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.837623629372156e-05,
+      "loss": 0.2449,
+      "step": 4835
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8375900458415195e-05,
+      "loss": 0.4689,
+      "step": 4836
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.837556462310883e-05,
+      "loss": 0.283,
+      "step": 4837
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8375228787802464e-05,
+      "loss": 0.4316,
+      "step": 4838
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.83748929524961e-05,
+      "loss": 0.3522,
+      "step": 4839
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.837455711718973e-05,
+      "loss": 0.2622,
+      "step": 4840
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8374221281883368e-05,
+      "loss": 0.301,
+      "step": 4841
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8373885446577e-05,
+      "loss": 0.392,
+      "step": 4842
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8373549611270634e-05,
+      "loss": 0.4061,
+      "step": 4843
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.837321377596427e-05,
+      "loss": 0.2817,
+      "step": 4844
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8372877940657903e-05,
+      "loss": 0.2327,
+      "step": 4845
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8372542105351538e-05,
+      "loss": 0.2429,
+      "step": 4846
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8372206270045172e-05,
+      "loss": 0.334,
+      "step": 4847
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8371870434738807e-05,
+      "loss": 0.5196,
+      "step": 4848
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8371534599432438e-05,
+      "loss": 0.3545,
+      "step": 4849
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8371198764126076e-05,
+      "loss": 0.3156,
+      "step": 4850
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8370862928819708e-05,
+      "loss": 0.314,
+      "step": 4851
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8370527093513342e-05,
+      "loss": 0.4812,
+      "step": 4852
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8370191258206977e-05,
+      "loss": 0.1912,
+      "step": 4853
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.836985542290061e-05,
+      "loss": 0.3181,
+      "step": 4854
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8369519587594243e-05,
+      "loss": 0.428,
+      "step": 4855
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.836918375228788e-05,
+      "loss": 0.4241,
+      "step": 4856
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8368847916981512e-05,
+      "loss": 0.25,
+      "step": 4857
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8368512081675147e-05,
+      "loss": 0.3539,
+      "step": 4858
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.836817624636878e-05,
+      "loss": 0.5289,
+      "step": 4859
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8367840411062416e-05,
+      "loss": 0.4394,
+      "step": 4860
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.836750457575605e-05,
+      "loss": 0.542,
+      "step": 4861
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8367168740449685e-05,
+      "loss": 0.4024,
+      "step": 4862
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.836683290514332e-05,
+      "loss": 0.3541,
+      "step": 4863
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.836649706983695e-05,
+      "loss": 0.4966,
+      "step": 4864
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.836616123453059e-05,
+      "loss": 0.4392,
+      "step": 4865
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.836582539922422e-05,
+      "loss": 0.3803,
+      "step": 4866
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8365489563917855e-05,
+      "loss": 0.4108,
+      "step": 4867
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.836515372861149e-05,
+      "loss": 0.3457,
+      "step": 4868
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8364817893305124e-05,
+      "loss": 0.225,
+      "step": 4869
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.836448205799876e-05,
+      "loss": 0.4316,
+      "step": 4870
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8364146222692394e-05,
+      "loss": 0.3927,
+      "step": 4871
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.836381038738603e-05,
+      "loss": 0.5536,
+      "step": 4872
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.836347455207966e-05,
+      "loss": 0.4533,
+      "step": 4873
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8363138716773298e-05,
+      "loss": 0.3908,
+      "step": 4874
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.836280288146693e-05,
+      "loss": 0.3184,
+      "step": 4875
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8362467046160564e-05,
+      "loss": 0.2913,
+      "step": 4876
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8362131210854198e-05,
+      "loss": 0.4162,
+      "step": 4877
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8361795375547833e-05,
+      "loss": 0.2865,
+      "step": 4878
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8361459540241468e-05,
+      "loss": 0.3777,
+      "step": 4879
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8361123704935102e-05,
+      "loss": 0.3583,
+      "step": 4880
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8360787869628737e-05,
+      "loss": 0.2269,
+      "step": 4881
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8360452034322368e-05,
+      "loss": 0.2538,
+      "step": 4882
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8360116199016006e-05,
+      "loss": 0.3797,
+      "step": 4883
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8359780363709637e-05,
+      "loss": 0.426,
+      "step": 4884
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8359444528403272e-05,
+      "loss": 0.2905,
+      "step": 4885
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8359108693096907e-05,
+      "loss": 0.424,
+      "step": 4886
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.835877285779054e-05,
+      "loss": 0.3564,
+      "step": 4887
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8358437022484176e-05,
+      "loss": 0.3416,
+      "step": 4888
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.835810118717781e-05,
+      "loss": 0.282,
+      "step": 4889
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8357765351871445e-05,
+      "loss": 0.4837,
+      "step": 4890
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8357429516565077e-05,
+      "loss": 0.3178,
+      "step": 4891
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8357093681258715e-05,
+      "loss": 0.28,
+      "step": 4892
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8356757845952346e-05,
+      "loss": 0.2977,
+      "step": 4893
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.835642201064598e-05,
+      "loss": 0.2873,
+      "step": 4894
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8356086175339615e-05,
+      "loss": 0.3695,
+      "step": 4895
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.835575034003325e-05,
+      "loss": 0.3374,
+      "step": 4896
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8355414504726884e-05,
+      "loss": 0.2952,
+      "step": 4897
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.835507866942052e-05,
+      "loss": 0.2607,
+      "step": 4898
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.835474283411415e-05,
+      "loss": 0.4978,
+      "step": 4899
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8354406998807785e-05,
+      "loss": 0.4499,
+      "step": 4900
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.835407116350142e-05,
+      "loss": 0.3923,
+      "step": 4901
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8353735328195054e-05,
+      "loss": 0.164,
+      "step": 4902
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.835339949288869e-05,
+      "loss": 0.259,
+      "step": 4903
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8353063657582324e-05,
+      "loss": 0.5931,
+      "step": 4904
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8352727822275958e-05,
+      "loss": 0.2184,
+      "step": 4905
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.835239198696959e-05,
+      "loss": 0.2803,
+      "step": 4906
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8352056151663228e-05,
+      "loss": 0.2488,
+      "step": 4907
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.835172031635686e-05,
+      "loss": 0.2267,
+      "step": 4908
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8351384481050493e-05,
+      "loss": 0.2709,
+      "step": 4909
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8351048645744128e-05,
+      "loss": 0.351,
+      "step": 4910
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8350712810437763e-05,
+      "loss": 0.2766,
+      "step": 4911
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8350376975131397e-05,
+      "loss": 0.2868,
+      "step": 4912
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8350041139825032e-05,
+      "loss": 0.2767,
+      "step": 4913
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8349705304518667e-05,
+      "loss": 0.3197,
+      "step": 4914
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8349369469212298e-05,
+      "loss": 0.3773,
+      "step": 4915
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8349033633905936e-05,
+      "loss": 0.2719,
+      "step": 4916
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8348697798599567e-05,
+      "loss": 0.4062,
+      "step": 4917
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8348361963293202e-05,
+      "loss": 0.5518,
+      "step": 4918
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8348026127986837e-05,
+      "loss": 0.5073,
+      "step": 4919
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.834769029268047e-05,
+      "loss": 0.3201,
+      "step": 4920
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8347354457374106e-05,
+      "loss": 0.422,
+      "step": 4921
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.834701862206774e-05,
+      "loss": 0.3991,
+      "step": 4922
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8346682786761375e-05,
+      "loss": 0.3718,
+      "step": 4923
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8346346951455006e-05,
+      "loss": 0.256,
+      "step": 4924
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8346011116148644e-05,
+      "loss": 0.3762,
+      "step": 4925
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8345675280842276e-05,
+      "loss": 0.3796,
+      "step": 4926
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.834533944553591e-05,
+      "loss": 0.2866,
+      "step": 4927
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8345003610229545e-05,
+      "loss": 0.2364,
+      "step": 4928
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.834466777492318e-05,
+      "loss": 0.3581,
+      "step": 4929
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8344331939616814e-05,
+      "loss": 0.3751,
+      "step": 4930
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.834399610431045e-05,
+      "loss": 0.2816,
+      "step": 4931
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8343660269004084e-05,
+      "loss": 0.2747,
+      "step": 4932
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8343324433697715e-05,
+      "loss": 0.2846,
+      "step": 4933
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8342988598391353e-05,
+      "loss": 0.4105,
+      "step": 4934
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8342652763084984e-05,
+      "loss": 0.3172,
+      "step": 4935
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.834231692777862e-05,
+      "loss": 0.2741,
+      "step": 4936
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8341981092472253e-05,
+      "loss": 0.307,
+      "step": 4937
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8341645257165888e-05,
+      "loss": 0.3143,
+      "step": 4938
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8341309421859523e-05,
+      "loss": 0.3528,
+      "step": 4939
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8340973586553157e-05,
+      "loss": 0.4997,
+      "step": 4940
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8340637751246792e-05,
+      "loss": 0.2425,
+      "step": 4941
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8340301915940423e-05,
+      "loss": 0.2483,
+      "step": 4942
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.833996608063406e-05,
+      "loss": 0.4635,
+      "step": 4943
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8339630245327693e-05,
+      "loss": 0.1884,
+      "step": 4944
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8339294410021327e-05,
+      "loss": 0.3285,
+      "step": 4945
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8338958574714962e-05,
+      "loss": 0.3297,
+      "step": 4946
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8338622739408596e-05,
+      "loss": 0.422,
+      "step": 4947
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8338286904102228e-05,
+      "loss": 0.2312,
+      "step": 4948
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8337951068795866e-05,
+      "loss": 0.3164,
+      "step": 4949
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8337615233489497e-05,
+      "loss": 0.3574,
+      "step": 4950
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.833727939818313e-05,
+      "loss": 0.2485,
+      "step": 4951
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8336943562876766e-05,
+      "loss": 0.3324,
+      "step": 4952
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.83366077275704e-05,
+      "loss": 0.2927,
+      "step": 4953
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8336271892264036e-05,
+      "loss": 0.3366,
+      "step": 4954
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.833593605695767e-05,
+      "loss": 0.3994,
+      "step": 4955
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8335600221651305e-05,
+      "loss": 0.399,
+      "step": 4956
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8335264386344936e-05,
+      "loss": 0.4233,
+      "step": 4957
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8334928551038574e-05,
+      "loss": 0.3788,
+      "step": 4958
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8334592715732205e-05,
+      "loss": 0.4062,
+      "step": 4959
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.833425688042584e-05,
+      "loss": 0.24,
+      "step": 4960
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8333921045119475e-05,
+      "loss": 0.4732,
+      "step": 4961
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.833358520981311e-05,
+      "loss": 0.3215,
+      "step": 4962
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8333249374506744e-05,
+      "loss": 0.3623,
+      "step": 4963
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8332913539200375e-05,
+      "loss": 0.3525,
+      "step": 4964
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8332577703894013e-05,
+      "loss": 0.3155,
+      "step": 4965
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8332241868587645e-05,
+      "loss": 0.2273,
+      "step": 4966
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.833190603328128e-05,
+      "loss": 0.3816,
+      "step": 4967
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8331570197974914e-05,
+      "loss": 0.4358,
+      "step": 4968
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.833123436266855e-05,
+      "loss": 0.41,
+      "step": 4969
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8330898527362183e-05,
+      "loss": 0.1654,
+      "step": 4970
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8330562692055818e-05,
+      "loss": 0.3366,
+      "step": 4971
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8330226856749452e-05,
+      "loss": 0.3691,
+      "step": 4972
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8329891021443084e-05,
+      "loss": 0.3692,
+      "step": 4973
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8329555186136722e-05,
+      "loss": 0.3822,
+      "step": 4974
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8329219350830353e-05,
+      "loss": 0.4253,
+      "step": 4975
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8328883515523988e-05,
+      "loss": 0.5311,
+      "step": 4976
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8328547680217622e-05,
+      "loss": 0.3144,
+      "step": 4977
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8328211844911257e-05,
+      "loss": 0.3494,
+      "step": 4978
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.832787600960489e-05,
+      "loss": 0.3589,
+      "step": 4979
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8327540174298526e-05,
+      "loss": 0.3058,
+      "step": 4980
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.832720433899216e-05,
+      "loss": 0.4128,
+      "step": 4981
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8326868503685792e-05,
+      "loss": 0.4158,
+      "step": 4982
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.832653266837943e-05,
+      "loss": 0.2483,
+      "step": 4983
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.832619683307306e-05,
+      "loss": 0.3159,
+      "step": 4984
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8325860997766696e-05,
+      "loss": 0.3829,
+      "step": 4985
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.832552516246033e-05,
+      "loss": 0.7633,
+      "step": 4986
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8325189327153965e-05,
+      "loss": 0.4528,
+      "step": 4987
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.83248534918476e-05,
+      "loss": 0.3791,
+      "step": 4988
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8324517656541235e-05,
+      "loss": 0.3505,
+      "step": 4989
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8324181821234866e-05,
+      "loss": 0.3664,
+      "step": 4990
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.83238459859285e-05,
+      "loss": 0.373,
+      "step": 4991
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8323510150622135e-05,
+      "loss": 0.3508,
+      "step": 4992
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.832317431531577e-05,
+      "loss": 0.2993,
+      "step": 4993
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8322838480009405e-05,
+      "loss": 0.4002,
+      "step": 4994
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.832250264470304e-05,
+      "loss": 0.3064,
+      "step": 4995
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8322166809396674e-05,
+      "loss": 0.2744,
+      "step": 4996
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8321830974090305e-05,
+      "loss": 0.5831,
+      "step": 4997
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8321495138783943e-05,
+      "loss": 0.2665,
+      "step": 4998
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8321159303477574e-05,
+      "loss": 0.5447,
+      "step": 4999
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.832082346817121e-05,
+      "loss": 0.4355,
+      "step": 5000
+    },
+    {
+      "epoch": 0.25,
+      "eval_accuracy": 0.873008176668944,
+      "eval_f1_microaverage": 0.873008176668944,
+      "eval_loss": 0.33953726291656494,
+      "eval_precision_macroaverage": 0.701831202619592,
+      "eval_precision_microaverage": 0.873008176668944,
+      "eval_recall_macroaverage": 0.583333570124535,
+      "eval_recall_microaverage": 0.873008176668944,
+      "eval_runtime": 1173.8976,
+      "eval_samples_per_second": 231.909,
+      "eval_steps_per_second": 3.624,
+      "step": 5000
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8320487632864844e-05,
+      "loss": 0.3621,
+      "step": 5001
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.832015179755848e-05,
+      "loss": 0.2417,
+      "step": 5002
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8319815962252113e-05,
+      "loss": 0.3156,
+      "step": 5003
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8319480126945748e-05,
+      "loss": 0.3752,
+      "step": 5004
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8319144291639382e-05,
+      "loss": 0.2984,
+      "step": 5005
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8318808456333014e-05,
+      "loss": 0.2947,
+      "step": 5006
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.831847262102665e-05,
+      "loss": 0.4906,
+      "step": 5007
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8318136785720283e-05,
+      "loss": 0.2897,
+      "step": 5008
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8317800950413918e-05,
+      "loss": 0.224,
+      "step": 5009
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8317465115107552e-05,
+      "loss": 0.3126,
+      "step": 5010
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8317129279801187e-05,
+      "loss": 0.2474,
+      "step": 5011
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.831679344449482e-05,
+      "loss": 0.3082,
+      "step": 5012
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8316457609188456e-05,
+      "loss": 0.2902,
+      "step": 5013
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.831612177388209e-05,
+      "loss": 0.4074,
+      "step": 5014
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8315785938575722e-05,
+      "loss": 0.2838,
+      "step": 5015
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.831545010326936e-05,
+      "loss": 0.3274,
+      "step": 5016
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.831511426796299e-05,
+      "loss": 0.5213,
+      "step": 5017
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8314778432656626e-05,
+      "loss": 0.3411,
+      "step": 5018
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.831444259735026e-05,
+      "loss": 0.3095,
+      "step": 5019
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8314106762043895e-05,
+      "loss": 0.3117,
+      "step": 5020
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.831377092673753e-05,
+      "loss": 0.2038,
+      "step": 5021
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8313435091431165e-05,
+      "loss": 0.2536,
+      "step": 5022
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.83130992561248e-05,
+      "loss": 0.284,
+      "step": 5023
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.831276342081843e-05,
+      "loss": 0.2629,
+      "step": 5024
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.831242758551207e-05,
+      "loss": 0.3269,
+      "step": 5025
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.83120917502057e-05,
+      "loss": 0.2921,
+      "step": 5026
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8311755914899334e-05,
+      "loss": 0.1509,
+      "step": 5027
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.831142007959297e-05,
+      "loss": 0.32,
+      "step": 5028
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8311084244286604e-05,
+      "loss": 0.5746,
+      "step": 5029
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.831074840898024e-05,
+      "loss": 0.3212,
+      "step": 5030
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8310412573673873e-05,
+      "loss": 0.2697,
+      "step": 5031
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8310076738367508e-05,
+      "loss": 0.3604,
+      "step": 5032
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.830974090306114e-05,
+      "loss": 0.4738,
+      "step": 5033
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8309405067754774e-05,
+      "loss": 0.3335,
+      "step": 5034
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8309069232448408e-05,
+      "loss": 0.404,
+      "step": 5035
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8308733397142043e-05,
+      "loss": 0.2826,
+      "step": 5036
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8308397561835677e-05,
+      "loss": 0.4913,
+      "step": 5037
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8308061726529312e-05,
+      "loss": 0.4134,
+      "step": 5038
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8307725891222943e-05,
+      "loss": 0.3998,
+      "step": 5039
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.830739005591658e-05,
+      "loss": 0.314,
+      "step": 5040
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8307054220610213e-05,
+      "loss": 0.4313,
+      "step": 5041
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8306718385303847e-05,
+      "loss": 0.2927,
+      "step": 5042
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8306382549997482e-05,
+      "loss": 0.3799,
+      "step": 5043
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8306046714691117e-05,
+      "loss": 0.3006,
+      "step": 5044
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.830571087938475e-05,
+      "loss": 0.2741,
+      "step": 5045
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8305375044078386e-05,
+      "loss": 0.2843,
+      "step": 5046
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.830503920877202e-05,
+      "loss": 0.3516,
+      "step": 5047
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8304703373465652e-05,
+      "loss": 0.4048,
+      "step": 5048
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.830436753815929e-05,
+      "loss": 0.282,
+      "step": 5049
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.830403170285292e-05,
+      "loss": 0.2631,
+      "step": 5050
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8303695867546556e-05,
+      "loss": 0.3634,
+      "step": 5051
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.830336003224019e-05,
+      "loss": 0.4174,
+      "step": 5052
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8303024196933825e-05,
+      "loss": 0.4239,
+      "step": 5053
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.830268836162746e-05,
+      "loss": 0.353,
+      "step": 5054
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8302352526321094e-05,
+      "loss": 0.314,
+      "step": 5055
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.830201669101473e-05,
+      "loss": 0.4094,
+      "step": 5056
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.830168085570836e-05,
+      "loss": 0.3047,
+      "step": 5057
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8301345020401998e-05,
+      "loss": 0.376,
+      "step": 5058
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.830100918509563e-05,
+      "loss": 0.3934,
+      "step": 5059
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8300673349789264e-05,
+      "loss": 0.2888,
+      "step": 5060
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.83003375144829e-05,
+      "loss": 0.331,
+      "step": 5061
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8300001679176533e-05,
+      "loss": 0.3821,
+      "step": 5062
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8299665843870168e-05,
+      "loss": 0.3765,
+      "step": 5063
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8299330008563803e-05,
+      "loss": 0.2141,
+      "step": 5064
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8298994173257437e-05,
+      "loss": 0.4731,
+      "step": 5065
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.829865833795107e-05,
+      "loss": 0.3592,
+      "step": 5066
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8298322502644707e-05,
+      "loss": 0.2427,
+      "step": 5067
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8297986667338338e-05,
+      "loss": 0.4398,
+      "step": 5068
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8297650832031973e-05,
+      "loss": 0.2203,
+      "step": 5069
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8297314996725607e-05,
+      "loss": 0.4896,
+      "step": 5070
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8296979161419242e-05,
+      "loss": 0.4306,
+      "step": 5071
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8296643326112877e-05,
+      "loss": 0.2973,
+      "step": 5072
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.829630749080651e-05,
+      "loss": 0.3672,
+      "step": 5073
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8295971655500146e-05,
+      "loss": 0.4743,
+      "step": 5074
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8295635820193777e-05,
+      "loss": 0.2642,
+      "step": 5075
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8295299984887415e-05,
+      "loss": 0.2961,
+      "step": 5076
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8294964149581046e-05,
+      "loss": 0.242,
+      "step": 5077
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.829462831427468e-05,
+      "loss": 0.4071,
+      "step": 5078
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8294292478968316e-05,
+      "loss": 0.4463,
+      "step": 5079
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.829395664366195e-05,
+      "loss": 0.3517,
+      "step": 5080
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.829362080835558e-05,
+      "loss": 0.2219,
+      "step": 5081
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.829328497304922e-05,
+      "loss": 0.3185,
+      "step": 5082
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.829294913774285e-05,
+      "loss": 0.2449,
+      "step": 5083
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8292613302436486e-05,
+      "loss": 0.2566,
+      "step": 5084
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.829227746713012e-05,
+      "loss": 0.2943,
+      "step": 5085
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8291941631823755e-05,
+      "loss": 0.3621,
+      "step": 5086
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.829160579651739e-05,
+      "loss": 0.4497,
+      "step": 5087
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8291269961211024e-05,
+      "loss": 0.4969,
+      "step": 5088
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.829093412590466e-05,
+      "loss": 0.235,
+      "step": 5089
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.829059829059829e-05,
+      "loss": 0.2434,
+      "step": 5090
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8290262455291928e-05,
+      "loss": 0.4343,
+      "step": 5091
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.828992661998556e-05,
+      "loss": 0.2628,
+      "step": 5092
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8289590784679194e-05,
+      "loss": 0.2145,
+      "step": 5093
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.828925494937283e-05,
+      "loss": 0.6217,
+      "step": 5094
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8288919114066463e-05,
+      "loss": 0.4449,
+      "step": 5095
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8288583278760098e-05,
+      "loss": 0.2917,
+      "step": 5096
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8288247443453733e-05,
+      "loss": 0.2524,
+      "step": 5097
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8287911608147367e-05,
+      "loss": 0.413,
+      "step": 5098
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8287575772841e-05,
+      "loss": 0.3596,
+      "step": 5099
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8287239937534637e-05,
+      "loss": 0.2582,
+      "step": 5100
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8286904102228268e-05,
+      "loss": 0.3501,
+      "step": 5101
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8286568266921902e-05,
+      "loss": 0.2954,
+      "step": 5102
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8286232431615537e-05,
+      "loss": 0.1922,
+      "step": 5103
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8285896596309172e-05,
+      "loss": 0.4466,
+      "step": 5104
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8285560761002806e-05,
+      "loss": 0.4791,
+      "step": 5105
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.828522492569644e-05,
+      "loss": 0.2763,
+      "step": 5106
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8284889090390076e-05,
+      "loss": 0.2646,
+      "step": 5107
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8284553255083707e-05,
+      "loss": 0.4111,
+      "step": 5108
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8284217419777345e-05,
+      "loss": 0.5451,
+      "step": 5109
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8283881584470976e-05,
+      "loss": 0.4621,
+      "step": 5110
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.828354574916461e-05,
+      "loss": 0.4032,
+      "step": 5111
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8283209913858246e-05,
+      "loss": 0.449,
+      "step": 5112
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.828287407855188e-05,
+      "loss": 0.4013,
+      "step": 5113
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8282538243245515e-05,
+      "loss": 0.26,
+      "step": 5114
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.828220240793915e-05,
+      "loss": 0.4126,
+      "step": 5115
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8281866572632784e-05,
+      "loss": 0.2148,
+      "step": 5116
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8281530737326415e-05,
+      "loss": 0.2673,
+      "step": 5117
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8281194902020053e-05,
+      "loss": 0.2139,
+      "step": 5118
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8280859066713685e-05,
+      "loss": 0.3936,
+      "step": 5119
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.828052323140732e-05,
+      "loss": 0.3309,
+      "step": 5120
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8280187396100954e-05,
+      "loss": 0.2354,
+      "step": 5121
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.827985156079459e-05,
+      "loss": 0.5205,
+      "step": 5122
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8279515725488223e-05,
+      "loss": 0.3683,
+      "step": 5123
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8279179890181858e-05,
+      "loss": 0.2357,
+      "step": 5124
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.827884405487549e-05,
+      "loss": 0.2587,
+      "step": 5125
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8278508219569124e-05,
+      "loss": 0.3387,
+      "step": 5126
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.827817238426276e-05,
+      "loss": 0.2057,
+      "step": 5127
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8277836548956393e-05,
+      "loss": 0.2566,
+      "step": 5128
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8277500713650028e-05,
+      "loss": 0.3693,
+      "step": 5129
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8277164878343662e-05,
+      "loss": 0.2953,
+      "step": 5130
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8276829043037297e-05,
+      "loss": 0.3201,
+      "step": 5131
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.827649320773093e-05,
+      "loss": 0.1752,
+      "step": 5132
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8276157372424566e-05,
+      "loss": 0.2623,
+      "step": 5133
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8275821537118198e-05,
+      "loss": 0.3366,
+      "step": 5134
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8275485701811832e-05,
+      "loss": 0.3212,
+      "step": 5135
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8275149866505467e-05,
+      "loss": 0.4569,
+      "step": 5136
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.82748140311991e-05,
+      "loss": 0.3464,
+      "step": 5137
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8274478195892736e-05,
+      "loss": 0.453,
+      "step": 5138
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.827414236058637e-05,
+      "loss": 0.1864,
+      "step": 5139
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8273806525280005e-05,
+      "loss": 0.3886,
+      "step": 5140
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8273470689973637e-05,
+      "loss": 0.3151,
+      "step": 5141
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8273134854667275e-05,
+      "loss": 0.1836,
+      "step": 5142
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8272799019360906e-05,
+      "loss": 0.3004,
+      "step": 5143
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.827246318405454e-05,
+      "loss": 0.2028,
+      "step": 5144
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8272127348748175e-05,
+      "loss": 0.3182,
+      "step": 5145
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.827179151344181e-05,
+      "loss": 0.342,
+      "step": 5146
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8271455678135445e-05,
+      "loss": 0.3673,
+      "step": 5147
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8271119842829076e-05,
+      "loss": 0.4819,
+      "step": 5148
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8270784007522714e-05,
+      "loss": 0.3887,
+      "step": 5149
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8270448172216345e-05,
+      "loss": 0.3423,
+      "step": 5150
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.827011233690998e-05,
+      "loss": 0.3931,
+      "step": 5151
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8269776501603614e-05,
+      "loss": 0.2788,
+      "step": 5152
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.826944066629725e-05,
+      "loss": 0.3557,
+      "step": 5153
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8269104830990884e-05,
+      "loss": 0.3744,
+      "step": 5154
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.826876899568452e-05,
+      "loss": 0.2426,
+      "step": 5155
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8268433160378153e-05,
+      "loss": 0.4906,
+      "step": 5156
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8268097325071784e-05,
+      "loss": 0.3629,
+      "step": 5157
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8267761489765422e-05,
+      "loss": 0.3652,
+      "step": 5158
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8267425654459054e-05,
+      "loss": 0.4097,
+      "step": 5159
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8267089819152688e-05,
+      "loss": 0.2418,
+      "step": 5160
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8266753983846323e-05,
+      "loss": 0.333,
+      "step": 5161
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8266418148539958e-05,
+      "loss": 0.4235,
+      "step": 5162
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8266082313233592e-05,
+      "loss": 0.3123,
+      "step": 5163
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8265746477927227e-05,
+      "loss": 0.1819,
+      "step": 5164
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.826541064262086e-05,
+      "loss": 0.2403,
+      "step": 5165
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8265074807314493e-05,
+      "loss": 0.246,
+      "step": 5166
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.826473897200813e-05,
+      "loss": 0.3473,
+      "step": 5167
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8264403136701762e-05,
+      "loss": 0.3246,
+      "step": 5168
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8264067301395397e-05,
+      "loss": 0.3428,
+      "step": 5169
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.826373146608903e-05,
+      "loss": 0.303,
+      "step": 5170
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8263395630782666e-05,
+      "loss": 0.2945,
+      "step": 5171
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8263059795476297e-05,
+      "loss": 0.3038,
+      "step": 5172
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8262723960169935e-05,
+      "loss": 0.4127,
+      "step": 5173
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8262388124863567e-05,
+      "loss": 0.2959,
+      "step": 5174
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.82620522895572e-05,
+      "loss": 0.2677,
+      "step": 5175
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8261716454250836e-05,
+      "loss": 0.2908,
+      "step": 5176
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.826138061894447e-05,
+      "loss": 0.44,
+      "step": 5177
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8261044783638105e-05,
+      "loss": 0.2295,
+      "step": 5178
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.826070894833174e-05,
+      "loss": 0.3744,
+      "step": 5179
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8260373113025374e-05,
+      "loss": 0.3025,
+      "step": 5180
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8260037277719006e-05,
+      "loss": 0.3593,
+      "step": 5181
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8259701442412644e-05,
+      "loss": 0.22,
+      "step": 5182
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8259365607106275e-05,
+      "loss": 0.3577,
+      "step": 5183
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.825902977179991e-05,
+      "loss": 0.3514,
+      "step": 5184
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8258693936493544e-05,
+      "loss": 0.3396,
+      "step": 5185
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.825835810118718e-05,
+      "loss": 0.3597,
+      "step": 5186
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8258022265880814e-05,
+      "loss": 0.4116,
+      "step": 5187
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8257686430574448e-05,
+      "loss": 0.2565,
+      "step": 5188
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8257350595268083e-05,
+      "loss": 0.3614,
+      "step": 5189
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8257014759961714e-05,
+      "loss": 0.2749,
+      "step": 5190
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8256678924655352e-05,
+      "loss": 0.4331,
+      "step": 5191
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8256343089348983e-05,
+      "loss": 0.2967,
+      "step": 5192
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8256007254042618e-05,
+      "loss": 0.2988,
+      "step": 5193
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8255671418736253e-05,
+      "loss": 0.4545,
+      "step": 5194
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8255335583429887e-05,
+      "loss": 0.357,
+      "step": 5195
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8254999748123522e-05,
+      "loss": 0.34,
+      "step": 5196
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8254663912817157e-05,
+      "loss": 0.4505,
+      "step": 5197
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.825432807751079e-05,
+      "loss": 0.3596,
+      "step": 5198
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8253992242204423e-05,
+      "loss": 0.5599,
+      "step": 5199
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.825365640689806e-05,
+      "loss": 0.2958,
+      "step": 5200
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8253320571591692e-05,
+      "loss": 0.8629,
+      "step": 5201
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8252984736285327e-05,
+      "loss": 0.348,
+      "step": 5202
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.825264890097896e-05,
+      "loss": 0.3443,
+      "step": 5203
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8252313065672596e-05,
+      "loss": 0.3709,
+      "step": 5204
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.825197723036623e-05,
+      "loss": 0.2388,
+      "step": 5205
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8251641395059865e-05,
+      "loss": 0.5106,
+      "step": 5206
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.82513055597535e-05,
+      "loss": 0.1872,
+      "step": 5207
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.825096972444713e-05,
+      "loss": 0.3554,
+      "step": 5208
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.825063388914077e-05,
+      "loss": 0.3355,
+      "step": 5209
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.82502980538344e-05,
+      "loss": 0.2213,
+      "step": 5210
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8249962218528035e-05,
+      "loss": 0.386,
+      "step": 5211
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.824962638322167e-05,
+      "loss": 0.3734,
+      "step": 5212
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8249290547915304e-05,
+      "loss": 0.2097,
+      "step": 5213
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.824895471260894e-05,
+      "loss": 0.2513,
+      "step": 5214
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8248618877302574e-05,
+      "loss": 0.4528,
+      "step": 5215
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8248283041996205e-05,
+      "loss": 0.4088,
+      "step": 5216
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.824794720668984e-05,
+      "loss": 0.3955,
+      "step": 5217
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8247611371383474e-05,
+      "loss": 0.4733,
+      "step": 5218
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.824727553607711e-05,
+      "loss": 0.5382,
+      "step": 5219
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8246939700770743e-05,
+      "loss": 0.5237,
+      "step": 5220
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8246603865464378e-05,
+      "loss": 0.3166,
+      "step": 5221
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8246268030158013e-05,
+      "loss": 0.1686,
+      "step": 5222
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8245932194851644e-05,
+      "loss": 0.3464,
+      "step": 5223
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8245596359545282e-05,
+      "loss": 0.3555,
+      "step": 5224
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8245260524238913e-05,
+      "loss": 0.3935,
+      "step": 5225
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8244924688932548e-05,
+      "loss": 0.4504,
+      "step": 5226
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8244588853626183e-05,
+      "loss": 0.3103,
+      "step": 5227
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8244253018319817e-05,
+      "loss": 0.2823,
+      "step": 5228
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8243917183013452e-05,
+      "loss": 0.3171,
+      "step": 5229
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8243581347707086e-05,
+      "loss": 0.3134,
+      "step": 5230
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.824324551240072e-05,
+      "loss": 0.3831,
+      "step": 5231
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8242909677094352e-05,
+      "loss": 0.3564,
+      "step": 5232
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.824257384178799e-05,
+      "loss": 0.3111,
+      "step": 5233
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.824223800648162e-05,
+      "loss": 0.3279,
+      "step": 5234
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8241902171175256e-05,
+      "loss": 0.2129,
+      "step": 5235
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.824156633586889e-05,
+      "loss": 0.3066,
+      "step": 5236
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8241230500562526e-05,
+      "loss": 0.2958,
+      "step": 5237
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.824089466525616e-05,
+      "loss": 0.1981,
+      "step": 5238
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8240558829949795e-05,
+      "loss": 0.4836,
+      "step": 5239
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.824022299464343e-05,
+      "loss": 0.2236,
+      "step": 5240
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.823988715933706e-05,
+      "loss": 0.4096,
+      "step": 5241
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.82395513240307e-05,
+      "loss": 0.3595,
+      "step": 5242
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.823921548872433e-05,
+      "loss": 0.4416,
+      "step": 5243
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8238879653417965e-05,
+      "loss": 0.411,
+      "step": 5244
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.82385438181116e-05,
+      "loss": 0.4015,
+      "step": 5245
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8238207982805234e-05,
+      "loss": 0.4263,
+      "step": 5246
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.823787214749887e-05,
+      "loss": 0.3757,
+      "step": 5247
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8237536312192503e-05,
+      "loss": 0.4881,
+      "step": 5248
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8237200476886138e-05,
+      "loss": 0.2946,
+      "step": 5249
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.823686464157977e-05,
+      "loss": 0.3813,
+      "step": 5250
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8236528806273407e-05,
+      "loss": 0.2911,
+      "step": 5251
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.823619297096704e-05,
+      "loss": 0.3649,
+      "step": 5252
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8235857135660673e-05,
+      "loss": 0.428,
+      "step": 5253
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8235521300354308e-05,
+      "loss": 0.3394,
+      "step": 5254
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8235185465047943e-05,
+      "loss": 0.3458,
+      "step": 5255
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8234849629741577e-05,
+      "loss": 0.4684,
+      "step": 5256
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8234513794435212e-05,
+      "loss": 0.3318,
+      "step": 5257
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8234177959128846e-05,
+      "loss": 0.2546,
+      "step": 5258
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8233842123822478e-05,
+      "loss": 0.29,
+      "step": 5259
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.8233506288516112e-05,
+      "loss": 0.3881,
+      "step": 5260
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8233170453209747e-05,
+      "loss": 0.2543,
+      "step": 5261
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.823283461790338e-05,
+      "loss": 0.6107,
+      "step": 5262
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8232498782597016e-05,
+      "loss": 0.2313,
+      "step": 5263
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.823216294729065e-05,
+      "loss": 0.2445,
+      "step": 5264
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8231827111984282e-05,
+      "loss": 0.3158,
+      "step": 5265
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.823149127667792e-05,
+      "loss": 0.316,
+      "step": 5266
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.823115544137155e-05,
+      "loss": 0.2646,
+      "step": 5267
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8230819606065186e-05,
+      "loss": 0.4956,
+      "step": 5268
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.823048377075882e-05,
+      "loss": 0.5229,
+      "step": 5269
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8230147935452455e-05,
+      "loss": 0.4174,
+      "step": 5270
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.822981210014609e-05,
+      "loss": 0.2131,
+      "step": 5271
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8229476264839725e-05,
+      "loss": 0.245,
+      "step": 5272
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.822914042953336e-05,
+      "loss": 0.2886,
+      "step": 5273
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.822880459422699e-05,
+      "loss": 0.3219,
+      "step": 5274
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.822846875892063e-05,
+      "loss": 0.2731,
+      "step": 5275
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.822813292361426e-05,
+      "loss": 0.2738,
+      "step": 5276
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8227797088307895e-05,
+      "loss": 0.3065,
+      "step": 5277
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.822746125300153e-05,
+      "loss": 0.2951,
+      "step": 5278
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8227125417695164e-05,
+      "loss": 0.3421,
+      "step": 5279
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.82267895823888e-05,
+      "loss": 0.4241,
+      "step": 5280
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8226453747082433e-05,
+      "loss": 0.3568,
+      "step": 5281
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8226117911776068e-05,
+      "loss": 0.5477,
+      "step": 5282
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.82257820764697e-05,
+      "loss": 0.3375,
+      "step": 5283
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8225446241163337e-05,
+      "loss": 0.2824,
+      "step": 5284
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.822511040585697e-05,
+      "loss": 0.3577,
+      "step": 5285
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8224774570550603e-05,
+      "loss": 0.2949,
+      "step": 5286
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8224438735244238e-05,
+      "loss": 0.4061,
+      "step": 5287
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8224102899937872e-05,
+      "loss": 0.561,
+      "step": 5288
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8223767064631507e-05,
+      "loss": 0.3901,
+      "step": 5289
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.822343122932514e-05,
+      "loss": 0.2304,
+      "step": 5290
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8223095394018776e-05,
+      "loss": 0.3299,
+      "step": 5291
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8222759558712408e-05,
+      "loss": 0.3016,
+      "step": 5292
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8222423723406046e-05,
+      "loss": 0.3734,
+      "step": 5293
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8222087888099677e-05,
+      "loss": 0.2984,
+      "step": 5294
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.822175205279331e-05,
+      "loss": 0.3925,
+      "step": 5295
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8221416217486946e-05,
+      "loss": 0.5279,
+      "step": 5296
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.822108038218058e-05,
+      "loss": 0.3561,
+      "step": 5297
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8220744546874215e-05,
+      "loss": 0.2759,
+      "step": 5298
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.822040871156785e-05,
+      "loss": 0.3189,
+      "step": 5299
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8220072876261485e-05,
+      "loss": 0.4693,
+      "step": 5300
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8219737040955116e-05,
+      "loss": 0.2669,
+      "step": 5301
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8219401205648754e-05,
+      "loss": 0.3602,
+      "step": 5302
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8219065370342385e-05,
+      "loss": 0.2155,
+      "step": 5303
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.821872953503602e-05,
+      "loss": 0.2807,
+      "step": 5304
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8218393699729655e-05,
+      "loss": 0.3178,
+      "step": 5305
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.821805786442329e-05,
+      "loss": 0.3197,
+      "step": 5306
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.821772202911692e-05,
+      "loss": 0.3052,
+      "step": 5307
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.821738619381056e-05,
+      "loss": 0.4818,
+      "step": 5308
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.821705035850419e-05,
+      "loss": 0.4503,
+      "step": 5309
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8216714523197824e-05,
+      "loss": 0.3055,
+      "step": 5310
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.821637868789146e-05,
+      "loss": 0.1883,
+      "step": 5311
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8216042852585094e-05,
+      "loss": 0.55,
+      "step": 5312
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.821570701727873e-05,
+      "loss": 0.3916,
+      "step": 5313
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8215371181972363e-05,
+      "loss": 0.3636,
+      "step": 5314
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8215035346665998e-05,
+      "loss": 0.2268,
+      "step": 5315
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.821469951135963e-05,
+      "loss": 0.2687,
+      "step": 5316
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8214363676053267e-05,
+      "loss": 0.4951,
+      "step": 5317
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8214027840746898e-05,
+      "loss": 0.4341,
+      "step": 5318
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8213692005440533e-05,
+      "loss": 0.4012,
+      "step": 5319
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8213356170134167e-05,
+      "loss": 0.3993,
+      "step": 5320
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8213020334827802e-05,
+      "loss": 0.4404,
+      "step": 5321
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8212684499521437e-05,
+      "loss": 0.2453,
+      "step": 5322
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.821234866421507e-05,
+      "loss": 0.4125,
+      "step": 5323
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8212012828908706e-05,
+      "loss": 0.2608,
+      "step": 5324
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8211676993602337e-05,
+      "loss": 0.2907,
+      "step": 5325
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8211341158295972e-05,
+      "loss": 0.4037,
+      "step": 5326
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8211005322989607e-05,
+      "loss": 0.3959,
+      "step": 5327
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.821066948768324e-05,
+      "loss": 0.3436,
+      "step": 5328
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8210333652376876e-05,
+      "loss": 0.3779,
+      "step": 5329
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.820999781707051e-05,
+      "loss": 0.4681,
+      "step": 5330
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8209661981764145e-05,
+      "loss": 0.3368,
+      "step": 5331
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8209326146457776e-05,
+      "loss": 0.3,
+      "step": 5332
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8208990311151415e-05,
+      "loss": 0.3184,
+      "step": 5333
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8208654475845046e-05,
+      "loss": 0.2984,
+      "step": 5334
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.820831864053868e-05,
+      "loss": 0.3029,
+      "step": 5335
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8207982805232315e-05,
+      "loss": 0.2377,
+      "step": 5336
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.820764696992595e-05,
+      "loss": 0.2806,
+      "step": 5337
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8207311134619584e-05,
+      "loss": 0.3164,
+      "step": 5338
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.820697529931322e-05,
+      "loss": 0.3031,
+      "step": 5339
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8206639464006854e-05,
+      "loss": 0.4757,
+      "step": 5340
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8206303628700485e-05,
+      "loss": 0.3838,
+      "step": 5341
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8205967793394123e-05,
+      "loss": 0.3455,
+      "step": 5342
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8205631958087754e-05,
+      "loss": 0.2338,
+      "step": 5343
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.820529612278139e-05,
+      "loss": 0.4749,
+      "step": 5344
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8204960287475024e-05,
+      "loss": 0.3157,
+      "step": 5345
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8204624452168658e-05,
+      "loss": 0.299,
+      "step": 5346
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8204288616862293e-05,
+      "loss": 0.3789,
+      "step": 5347
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8203952781555927e-05,
+      "loss": 0.3753,
+      "step": 5348
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8203616946249562e-05,
+      "loss": 0.3442,
+      "step": 5349
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8203281110943193e-05,
+      "loss": 0.2787,
+      "step": 5350
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8202945275636828e-05,
+      "loss": 0.4812,
+      "step": 5351
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8202609440330463e-05,
+      "loss": 0.1835,
+      "step": 5352
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8202273605024097e-05,
+      "loss": 0.2184,
+      "step": 5353
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8201937769717732e-05,
+      "loss": 0.2218,
+      "step": 5354
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8201601934411367e-05,
+      "loss": 0.2538,
+      "step": 5355
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8201266099104998e-05,
+      "loss": 0.3599,
+      "step": 5356
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8200930263798636e-05,
+      "loss": 0.4933,
+      "step": 5357
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8200594428492267e-05,
+      "loss": 0.2098,
+      "step": 5358
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8200258593185902e-05,
+      "loss": 0.3433,
+      "step": 5359
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8199922757879536e-05,
+      "loss": 0.3132,
+      "step": 5360
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.819958692257317e-05,
+      "loss": 0.3521,
+      "step": 5361
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8199251087266806e-05,
+      "loss": 0.3025,
+      "step": 5362
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.819891525196044e-05,
+      "loss": 0.255,
+      "step": 5363
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8198579416654075e-05,
+      "loss": 0.2664,
+      "step": 5364
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8198243581347706e-05,
+      "loss": 0.24,
+      "step": 5365
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8197907746041344e-05,
+      "loss": 0.3185,
+      "step": 5366
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8197571910734976e-05,
+      "loss": 0.264,
+      "step": 5367
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.819723607542861e-05,
+      "loss": 0.2527,
+      "step": 5368
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8196900240122245e-05,
+      "loss": 0.2697,
+      "step": 5369
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.819656440481588e-05,
+      "loss": 0.3282,
+      "step": 5370
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8196228569509514e-05,
+      "loss": 0.2494,
+      "step": 5371
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.819589273420315e-05,
+      "loss": 0.4616,
+      "step": 5372
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8195556898896783e-05,
+      "loss": 0.3713,
+      "step": 5373
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8195221063590415e-05,
+      "loss": 0.5062,
+      "step": 5374
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8194885228284053e-05,
+      "loss": 0.4214,
+      "step": 5375
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8194549392977684e-05,
+      "loss": 0.2994,
+      "step": 5376
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.819421355767132e-05,
+      "loss": 0.2285,
+      "step": 5377
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8193877722364953e-05,
+      "loss": 0.4818,
+      "step": 5378
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8193541887058588e-05,
+      "loss": 0.3491,
+      "step": 5379
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8193206051752223e-05,
+      "loss": 0.2387,
+      "step": 5380
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8192870216445857e-05,
+      "loss": 0.3615,
+      "step": 5381
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8192534381139492e-05,
+      "loss": 0.4065,
+      "step": 5382
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8192198545833123e-05,
+      "loss": 0.285,
+      "step": 5383
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.819186271052676e-05,
+      "loss": 0.4062,
+      "step": 5384
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8191526875220392e-05,
+      "loss": 0.3829,
+      "step": 5385
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8191191039914027e-05,
+      "loss": 0.35,
+      "step": 5386
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8190855204607662e-05,
+      "loss": 0.2079,
+      "step": 5387
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8190519369301296e-05,
+      "loss": 0.2781,
+      "step": 5388
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.819018353399493e-05,
+      "loss": 0.1793,
+      "step": 5389
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8189847698688566e-05,
+      "loss": 0.4511,
+      "step": 5390
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.81895118633822e-05,
+      "loss": 0.1911,
+      "step": 5391
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.818917602807583e-05,
+      "loss": 0.3526,
+      "step": 5392
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.818884019276947e-05,
+      "loss": 0.4347,
+      "step": 5393
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.81885043574631e-05,
+      "loss": 0.2213,
+      "step": 5394
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8188168522156736e-05,
+      "loss": 0.2792,
+      "step": 5395
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.818783268685037e-05,
+      "loss": 0.1825,
+      "step": 5396
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8187496851544005e-05,
+      "loss": 0.3125,
+      "step": 5397
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8187161016237636e-05,
+      "loss": 0.3636,
+      "step": 5398
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8186825180931274e-05,
+      "loss": 0.4508,
+      "step": 5399
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8186489345624905e-05,
+      "loss": 0.3087,
+      "step": 5400
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.818615351031854e-05,
+      "loss": 0.2888,
+      "step": 5401
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8185817675012175e-05,
+      "loss": 0.2057,
+      "step": 5402
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.818548183970581e-05,
+      "loss": 0.4357,
+      "step": 5403
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8185146004399444e-05,
+      "loss": 0.3428,
+      "step": 5404
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.818481016909308e-05,
+      "loss": 0.2174,
+      "step": 5405
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8184474333786713e-05,
+      "loss": 0.2051,
+      "step": 5406
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8184138498480345e-05,
+      "loss": 0.3032,
+      "step": 5407
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8183802663173983e-05,
+      "loss": 0.3535,
+      "step": 5408
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8183466827867614e-05,
+      "loss": 0.3571,
+      "step": 5409
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.818313099256125e-05,
+      "loss": 0.2772,
+      "step": 5410
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8182795157254883e-05,
+      "loss": 0.5575,
+      "step": 5411
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8182459321948518e-05,
+      "loss": 0.2775,
+      "step": 5412
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8182123486642152e-05,
+      "loss": 0.3907,
+      "step": 5413
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8181787651335787e-05,
+      "loss": 0.249,
+      "step": 5414
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8181451816029422e-05,
+      "loss": 0.2684,
+      "step": 5415
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8181115980723053e-05,
+      "loss": 0.2744,
+      "step": 5416
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.818078014541669e-05,
+      "loss": 0.2793,
+      "step": 5417
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8180444310110322e-05,
+      "loss": 0.3685,
+      "step": 5418
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8180108474803957e-05,
+      "loss": 0.2723,
+      "step": 5419
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.817977263949759e-05,
+      "loss": 0.2723,
+      "step": 5420
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8179436804191226e-05,
+      "loss": 0.3996,
+      "step": 5421
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.817910096888486e-05,
+      "loss": 0.2892,
+      "step": 5422
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8178765133578496e-05,
+      "loss": 0.3859,
+      "step": 5423
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.817842929827213e-05,
+      "loss": 0.3633,
+      "step": 5424
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.817809346296576e-05,
+      "loss": 0.4472,
+      "step": 5425
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.81777576276594e-05,
+      "loss": 0.3781,
+      "step": 5426
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.817742179235303e-05,
+      "loss": 0.4253,
+      "step": 5427
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8177085957046665e-05,
+      "loss": 0.3219,
+      "step": 5428
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.81767501217403e-05,
+      "loss": 0.2392,
+      "step": 5429
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8176414286433935e-05,
+      "loss": 0.3063,
+      "step": 5430
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.817607845112757e-05,
+      "loss": 0.3221,
+      "step": 5431
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8175742615821204e-05,
+      "loss": 0.3453,
+      "step": 5432
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.817540678051484e-05,
+      "loss": 0.4435,
+      "step": 5433
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.817507094520847e-05,
+      "loss": 0.2965,
+      "step": 5434
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8174735109902108e-05,
+      "loss": 0.2892,
+      "step": 5435
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.817439927459574e-05,
+      "loss": 0.2666,
+      "step": 5436
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8174063439289374e-05,
+      "loss": 0.3864,
+      "step": 5437
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.817372760398301e-05,
+      "loss": 0.3133,
+      "step": 5438
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8173391768676643e-05,
+      "loss": 0.2962,
+      "step": 5439
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8173055933370278e-05,
+      "loss": 0.2919,
+      "step": 5440
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8172720098063912e-05,
+      "loss": 0.3597,
+      "step": 5441
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8172384262757544e-05,
+      "loss": 0.4685,
+      "step": 5442
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8172048427451178e-05,
+      "loss": 0.3819,
+      "step": 5443
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8171712592144813e-05,
+      "loss": 0.2191,
+      "step": 5444
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8171376756838448e-05,
+      "loss": 0.3583,
+      "step": 5445
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8171040921532082e-05,
+      "loss": 0.2713,
+      "step": 5446
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8170705086225717e-05,
+      "loss": 0.2163,
+      "step": 5447
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.817036925091935e-05,
+      "loss": 0.6083,
+      "step": 5448
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8170033415612983e-05,
+      "loss": 0.2031,
+      "step": 5449
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.816969758030662e-05,
+      "loss": 0.3703,
+      "step": 5450
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8169361745000252e-05,
+      "loss": 0.2691,
+      "step": 5451
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8169025909693887e-05,
+      "loss": 0.3311,
+      "step": 5452
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.816869007438752e-05,
+      "loss": 0.3539,
+      "step": 5453
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8168354239081156e-05,
+      "loss": 0.3126,
+      "step": 5454
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.816801840377479e-05,
+      "loss": 0.4333,
+      "step": 5455
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.8167682568468425e-05,
+      "loss": 0.3162,
+      "step": 5456
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.816734673316206e-05,
+      "loss": 0.3751,
+      "step": 5457
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.816701089785569e-05,
+      "loss": 0.612,
+      "step": 5458
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.816667506254933e-05,
+      "loss": 0.2331,
+      "step": 5459
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.816633922724296e-05,
+      "loss": 0.335,
+      "step": 5460
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8166003391936595e-05,
+      "loss": 0.3168,
+      "step": 5461
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.816566755663023e-05,
+      "loss": 0.4581,
+      "step": 5462
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8165331721323864e-05,
+      "loss": 0.2666,
+      "step": 5463
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.81649958860175e-05,
+      "loss": 0.2992,
+      "step": 5464
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8164660050711134e-05,
+      "loss": 0.5516,
+      "step": 5465
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.816432421540477e-05,
+      "loss": 0.1793,
+      "step": 5466
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.81639883800984e-05,
+      "loss": 0.4099,
+      "step": 5467
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8163652544792038e-05,
+      "loss": 0.3101,
+      "step": 5468
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.816331670948567e-05,
+      "loss": 0.2626,
+      "step": 5469
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8162980874179304e-05,
+      "loss": 0.2289,
+      "step": 5470
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8162645038872938e-05,
+      "loss": 0.3435,
+      "step": 5471
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8162309203566573e-05,
+      "loss": 0.2906,
+      "step": 5472
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8161973368260208e-05,
+      "loss": 0.4372,
+      "step": 5473
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8161637532953842e-05,
+      "loss": 0.3125,
+      "step": 5474
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8161301697647477e-05,
+      "loss": 0.569,
+      "step": 5475
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8160965862341108e-05,
+      "loss": 0.2931,
+      "step": 5476
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8160630027034746e-05,
+      "loss": 0.2549,
+      "step": 5477
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8160294191728377e-05,
+      "loss": 0.2259,
+      "step": 5478
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8159958356422012e-05,
+      "loss": 0.4088,
+      "step": 5479
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8159622521115647e-05,
+      "loss": 0.3324,
+      "step": 5480
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.815928668580928e-05,
+      "loss": 0.4723,
+      "step": 5481
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8158950850502916e-05,
+      "loss": 0.2229,
+      "step": 5482
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.815861501519655e-05,
+      "loss": 0.296,
+      "step": 5483
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8158279179890185e-05,
+      "loss": 0.2791,
+      "step": 5484
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8157943344583817e-05,
+      "loss": 0.2897,
+      "step": 5485
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.815760750927745e-05,
+      "loss": 0.3209,
+      "step": 5486
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8157271673971086e-05,
+      "loss": 0.3774,
+      "step": 5487
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.815693583866472e-05,
+      "loss": 0.3934,
+      "step": 5488
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8156600003358355e-05,
+      "loss": 0.5184,
+      "step": 5489
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.815626416805199e-05,
+      "loss": 0.2421,
+      "step": 5490
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.815592833274562e-05,
+      "loss": 0.162,
+      "step": 5491
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.815559249743926e-05,
+      "loss": 0.2732,
+      "step": 5492
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.815525666213289e-05,
+      "loss": 0.2383,
+      "step": 5493
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8154920826826525e-05,
+      "loss": 0.3365,
+      "step": 5494
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.815458499152016e-05,
+      "loss": 0.2477,
+      "step": 5495
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8154249156213794e-05,
+      "loss": 0.3039,
+      "step": 5496
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.815391332090743e-05,
+      "loss": 0.3262,
+      "step": 5497
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8153577485601064e-05,
+      "loss": 0.322,
+      "step": 5498
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8153241650294698e-05,
+      "loss": 0.1846,
+      "step": 5499
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.815290581498833e-05,
+      "loss": 0.1046,
+      "step": 5500
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8152569979681968e-05,
+      "loss": 0.2381,
+      "step": 5501
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.81522341443756e-05,
+      "loss": 0.3879,
+      "step": 5502
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8151898309069233e-05,
+      "loss": 0.3395,
+      "step": 5503
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8151562473762868e-05,
+      "loss": 0.5177,
+      "step": 5504
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8151226638456503e-05,
+      "loss": 0.428,
+      "step": 5505
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8150890803150137e-05,
+      "loss": 0.2706,
+      "step": 5506
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.815055496784377e-05,
+      "loss": 0.2931,
+      "step": 5507
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8150219132537407e-05,
+      "loss": 0.1999,
+      "step": 5508
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8149883297231038e-05,
+      "loss": 0.2641,
+      "step": 5509
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8149547461924673e-05,
+      "loss": 0.5415,
+      "step": 5510
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8149211626618307e-05,
+      "loss": 0.3091,
+      "step": 5511
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8148875791311942e-05,
+      "loss": 0.2773,
+      "step": 5512
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8148539956005577e-05,
+      "loss": 0.3355,
+      "step": 5513
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.814820412069921e-05,
+      "loss": 0.3648,
+      "step": 5514
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8147868285392846e-05,
+      "loss": 0.2154,
+      "step": 5515
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8147532450086477e-05,
+      "loss": 0.2893,
+      "step": 5516
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8147196614780115e-05,
+      "loss": 0.3892,
+      "step": 5517
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8146860779473746e-05,
+      "loss": 0.1894,
+      "step": 5518
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.814652494416738e-05,
+      "loss": 0.4344,
+      "step": 5519
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8146189108861016e-05,
+      "loss": 0.4155,
+      "step": 5520
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.814585327355465e-05,
+      "loss": 0.4203,
+      "step": 5521
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8145517438248285e-05,
+      "loss": 0.3372,
+      "step": 5522
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.814518160294192e-05,
+      "loss": 0.5463,
+      "step": 5523
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8144845767635554e-05,
+      "loss": 0.2548,
+      "step": 5524
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8144509932329185e-05,
+      "loss": 0.2109,
+      "step": 5525
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8144174097022824e-05,
+      "loss": 0.3902,
+      "step": 5526
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8143838261716455e-05,
+      "loss": 0.2195,
+      "step": 5527
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.814350242641009e-05,
+      "loss": 0.309,
+      "step": 5528
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8143166591103724e-05,
+      "loss": 0.3905,
+      "step": 5529
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.814283075579736e-05,
+      "loss": 0.2379,
+      "step": 5530
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8142494920490993e-05,
+      "loss": 0.2793,
+      "step": 5531
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8142159085184628e-05,
+      "loss": 0.3228,
+      "step": 5532
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.814182324987826e-05,
+      "loss": 0.2726,
+      "step": 5533
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8141487414571894e-05,
+      "loss": 0.3085,
+      "step": 5534
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.814115157926553e-05,
+      "loss": 0.3259,
+      "step": 5535
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8140815743959163e-05,
+      "loss": 0.2982,
+      "step": 5536
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8140479908652798e-05,
+      "loss": 0.3698,
+      "step": 5537
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8140144073346433e-05,
+      "loss": 0.3179,
+      "step": 5538
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8139808238040067e-05,
+      "loss": 0.3156,
+      "step": 5539
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.81394724027337e-05,
+      "loss": 0.2716,
+      "step": 5540
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8139136567427336e-05,
+      "loss": 0.4263,
+      "step": 5541
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8138800732120968e-05,
+      "loss": 0.2467,
+      "step": 5542
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8138464896814602e-05,
+      "loss": 0.4184,
+      "step": 5543
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8138129061508237e-05,
+      "loss": 0.2902,
+      "step": 5544
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.813779322620187e-05,
+      "loss": 0.3066,
+      "step": 5545
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8137457390895506e-05,
+      "loss": 0.2547,
+      "step": 5546
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.813712155558914e-05,
+      "loss": 0.2458,
+      "step": 5547
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8136785720282776e-05,
+      "loss": 0.3988,
+      "step": 5548
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8136449884976407e-05,
+      "loss": 0.1957,
+      "step": 5549
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8136114049670045e-05,
+      "loss": 0.2946,
+      "step": 5550
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8135778214363676e-05,
+      "loss": 0.4094,
+      "step": 5551
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.813544237905731e-05,
+      "loss": 0.3294,
+      "step": 5552
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8135106543750945e-05,
+      "loss": 0.3169,
+      "step": 5553
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.813477070844458e-05,
+      "loss": 0.4281,
+      "step": 5554
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8134434873138215e-05,
+      "loss": 0.3472,
+      "step": 5555
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.813409903783185e-05,
+      "loss": 0.4818,
+      "step": 5556
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8133763202525484e-05,
+      "loss": 0.1934,
+      "step": 5557
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8133427367219115e-05,
+      "loss": 0.2397,
+      "step": 5558
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8133091531912753e-05,
+      "loss": 0.1917,
+      "step": 5559
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8132755696606385e-05,
+      "loss": 0.5137,
+      "step": 5560
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.813241986130002e-05,
+      "loss": 0.3654,
+      "step": 5561
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8132084025993654e-05,
+      "loss": 0.2301,
+      "step": 5562
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.813174819068729e-05,
+      "loss": 0.2462,
+      "step": 5563
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8131412355380923e-05,
+      "loss": 0.576,
+      "step": 5564
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8131076520074558e-05,
+      "loss": 0.274,
+      "step": 5565
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8130740684768192e-05,
+      "loss": 0.3673,
+      "step": 5566
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8130404849461824e-05,
+      "loss": 0.2315,
+      "step": 5567
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8130069014155462e-05,
+      "loss": 0.3008,
+      "step": 5568
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8129733178849093e-05,
+      "loss": 0.2883,
+      "step": 5569
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8129397343542728e-05,
+      "loss": 0.3794,
+      "step": 5570
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8129061508236362e-05,
+      "loss": 0.3321,
+      "step": 5571
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8128725672929997e-05,
+      "loss": 0.3535,
+      "step": 5572
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.812838983762363e-05,
+      "loss": 0.3857,
+      "step": 5573
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8128054002317266e-05,
+      "loss": 0.3277,
+      "step": 5574
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.81277181670109e-05,
+      "loss": 0.3272,
+      "step": 5575
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8127382331704532e-05,
+      "loss": 0.4014,
+      "step": 5576
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8127046496398167e-05,
+      "loss": 0.371,
+      "step": 5577
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.81267106610918e-05,
+      "loss": 0.3288,
+      "step": 5578
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8126374825785436e-05,
+      "loss": 0.2758,
+      "step": 5579
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.812603899047907e-05,
+      "loss": 0.3564,
+      "step": 5580
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8125703155172705e-05,
+      "loss": 0.3546,
+      "step": 5581
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8125367319866337e-05,
+      "loss": 0.3371,
+      "step": 5582
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8125031484559975e-05,
+      "loss": 0.3291,
+      "step": 5583
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8124695649253606e-05,
+      "loss": 0.3797,
+      "step": 5584
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.812435981394724e-05,
+      "loss": 0.4267,
+      "step": 5585
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8124023978640875e-05,
+      "loss": 0.2051,
+      "step": 5586
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.812368814333451e-05,
+      "loss": 0.2237,
+      "step": 5587
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8123352308028145e-05,
+      "loss": 0.4279,
+      "step": 5588
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.812301647272178e-05,
+      "loss": 0.1759,
+      "step": 5589
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8122680637415414e-05,
+      "loss": 0.1613,
+      "step": 5590
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8122344802109045e-05,
+      "loss": 0.4716,
+      "step": 5591
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8122008966802683e-05,
+      "loss": 0.3414,
+      "step": 5592
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8121673131496314e-05,
+      "loss": 0.1959,
+      "step": 5593
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.812133729618995e-05,
+      "loss": 0.3504,
+      "step": 5594
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8121001460883584e-05,
+      "loss": 0.4925,
+      "step": 5595
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.812066562557722e-05,
+      "loss": 0.3885,
+      "step": 5596
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8120329790270853e-05,
+      "loss": 0.3283,
+      "step": 5597
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8119993954964488e-05,
+      "loss": 0.3674,
+      "step": 5598
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8119658119658122e-05,
+      "loss": 0.2906,
+      "step": 5599
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8119322284351754e-05,
+      "loss": 0.2503,
+      "step": 5600
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.811898644904539e-05,
+      "loss": 0.3085,
+      "step": 5601
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8118650613739023e-05,
+      "loss": 0.534,
+      "step": 5602
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8118314778432658e-05,
+      "loss": 0.1824,
+      "step": 5603
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8117978943126292e-05,
+      "loss": 0.2143,
+      "step": 5604
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8117643107819927e-05,
+      "loss": 0.2719,
+      "step": 5605
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.811730727251356e-05,
+      "loss": 0.3195,
+      "step": 5606
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8116971437207196e-05,
+      "loss": 0.3888,
+      "step": 5607
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.811663560190083e-05,
+      "loss": 0.4212,
+      "step": 5608
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8116299766594462e-05,
+      "loss": 0.2244,
+      "step": 5609
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.81159639312881e-05,
+      "loss": 0.4013,
+      "step": 5610
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.811562809598173e-05,
+      "loss": 0.3989,
+      "step": 5611
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8115292260675366e-05,
+      "loss": 0.2984,
+      "step": 5612
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8114956425369e-05,
+      "loss": 0.2594,
+      "step": 5613
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8114620590062635e-05,
+      "loss": 0.2643,
+      "step": 5614
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.811428475475627e-05,
+      "loss": 0.3675,
+      "step": 5615
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8113948919449905e-05,
+      "loss": 0.3836,
+      "step": 5616
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.811361308414354e-05,
+      "loss": 0.3944,
+      "step": 5617
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.811327724883717e-05,
+      "loss": 0.6276,
+      "step": 5618
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.811294141353081e-05,
+      "loss": 0.3012,
+      "step": 5619
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.811260557822444e-05,
+      "loss": 0.3465,
+      "step": 5620
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8112269742918074e-05,
+      "loss": 0.2593,
+      "step": 5621
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.811193390761171e-05,
+      "loss": 0.4113,
+      "step": 5622
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8111598072305344e-05,
+      "loss": 0.379,
+      "step": 5623
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8111262236998975e-05,
+      "loss": 0.308,
+      "step": 5624
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8110926401692613e-05,
+      "loss": 0.4129,
+      "step": 5625
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8110590566386244e-05,
+      "loss": 0.3691,
+      "step": 5626
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.811025473107988e-05,
+      "loss": 0.4216,
+      "step": 5627
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8109918895773514e-05,
+      "loss": 0.4935,
+      "step": 5628
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8109583060467148e-05,
+      "loss": 0.4779,
+      "step": 5629
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8109247225160783e-05,
+      "loss": 0.268,
+      "step": 5630
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8108911389854417e-05,
+      "loss": 0.2866,
+      "step": 5631
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8108575554548052e-05,
+      "loss": 0.2106,
+      "step": 5632
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8108239719241683e-05,
+      "loss": 0.405,
+      "step": 5633
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.810790388393532e-05,
+      "loss": 0.4127,
+      "step": 5634
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8107568048628953e-05,
+      "loss": 0.2385,
+      "step": 5635
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8107232213322587e-05,
+      "loss": 0.5442,
+      "step": 5636
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8106896378016222e-05,
+      "loss": 0.4435,
+      "step": 5637
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8106560542709857e-05,
+      "loss": 0.3536,
+      "step": 5638
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.810622470740349e-05,
+      "loss": 0.2772,
+      "step": 5639
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8105888872097126e-05,
+      "loss": 0.2712,
+      "step": 5640
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.810555303679076e-05,
+      "loss": 0.3671,
+      "step": 5641
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8105217201484392e-05,
+      "loss": 0.3499,
+      "step": 5642
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.810488136617803e-05,
+      "loss": 0.3035,
+      "step": 5643
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.810454553087166e-05,
+      "loss": 0.44,
+      "step": 5644
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8104209695565296e-05,
+      "loss": 0.3183,
+      "step": 5645
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.810387386025893e-05,
+      "loss": 0.2929,
+      "step": 5646
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8103538024952565e-05,
+      "loss": 0.4008,
+      "step": 5647
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.81032021896462e-05,
+      "loss": 0.3671,
+      "step": 5648
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8102866354339834e-05,
+      "loss": 0.3491,
+      "step": 5649
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.810253051903347e-05,
+      "loss": 0.4461,
+      "step": 5650
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.81021946837271e-05,
+      "loss": 0.3687,
+      "step": 5651
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8101858848420738e-05,
+      "loss": 0.2555,
+      "step": 5652
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.810152301311437e-05,
+      "loss": 0.2398,
+      "step": 5653
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8101187177808004e-05,
+      "loss": 0.3813,
+      "step": 5654
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.810085134250164e-05,
+      "loss": 0.4194,
+      "step": 5655
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8100515507195273e-05,
+      "loss": 0.3001,
+      "step": 5656
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.8100179671888908e-05,
+      "loss": 0.1959,
+      "step": 5657
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8099843836582543e-05,
+      "loss": 0.2887,
+      "step": 5658
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8099508001276177e-05,
+      "loss": 0.2892,
+      "step": 5659
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.809917216596981e-05,
+      "loss": 0.3045,
+      "step": 5660
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8098836330663447e-05,
+      "loss": 0.4079,
+      "step": 5661
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8098500495357078e-05,
+      "loss": 0.2527,
+      "step": 5662
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8098164660050713e-05,
+      "loss": 0.3024,
+      "step": 5663
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8097828824744347e-05,
+      "loss": 0.2559,
+      "step": 5664
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8097492989437982e-05,
+      "loss": 0.3478,
+      "step": 5665
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8097157154131617e-05,
+      "loss": 0.1739,
+      "step": 5666
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.809682131882525e-05,
+      "loss": 0.2987,
+      "step": 5667
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8096485483518882e-05,
+      "loss": 0.2628,
+      "step": 5668
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8096149648212517e-05,
+      "loss": 0.2706,
+      "step": 5669
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8095813812906152e-05,
+      "loss": 0.2753,
+      "step": 5670
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8095477977599786e-05,
+      "loss": 0.4296,
+      "step": 5671
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.809514214229342e-05,
+      "loss": 0.4805,
+      "step": 5672
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8094806306987056e-05,
+      "loss": 0.3586,
+      "step": 5673
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.809447047168069e-05,
+      "loss": 0.4402,
+      "step": 5674
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.809413463637432e-05,
+      "loss": 0.3655,
+      "step": 5675
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.809379880106796e-05,
+      "loss": 0.2356,
+      "step": 5676
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.809346296576159e-05,
+      "loss": 0.3366,
+      "step": 5677
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8093127130455226e-05,
+      "loss": 0.3194,
+      "step": 5678
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.809279129514886e-05,
+      "loss": 0.3303,
+      "step": 5679
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8092455459842495e-05,
+      "loss": 0.3226,
+      "step": 5680
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.809211962453613e-05,
+      "loss": 0.3324,
+      "step": 5681
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8091783789229764e-05,
+      "loss": 0.5754,
+      "step": 5682
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.80914479539234e-05,
+      "loss": 0.2091,
+      "step": 5683
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.809111211861703e-05,
+      "loss": 0.2643,
+      "step": 5684
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8090776283310665e-05,
+      "loss": 0.2551,
+      "step": 5685
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.80904404480043e-05,
+      "loss": 0.3393,
+      "step": 5686
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8090104612697934e-05,
+      "loss": 0.6104,
+      "step": 5687
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.808976877739157e-05,
+      "loss": 0.4698,
+      "step": 5688
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8089432942085203e-05,
+      "loss": 0.2759,
+      "step": 5689
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8089097106778838e-05,
+      "loss": 0.2814,
+      "step": 5690
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.808876127147247e-05,
+      "loss": 0.342,
+      "step": 5691
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8088425436166107e-05,
+      "loss": 0.1839,
+      "step": 5692
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.808808960085974e-05,
+      "loss": 0.2384,
+      "step": 5693
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8087753765553373e-05,
+      "loss": 0.4696,
+      "step": 5694
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8087417930247008e-05,
+      "loss": 0.5536,
+      "step": 5695
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8087082094940642e-05,
+      "loss": 0.1701,
+      "step": 5696
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8086746259634277e-05,
+      "loss": 0.2694,
+      "step": 5697
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8086410424327912e-05,
+      "loss": 0.5818,
+      "step": 5698
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8086074589021546e-05,
+      "loss": 0.483,
+      "step": 5699
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8085738753715178e-05,
+      "loss": 0.3779,
+      "step": 5700
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8085402918408816e-05,
+      "loss": 0.2797,
+      "step": 5701
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8085067083102447e-05,
+      "loss": 0.3166,
+      "step": 5702
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.808473124779608e-05,
+      "loss": 0.3729,
+      "step": 5703
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8084395412489716e-05,
+      "loss": 0.4701,
+      "step": 5704
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.808405957718335e-05,
+      "loss": 0.3518,
+      "step": 5705
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8083723741876986e-05,
+      "loss": 0.3125,
+      "step": 5706
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.808338790657062e-05,
+      "loss": 0.2266,
+      "step": 5707
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8083052071264255e-05,
+      "loss": 0.4633,
+      "step": 5708
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8082716235957886e-05,
+      "loss": 0.3364,
+      "step": 5709
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8082380400651524e-05,
+      "loss": 0.298,
+      "step": 5710
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8082044565345155e-05,
+      "loss": 0.4166,
+      "step": 5711
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.808170873003879e-05,
+      "loss": 0.4724,
+      "step": 5712
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8081372894732425e-05,
+      "loss": 0.3877,
+      "step": 5713
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.808103705942606e-05,
+      "loss": 0.3435,
+      "step": 5714
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.808070122411969e-05,
+      "loss": 0.3678,
+      "step": 5715
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.808036538881333e-05,
+      "loss": 0.2718,
+      "step": 5716
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.808002955350696e-05,
+      "loss": 0.2664,
+      "step": 5717
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8079693718200595e-05,
+      "loss": 0.2101,
+      "step": 5718
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.807935788289423e-05,
+      "loss": 0.324,
+      "step": 5719
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8079022047587864e-05,
+      "loss": 0.3538,
+      "step": 5720
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.80786862122815e-05,
+      "loss": 0.3196,
+      "step": 5721
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8078350376975133e-05,
+      "loss": 0.3731,
+      "step": 5722
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8078014541668768e-05,
+      "loss": 0.3076,
+      "step": 5723
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.80776787063624e-05,
+      "loss": 0.4288,
+      "step": 5724
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8077342871056037e-05,
+      "loss": 0.2997,
+      "step": 5725
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.807700703574967e-05,
+      "loss": 0.3833,
+      "step": 5726
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8076671200443303e-05,
+      "loss": 0.3034,
+      "step": 5727
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8076335365136938e-05,
+      "loss": 0.3133,
+      "step": 5728
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8075999529830572e-05,
+      "loss": 0.436,
+      "step": 5729
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8075663694524207e-05,
+      "loss": 0.2846,
+      "step": 5730
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.807532785921784e-05,
+      "loss": 0.3165,
+      "step": 5731
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8074992023911476e-05,
+      "loss": 0.4198,
+      "step": 5732
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8074656188605107e-05,
+      "loss": 0.4602,
+      "step": 5733
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8074320353298745e-05,
+      "loss": 0.1826,
+      "step": 5734
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8073984517992377e-05,
+      "loss": 0.4179,
+      "step": 5735
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.807364868268601e-05,
+      "loss": 0.2858,
+      "step": 5736
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8073312847379646e-05,
+      "loss": 0.3428,
+      "step": 5737
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.807297701207328e-05,
+      "loss": 0.332,
+      "step": 5738
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8072641176766915e-05,
+      "loss": 0.4022,
+      "step": 5739
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.807230534146055e-05,
+      "loss": 0.2218,
+      "step": 5740
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8071969506154185e-05,
+      "loss": 0.5151,
+      "step": 5741
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8071633670847816e-05,
+      "loss": 0.3513,
+      "step": 5742
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8071297835541454e-05,
+      "loss": 0.3247,
+      "step": 5743
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8070962000235085e-05,
+      "loss": 0.3278,
+      "step": 5744
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.807062616492872e-05,
+      "loss": 0.3296,
+      "step": 5745
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8070290329622354e-05,
+      "loss": 0.4585,
+      "step": 5746
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.806995449431599e-05,
+      "loss": 0.2769,
+      "step": 5747
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8069618659009624e-05,
+      "loss": 0.3055,
+      "step": 5748
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.806928282370326e-05,
+      "loss": 0.2857,
+      "step": 5749
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8068946988396893e-05,
+      "loss": 0.3271,
+      "step": 5750
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8068611153090524e-05,
+      "loss": 0.3562,
+      "step": 5751
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8068275317784162e-05,
+      "loss": 0.2286,
+      "step": 5752
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8067939482477794e-05,
+      "loss": 0.4335,
+      "step": 5753
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8067603647171428e-05,
+      "loss": 0.5236,
+      "step": 5754
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8067267811865063e-05,
+      "loss": 0.4421,
+      "step": 5755
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8066931976558698e-05,
+      "loss": 0.1948,
+      "step": 5756
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8066596141252332e-05,
+      "loss": 0.4259,
+      "step": 5757
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8066260305945967e-05,
+      "loss": 0.4465,
+      "step": 5758
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8065924470639598e-05,
+      "loss": 0.3088,
+      "step": 5759
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8065588635333233e-05,
+      "loss": 0.2374,
+      "step": 5760
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8065252800026867e-05,
+      "loss": 0.2679,
+      "step": 5761
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8064916964720502e-05,
+      "loss": 0.2157,
+      "step": 5762
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8064581129414137e-05,
+      "loss": 0.4258,
+      "step": 5763
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.806424529410777e-05,
+      "loss": 0.3058,
+      "step": 5764
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8063909458801406e-05,
+      "loss": 0.2995,
+      "step": 5765
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8063573623495037e-05,
+      "loss": 0.4242,
+      "step": 5766
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8063237788188675e-05,
+      "loss": 0.3265,
+      "step": 5767
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8062901952882307e-05,
+      "loss": 0.3337,
+      "step": 5768
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.806256611757594e-05,
+      "loss": 0.4267,
+      "step": 5769
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8062230282269576e-05,
+      "loss": 0.2628,
+      "step": 5770
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.806189444696321e-05,
+      "loss": 0.5595,
+      "step": 5771
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8061558611656845e-05,
+      "loss": 0.3525,
+      "step": 5772
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.806122277635048e-05,
+      "loss": 0.5372,
+      "step": 5773
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8060886941044114e-05,
+      "loss": 0.3782,
+      "step": 5774
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8060551105737746e-05,
+      "loss": 0.3284,
+      "step": 5775
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8060215270431384e-05,
+      "loss": 0.3917,
+      "step": 5776
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8059879435125015e-05,
+      "loss": 0.2379,
+      "step": 5777
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.805954359981865e-05,
+      "loss": 0.4533,
+      "step": 5778
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8059207764512284e-05,
+      "loss": 0.4682,
+      "step": 5779
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.805887192920592e-05,
+      "loss": 0.3763,
+      "step": 5780
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8058536093899554e-05,
+      "loss": 0.1791,
+      "step": 5781
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8058200258593188e-05,
+      "loss": 0.3386,
+      "step": 5782
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8057864423286823e-05,
+      "loss": 0.4078,
+      "step": 5783
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8057528587980454e-05,
+      "loss": 0.3163,
+      "step": 5784
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8057192752674092e-05,
+      "loss": 0.3535,
+      "step": 5785
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8056856917367723e-05,
+      "loss": 0.2996,
+      "step": 5786
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8056521082061358e-05,
+      "loss": 0.1638,
+      "step": 5787
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8056185246754993e-05,
+      "loss": 0.238,
+      "step": 5788
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8055849411448627e-05,
+      "loss": 0.4125,
+      "step": 5789
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8055513576142262e-05,
+      "loss": 0.4525,
+      "step": 5790
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8055177740835897e-05,
+      "loss": 0.2637,
+      "step": 5791
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.805484190552953e-05,
+      "loss": 0.2626,
+      "step": 5792
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8054506070223163e-05,
+      "loss": 0.5111,
+      "step": 5793
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.80541702349168e-05,
+      "loss": 0.43,
+      "step": 5794
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8053834399610432e-05,
+      "loss": 0.3793,
+      "step": 5795
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8053498564304067e-05,
+      "loss": 0.2141,
+      "step": 5796
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.80531627289977e-05,
+      "loss": 0.2764,
+      "step": 5797
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8052826893691336e-05,
+      "loss": 0.4046,
+      "step": 5798
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.805249105838497e-05,
+      "loss": 0.2444,
+      "step": 5799
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8052155223078605e-05,
+      "loss": 0.4378,
+      "step": 5800
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.805181938777224e-05,
+      "loss": 0.36,
+      "step": 5801
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.805148355246587e-05,
+      "loss": 0.2847,
+      "step": 5802
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8051147717159506e-05,
+      "loss": 0.3208,
+      "step": 5803
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.805081188185314e-05,
+      "loss": 0.2524,
+      "step": 5804
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8050476046546775e-05,
+      "loss": 0.3462,
+      "step": 5805
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.805014021124041e-05,
+      "loss": 0.4031,
+      "step": 5806
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8049804375934044e-05,
+      "loss": 0.3989,
+      "step": 5807
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8049468540627676e-05,
+      "loss": 0.3053,
+      "step": 5808
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8049132705321314e-05,
+      "loss": 0.3641,
+      "step": 5809
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8048796870014945e-05,
+      "loss": 0.3038,
+      "step": 5810
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.804846103470858e-05,
+      "loss": 0.2239,
+      "step": 5811
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8048125199402214e-05,
+      "loss": 0.278,
+      "step": 5812
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.804778936409585e-05,
+      "loss": 0.5312,
+      "step": 5813
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8047453528789483e-05,
+      "loss": 0.2667,
+      "step": 5814
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8047117693483118e-05,
+      "loss": 0.1796,
+      "step": 5815
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8046781858176753e-05,
+      "loss": 0.4282,
+      "step": 5816
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8046446022870384e-05,
+      "loss": 0.2604,
+      "step": 5817
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8046110187564022e-05,
+      "loss": 0.3175,
+      "step": 5818
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8045774352257653e-05,
+      "loss": 0.372,
+      "step": 5819
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8045438516951288e-05,
+      "loss": 0.4343,
+      "step": 5820
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8045102681644923e-05,
+      "loss": 0.4315,
+      "step": 5821
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8044766846338557e-05,
+      "loss": 0.3033,
+      "step": 5822
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8044431011032192e-05,
+      "loss": 0.235,
+      "step": 5823
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8044095175725826e-05,
+      "loss": 0.3871,
+      "step": 5824
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.804375934041946e-05,
+      "loss": 0.3424,
+      "step": 5825
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8043423505113092e-05,
+      "loss": 0.3479,
+      "step": 5826
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.804308766980673e-05,
+      "loss": 0.4143,
+      "step": 5827
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.804275183450036e-05,
+      "loss": 0.3468,
+      "step": 5828
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8042415999193996e-05,
+      "loss": 0.4128,
+      "step": 5829
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.804208016388763e-05,
+      "loss": 0.4435,
+      "step": 5830
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8041744328581266e-05,
+      "loss": 0.3138,
+      "step": 5831
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.80414084932749e-05,
+      "loss": 0.397,
+      "step": 5832
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8041072657968535e-05,
+      "loss": 0.3419,
+      "step": 5833
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.804073682266217e-05,
+      "loss": 0.359,
+      "step": 5834
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.80404009873558e-05,
+      "loss": 0.305,
+      "step": 5835
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.804006515204944e-05,
+      "loss": 0.4547,
+      "step": 5836
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.803972931674307e-05,
+      "loss": 0.3725,
+      "step": 5837
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8039393481436705e-05,
+      "loss": 0.3776,
+      "step": 5838
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.803905764613034e-05,
+      "loss": 0.4336,
+      "step": 5839
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8038721810823974e-05,
+      "loss": 0.4188,
+      "step": 5840
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.803838597551761e-05,
+      "loss": 0.268,
+      "step": 5841
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8038050140211243e-05,
+      "loss": 0.4445,
+      "step": 5842
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8037714304904878e-05,
+      "loss": 0.3113,
+      "step": 5843
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.803737846959851e-05,
+      "loss": 0.3185,
+      "step": 5844
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8037042634292147e-05,
+      "loss": 0.3306,
+      "step": 5845
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.803670679898578e-05,
+      "loss": 0.3383,
+      "step": 5846
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8036370963679413e-05,
+      "loss": 0.4418,
+      "step": 5847
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8036035128373048e-05,
+      "loss": 0.4529,
+      "step": 5848
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8035699293066683e-05,
+      "loss": 0.4089,
+      "step": 5849
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8035363457760314e-05,
+      "loss": 0.2659,
+      "step": 5850
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8035027622453952e-05,
+      "loss": 0.4834,
+      "step": 5851
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8034691787147583e-05,
+      "loss": 0.2187,
+      "step": 5852
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8034355951841218e-05,
+      "loss": 0.3892,
+      "step": 5853
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8034020116534852e-05,
+      "loss": 0.4928,
+      "step": 5854
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.8033684281228487e-05,
+      "loss": 0.2446,
+      "step": 5855
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.803334844592212e-05,
+      "loss": 0.1267,
+      "step": 5856
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8033012610615756e-05,
+      "loss": 0.3569,
+      "step": 5857
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.803267677530939e-05,
+      "loss": 0.2527,
+      "step": 5858
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8032340940003022e-05,
+      "loss": 0.2452,
+      "step": 5859
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.803200510469666e-05,
+      "loss": 0.1947,
+      "step": 5860
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.803166926939029e-05,
+      "loss": 0.424,
+      "step": 5861
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8031333434083926e-05,
+      "loss": 0.3124,
+      "step": 5862
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.803099759877756e-05,
+      "loss": 0.4109,
+      "step": 5863
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8030661763471195e-05,
+      "loss": 0.306,
+      "step": 5864
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.803032592816483e-05,
+      "loss": 0.291,
+      "step": 5865
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.802999009285846e-05,
+      "loss": 0.3014,
+      "step": 5866
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.80296542575521e-05,
+      "loss": 0.2902,
+      "step": 5867
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.802931842224573e-05,
+      "loss": 0.4503,
+      "step": 5868
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8028982586939365e-05,
+      "loss": 0.4063,
+      "step": 5869
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8028646751633e-05,
+      "loss": 0.2755,
+      "step": 5870
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8028310916326635e-05,
+      "loss": 0.5002,
+      "step": 5871
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.802797508102027e-05,
+      "loss": 0.2983,
+      "step": 5872
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8027639245713904e-05,
+      "loss": 0.2976,
+      "step": 5873
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.802730341040754e-05,
+      "loss": 0.4009,
+      "step": 5874
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.802696757510117e-05,
+      "loss": 0.4284,
+      "step": 5875
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8026631739794808e-05,
+      "loss": 0.2533,
+      "step": 5876
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.802629590448844e-05,
+      "loss": 0.3672,
+      "step": 5877
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8025960069182074e-05,
+      "loss": 0.5778,
+      "step": 5878
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.802562423387571e-05,
+      "loss": 0.463,
+      "step": 5879
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8025288398569343e-05,
+      "loss": 0.3378,
+      "step": 5880
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8024952563262978e-05,
+      "loss": 0.3443,
+      "step": 5881
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8024616727956612e-05,
+      "loss": 0.2424,
+      "step": 5882
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8024280892650247e-05,
+      "loss": 0.2853,
+      "step": 5883
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8023945057343878e-05,
+      "loss": 0.2567,
+      "step": 5884
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8023609222037516e-05,
+      "loss": 0.307,
+      "step": 5885
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8023273386731148e-05,
+      "loss": 0.2174,
+      "step": 5886
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8022937551424782e-05,
+      "loss": 0.2457,
+      "step": 5887
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8022601716118417e-05,
+      "loss": 0.2481,
+      "step": 5888
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.802226588081205e-05,
+      "loss": 0.3838,
+      "step": 5889
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8021930045505686e-05,
+      "loss": 0.1555,
+      "step": 5890
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.802159421019932e-05,
+      "loss": 0.2718,
+      "step": 5891
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8021258374892955e-05,
+      "loss": 0.3807,
+      "step": 5892
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8020922539586587e-05,
+      "loss": 0.3211,
+      "step": 5893
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.802058670428022e-05,
+      "loss": 0.4387,
+      "step": 5894
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8020250868973856e-05,
+      "loss": 0.2107,
+      "step": 5895
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.801991503366749e-05,
+      "loss": 0.7031,
+      "step": 5896
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8019579198361125e-05,
+      "loss": 0.2681,
+      "step": 5897
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.801924336305476e-05,
+      "loss": 0.4363,
+      "step": 5898
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.801890752774839e-05,
+      "loss": 0.2392,
+      "step": 5899
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.801857169244203e-05,
+      "loss": 0.3662,
+      "step": 5900
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.801823585713566e-05,
+      "loss": 0.4076,
+      "step": 5901
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8017900021829295e-05,
+      "loss": 0.3394,
+      "step": 5902
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.801756418652293e-05,
+      "loss": 0.5299,
+      "step": 5903
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8017228351216564e-05,
+      "loss": 0.316,
+      "step": 5904
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.80168925159102e-05,
+      "loss": 0.2356,
+      "step": 5905
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8016556680603834e-05,
+      "loss": 0.3424,
+      "step": 5906
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.801622084529747e-05,
+      "loss": 0.4316,
+      "step": 5907
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.80158850099911e-05,
+      "loss": 0.3109,
+      "step": 5908
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8015549174684738e-05,
+      "loss": 0.2769,
+      "step": 5909
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.801521333937837e-05,
+      "loss": 0.241,
+      "step": 5910
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8014877504072004e-05,
+      "loss": 0.3009,
+      "step": 5911
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8014541668765638e-05,
+      "loss": 0.2651,
+      "step": 5912
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8014205833459273e-05,
+      "loss": 0.2724,
+      "step": 5913
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8013869998152907e-05,
+      "loss": 0.2179,
+      "step": 5914
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8013534162846542e-05,
+      "loss": 0.4253,
+      "step": 5915
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8013198327540177e-05,
+      "loss": 0.4795,
+      "step": 5916
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8012862492233808e-05,
+      "loss": 0.2752,
+      "step": 5917
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8012526656927446e-05,
+      "loss": 0.4672,
+      "step": 5918
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8012190821621077e-05,
+      "loss": 0.2871,
+      "step": 5919
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8011854986314712e-05,
+      "loss": 0.3545,
+      "step": 5920
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8011519151008347e-05,
+      "loss": 0.3502,
+      "step": 5921
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.801118331570198e-05,
+      "loss": 0.4184,
+      "step": 5922
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8010847480395616e-05,
+      "loss": 0.1885,
+      "step": 5923
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.801051164508925e-05,
+      "loss": 0.5936,
+      "step": 5924
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8010175809782885e-05,
+      "loss": 0.3597,
+      "step": 5925
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8009839974476516e-05,
+      "loss": 0.1848,
+      "step": 5926
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8009504139170155e-05,
+      "loss": 0.2155,
+      "step": 5927
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8009168303863786e-05,
+      "loss": 0.2736,
+      "step": 5928
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.800883246855742e-05,
+      "loss": 0.4194,
+      "step": 5929
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8008496633251055e-05,
+      "loss": 0.1933,
+      "step": 5930
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.800816079794469e-05,
+      "loss": 0.2985,
+      "step": 5931
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8007824962638324e-05,
+      "loss": 0.4772,
+      "step": 5932
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.800748912733196e-05,
+      "loss": 0.2745,
+      "step": 5933
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8007153292025594e-05,
+      "loss": 0.2654,
+      "step": 5934
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8006817456719225e-05,
+      "loss": 0.2971,
+      "step": 5935
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8006481621412863e-05,
+      "loss": 0.4464,
+      "step": 5936
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8006145786106494e-05,
+      "loss": 0.3015,
+      "step": 5937
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.800580995080013e-05,
+      "loss": 0.1595,
+      "step": 5938
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8005474115493763e-05,
+      "loss": 0.2897,
+      "step": 5939
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8005138280187398e-05,
+      "loss": 0.3207,
+      "step": 5940
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.800480244488103e-05,
+      "loss": 0.2979,
+      "step": 5941
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8004466609574667e-05,
+      "loss": 0.491,
+      "step": 5942
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.80041307742683e-05,
+      "loss": 0.3043,
+      "step": 5943
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8003794938961933e-05,
+      "loss": 0.2668,
+      "step": 5944
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8003459103655568e-05,
+      "loss": 0.2672,
+      "step": 5945
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8003123268349203e-05,
+      "loss": 0.2359,
+      "step": 5946
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8002787433042837e-05,
+      "loss": 0.2175,
+      "step": 5947
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8002451597736472e-05,
+      "loss": 0.2173,
+      "step": 5948
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8002115762430107e-05,
+      "loss": 0.4156,
+      "step": 5949
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8001779927123738e-05,
+      "loss": 0.4427,
+      "step": 5950
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8001444091817376e-05,
+      "loss": 0.1679,
+      "step": 5951
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8001108256511007e-05,
+      "loss": 0.4318,
+      "step": 5952
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8000772421204642e-05,
+      "loss": 0.2562,
+      "step": 5953
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.8000436585898276e-05,
+      "loss": 0.2555,
+      "step": 5954
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.800010075059191e-05,
+      "loss": 0.3952,
+      "step": 5955
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7999764915285546e-05,
+      "loss": 0.3795,
+      "step": 5956
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.799942907997918e-05,
+      "loss": 0.3416,
+      "step": 5957
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7999093244672815e-05,
+      "loss": 0.1555,
+      "step": 5958
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7998757409366446e-05,
+      "loss": 0.2393,
+      "step": 5959
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7998421574060084e-05,
+      "loss": 0.4501,
+      "step": 5960
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7998085738753716e-05,
+      "loss": 0.1934,
+      "step": 5961
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.799774990344735e-05,
+      "loss": 0.4442,
+      "step": 5962
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7997414068140985e-05,
+      "loss": 0.2138,
+      "step": 5963
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.799707823283462e-05,
+      "loss": 0.2535,
+      "step": 5964
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7996742397528254e-05,
+      "loss": 0.3305,
+      "step": 5965
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.799640656222189e-05,
+      "loss": 0.2568,
+      "step": 5966
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7996070726915523e-05,
+      "loss": 0.37,
+      "step": 5967
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7995734891609155e-05,
+      "loss": 0.4066,
+      "step": 5968
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7995399056302793e-05,
+      "loss": 0.4956,
+      "step": 5969
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7995063220996424e-05,
+      "loss": 0.4505,
+      "step": 5970
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.799472738569006e-05,
+      "loss": 0.6052,
+      "step": 5971
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7994391550383693e-05,
+      "loss": 0.4604,
+      "step": 5972
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7994055715077328e-05,
+      "loss": 0.211,
+      "step": 5973
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7993719879770963e-05,
+      "loss": 0.3498,
+      "step": 5974
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7993384044464597e-05,
+      "loss": 0.3001,
+      "step": 5975
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7993048209158232e-05,
+      "loss": 0.3484,
+      "step": 5976
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7992712373851863e-05,
+      "loss": 0.3032,
+      "step": 5977
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.79923765385455e-05,
+      "loss": 0.2088,
+      "step": 5978
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7992040703239132e-05,
+      "loss": 0.3423,
+      "step": 5979
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7991704867932767e-05,
+      "loss": 0.3907,
+      "step": 5980
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7991369032626402e-05,
+      "loss": 0.4093,
+      "step": 5981
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7991033197320036e-05,
+      "loss": 0.3527,
+      "step": 5982
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7990697362013668e-05,
+      "loss": 0.3368,
+      "step": 5983
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7990361526707306e-05,
+      "loss": 0.4407,
+      "step": 5984
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7990025691400937e-05,
+      "loss": 0.3983,
+      "step": 5985
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.798968985609457e-05,
+      "loss": 0.1787,
+      "step": 5986
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7989354020788206e-05,
+      "loss": 0.4016,
+      "step": 5987
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.798901818548184e-05,
+      "loss": 0.2149,
+      "step": 5988
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7988682350175476e-05,
+      "loss": 0.3425,
+      "step": 5989
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.798834651486911e-05,
+      "loss": 0.5138,
+      "step": 5990
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7988010679562745e-05,
+      "loss": 0.3477,
+      "step": 5991
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7987674844256376e-05,
+      "loss": 0.3163,
+      "step": 5992
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7987339008950014e-05,
+      "loss": 0.2406,
+      "step": 5993
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7987003173643645e-05,
+      "loss": 0.2758,
+      "step": 5994
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.798666733833728e-05,
+      "loss": 0.3274,
+      "step": 5995
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7986331503030915e-05,
+      "loss": 0.3911,
+      "step": 5996
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.798599566772455e-05,
+      "loss": 0.2334,
+      "step": 5997
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7985659832418184e-05,
+      "loss": 0.2962,
+      "step": 5998
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.798532399711182e-05,
+      "loss": 0.1739,
+      "step": 5999
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7984988161805453e-05,
+      "loss": 0.343,
+      "step": 6000
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7984652326499085e-05,
+      "loss": 0.3575,
+      "step": 6001
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7984316491192723e-05,
+      "loss": 0.3461,
+      "step": 6002
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7983980655886354e-05,
+      "loss": 0.2944,
+      "step": 6003
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.798364482057999e-05,
+      "loss": 0.4915,
+      "step": 6004
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7983308985273623e-05,
+      "loss": 0.3188,
+      "step": 6005
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7982973149967258e-05,
+      "loss": 0.3265,
+      "step": 6006
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7982637314660892e-05,
+      "loss": 0.3378,
+      "step": 6007
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7982301479354527e-05,
+      "loss": 0.2763,
+      "step": 6008
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7981965644048162e-05,
+      "loss": 0.3141,
+      "step": 6009
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7981629808741793e-05,
+      "loss": 0.2952,
+      "step": 6010
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.798129397343543e-05,
+      "loss": 0.2236,
+      "step": 6011
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7980958138129062e-05,
+      "loss": 0.189,
+      "step": 6012
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7980622302822697e-05,
+      "loss": 0.3946,
+      "step": 6013
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.798028646751633e-05,
+      "loss": 0.314,
+      "step": 6014
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7979950632209966e-05,
+      "loss": 0.5816,
+      "step": 6015
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.79796147969036e-05,
+      "loss": 0.1557,
+      "step": 6016
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7979278961597235e-05,
+      "loss": 0.5404,
+      "step": 6017
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.797894312629087e-05,
+      "loss": 0.3147,
+      "step": 6018
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.79786072909845e-05,
+      "loss": 0.2586,
+      "step": 6019
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.797827145567814e-05,
+      "loss": 0.4279,
+      "step": 6020
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.797793562037177e-05,
+      "loss": 0.4806,
+      "step": 6021
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7977599785065405e-05,
+      "loss": 0.3181,
+      "step": 6022
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.797726394975904e-05,
+      "loss": 0.3574,
+      "step": 6023
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7976928114452675e-05,
+      "loss": 0.2537,
+      "step": 6024
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.797659227914631e-05,
+      "loss": 0.241,
+      "step": 6025
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7976256443839944e-05,
+      "loss": 0.22,
+      "step": 6026
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.797592060853358e-05,
+      "loss": 0.2953,
+      "step": 6027
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.797558477322721e-05,
+      "loss": 0.2709,
+      "step": 6028
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7975248937920844e-05,
+      "loss": 0.2964,
+      "step": 6029
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.797491310261448e-05,
+      "loss": 0.3518,
+      "step": 6030
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7974577267308114e-05,
+      "loss": 0.5297,
+      "step": 6031
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.797424143200175e-05,
+      "loss": 0.3335,
+      "step": 6032
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7973905596695383e-05,
+      "loss": 0.4334,
+      "step": 6033
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7973569761389014e-05,
+      "loss": 0.2973,
+      "step": 6034
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7973233926082652e-05,
+      "loss": 0.3892,
+      "step": 6035
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7972898090776284e-05,
+      "loss": 0.2577,
+      "step": 6036
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7972562255469918e-05,
+      "loss": 0.3869,
+      "step": 6037
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7972226420163553e-05,
+      "loss": 0.3411,
+      "step": 6038
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7971890584857188e-05,
+      "loss": 0.3057,
+      "step": 6039
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7971554749550822e-05,
+      "loss": 0.5836,
+      "step": 6040
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7971218914244457e-05,
+      "loss": 0.247,
+      "step": 6041
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.797088307893809e-05,
+      "loss": 0.3127,
+      "step": 6042
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7970547243631723e-05,
+      "loss": 0.2562,
+      "step": 6043
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.797021140832536e-05,
+      "loss": 0.3696,
+      "step": 6044
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7969875573018992e-05,
+      "loss": 0.3058,
+      "step": 6045
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7969539737712627e-05,
+      "loss": 0.3198,
+      "step": 6046
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.796920390240626e-05,
+      "loss": 0.4172,
+      "step": 6047
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7968868067099896e-05,
+      "loss": 0.2563,
+      "step": 6048
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.796853223179353e-05,
+      "loss": 0.3491,
+      "step": 6049
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7968196396487162e-05,
+      "loss": 0.2658,
+      "step": 6050
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.79678605611808e-05,
+      "loss": 0.4075,
+      "step": 6051
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.796752472587443e-05,
+      "loss": 0.5106,
+      "step": 6052
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7967188890568066e-05,
+      "loss": 0.3671,
+      "step": 6053
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.79668530552617e-05,
+      "loss": 0.3387,
+      "step": 6054
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7966517219955335e-05,
+      "loss": 0.276,
+      "step": 6055
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.796618138464897e-05,
+      "loss": 0.3929,
+      "step": 6056
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7965845549342604e-05,
+      "loss": 0.4486,
+      "step": 6057
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.796550971403624e-05,
+      "loss": 0.3271,
+      "step": 6058
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.796517387872987e-05,
+      "loss": 0.2391,
+      "step": 6059
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.796483804342351e-05,
+      "loss": 0.4118,
+      "step": 6060
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.796450220811714e-05,
+      "loss": 0.2927,
+      "step": 6061
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7964166372810774e-05,
+      "loss": 0.5224,
+      "step": 6062
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.796383053750441e-05,
+      "loss": 0.2579,
+      "step": 6063
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7963494702198044e-05,
+      "loss": 0.4021,
+      "step": 6064
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7963158866891678e-05,
+      "loss": 0.3496,
+      "step": 6065
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7962823031585313e-05,
+      "loss": 0.4167,
+      "step": 6066
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7962487196278948e-05,
+      "loss": 0.2971,
+      "step": 6067
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.796215136097258e-05,
+      "loss": 0.3955,
+      "step": 6068
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7961815525666217e-05,
+      "loss": 0.3468,
+      "step": 6069
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7961479690359848e-05,
+      "loss": 0.3156,
+      "step": 6070
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7961143855053483e-05,
+      "loss": 0.2299,
+      "step": 6071
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7960808019747117e-05,
+      "loss": 0.2125,
+      "step": 6072
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7960472184440752e-05,
+      "loss": 0.2462,
+      "step": 6073
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7960136349134383e-05,
+      "loss": 0.3486,
+      "step": 6074
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.795980051382802e-05,
+      "loss": 0.4019,
+      "step": 6075
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7959464678521653e-05,
+      "loss": 0.3509,
+      "step": 6076
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7959128843215287e-05,
+      "loss": 0.1904,
+      "step": 6077
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7958793007908922e-05,
+      "loss": 0.2887,
+      "step": 6078
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7958457172602557e-05,
+      "loss": 0.3267,
+      "step": 6079
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.795812133729619e-05,
+      "loss": 0.4548,
+      "step": 6080
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7957785501989826e-05,
+      "loss": 0.3653,
+      "step": 6081
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.795744966668346e-05,
+      "loss": 0.3265,
+      "step": 6082
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7957113831377092e-05,
+      "loss": 0.2788,
+      "step": 6083
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.795677799607073e-05,
+      "loss": 0.4445,
+      "step": 6084
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.795644216076436e-05,
+      "loss": 0.3301,
+      "step": 6085
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7956106325457996e-05,
+      "loss": 0.3244,
+      "step": 6086
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.795577049015163e-05,
+      "loss": 0.4501,
+      "step": 6087
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7955434654845265e-05,
+      "loss": 0.433,
+      "step": 6088
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.79550988195389e-05,
+      "loss": 0.2239,
+      "step": 6089
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7954762984232534e-05,
+      "loss": 0.3338,
+      "step": 6090
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.795442714892617e-05,
+      "loss": 0.2807,
+      "step": 6091
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.79540913136198e-05,
+      "loss": 0.351,
+      "step": 6092
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7953755478313438e-05,
+      "loss": 0.2759,
+      "step": 6093
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.795341964300707e-05,
+      "loss": 0.2515,
+      "step": 6094
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7953083807700704e-05,
+      "loss": 0.2809,
+      "step": 6095
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.795274797239434e-05,
+      "loss": 0.5455,
+      "step": 6096
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7952412137087973e-05,
+      "loss": 0.4155,
+      "step": 6097
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7952076301781608e-05,
+      "loss": 0.4153,
+      "step": 6098
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7951740466475243e-05,
+      "loss": 0.2533,
+      "step": 6099
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7951404631168877e-05,
+      "loss": 0.503,
+      "step": 6100
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.795106879586251e-05,
+      "loss": 0.3874,
+      "step": 6101
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7950732960556147e-05,
+      "loss": 0.5017,
+      "step": 6102
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7950397125249778e-05,
+      "loss": 0.3062,
+      "step": 6103
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7950061289943413e-05,
+      "loss": 0.4133,
+      "step": 6104
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7949725454637047e-05,
+      "loss": 0.5092,
+      "step": 6105
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7949389619330682e-05,
+      "loss": 0.3943,
+      "step": 6106
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7949053784024316e-05,
+      "loss": 0.3139,
+      "step": 6107
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.794871794871795e-05,
+      "loss": 0.4251,
+      "step": 6108
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7948382113411586e-05,
+      "loss": 0.4092,
+      "step": 6109
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7948046278105217e-05,
+      "loss": 0.2639,
+      "step": 6110
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7947710442798855e-05,
+      "loss": 0.3013,
+      "step": 6111
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7947374607492486e-05,
+      "loss": 0.3599,
+      "step": 6112
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.794703877218612e-05,
+      "loss": 0.3602,
+      "step": 6113
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7946702936879756e-05,
+      "loss": 0.4313,
+      "step": 6114
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.794636710157339e-05,
+      "loss": 0.4385,
+      "step": 6115
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7946031266267025e-05,
+      "loss": 0.3228,
+      "step": 6116
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.794569543096066e-05,
+      "loss": 0.4214,
+      "step": 6117
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.794535959565429e-05,
+      "loss": 0.3624,
+      "step": 6118
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7945023760347925e-05,
+      "loss": 0.496,
+      "step": 6119
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.794468792504156e-05,
+      "loss": 0.4122,
+      "step": 6120
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7944352089735195e-05,
+      "loss": 0.3386,
+      "step": 6121
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.794401625442883e-05,
+      "loss": 0.3263,
+      "step": 6122
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7943680419122464e-05,
+      "loss": 0.3711,
+      "step": 6123
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.79433445838161e-05,
+      "loss": 0.2223,
+      "step": 6124
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.794300874850973e-05,
+      "loss": 0.2872,
+      "step": 6125
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7942672913203368e-05,
+      "loss": 0.4029,
+      "step": 6126
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7942337077897e-05,
+      "loss": 0.2963,
+      "step": 6127
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7942001242590634e-05,
+      "loss": 0.3665,
+      "step": 6128
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.794166540728427e-05,
+      "loss": 0.2371,
+      "step": 6129
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7941329571977903e-05,
+      "loss": 0.2941,
+      "step": 6130
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7940993736671538e-05,
+      "loss": 0.3129,
+      "step": 6131
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7940657901365173e-05,
+      "loss": 0.2961,
+      "step": 6132
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7940322066058807e-05,
+      "loss": 0.2503,
+      "step": 6133
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.793998623075244e-05,
+      "loss": 0.5556,
+      "step": 6134
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7939650395446076e-05,
+      "loss": 0.4415,
+      "step": 6135
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7939314560139708e-05,
+      "loss": 0.5127,
+      "step": 6136
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7938978724833342e-05,
+      "loss": 0.4574,
+      "step": 6137
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7938642889526977e-05,
+      "loss": 0.313,
+      "step": 6138
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.793830705422061e-05,
+      "loss": 0.4221,
+      "step": 6139
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7937971218914246e-05,
+      "loss": 0.237,
+      "step": 6140
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.793763538360788e-05,
+      "loss": 0.3004,
+      "step": 6141
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7937299548301516e-05,
+      "loss": 0.2992,
+      "step": 6142
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7936963712995147e-05,
+      "loss": 0.3099,
+      "step": 6143
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7936627877688785e-05,
+      "loss": 0.2712,
+      "step": 6144
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7936292042382416e-05,
+      "loss": 0.2831,
+      "step": 6145
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.793595620707605e-05,
+      "loss": 0.2703,
+      "step": 6146
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7935620371769685e-05,
+      "loss": 0.3901,
+      "step": 6147
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.793528453646332e-05,
+      "loss": 0.5589,
+      "step": 6148
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7934948701156955e-05,
+      "loss": 0.4464,
+      "step": 6149
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.793461286585059e-05,
+      "loss": 0.4566,
+      "step": 6150
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7934277030544224e-05,
+      "loss": 0.2698,
+      "step": 6151
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7933941195237855e-05,
+      "loss": 0.2996,
+      "step": 6152
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7933605359931493e-05,
+      "loss": 0.3809,
+      "step": 6153
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7933269524625125e-05,
+      "loss": 0.2812,
+      "step": 6154
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.793293368931876e-05,
+      "loss": 0.326,
+      "step": 6155
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7932597854012394e-05,
+      "loss": 0.2355,
+      "step": 6156
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.793226201870603e-05,
+      "loss": 0.2698,
+      "step": 6157
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7931926183399663e-05,
+      "loss": 0.3519,
+      "step": 6158
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7931590348093298e-05,
+      "loss": 0.4157,
+      "step": 6159
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7931254512786932e-05,
+      "loss": 0.2678,
+      "step": 6160
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7930918677480564e-05,
+      "loss": 0.2251,
+      "step": 6161
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7930582842174202e-05,
+      "loss": 0.2352,
+      "step": 6162
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7930247006867833e-05,
+      "loss": 0.3557,
+      "step": 6163
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7929911171561468e-05,
+      "loss": 0.2196,
+      "step": 6164
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7929575336255102e-05,
+      "loss": 0.2828,
+      "step": 6165
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7929239500948737e-05,
+      "loss": 0.3223,
+      "step": 6166
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7928903665642368e-05,
+      "loss": 0.4639,
+      "step": 6167
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7928567830336006e-05,
+      "loss": 0.1923,
+      "step": 6168
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7928231995029638e-05,
+      "loss": 0.3088,
+      "step": 6169
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7927896159723272e-05,
+      "loss": 0.4592,
+      "step": 6170
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7927560324416907e-05,
+      "loss": 0.2235,
+      "step": 6171
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.792722448911054e-05,
+      "loss": 0.1252,
+      "step": 6172
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7926888653804176e-05,
+      "loss": 0.3943,
+      "step": 6173
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.792655281849781e-05,
+      "loss": 0.3019,
+      "step": 6174
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7926216983191445e-05,
+      "loss": 0.2238,
+      "step": 6175
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7925881147885077e-05,
+      "loss": 0.2405,
+      "step": 6176
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7925545312578715e-05,
+      "loss": 0.3975,
+      "step": 6177
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7925209477272346e-05,
+      "loss": 0.4842,
+      "step": 6178
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.792487364196598e-05,
+      "loss": 0.3561,
+      "step": 6179
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7924537806659615e-05,
+      "loss": 0.5485,
+      "step": 6180
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.792420197135325e-05,
+      "loss": 0.269,
+      "step": 6181
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7923866136046885e-05,
+      "loss": 0.3888,
+      "step": 6182
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.792353030074052e-05,
+      "loss": 0.4533,
+      "step": 6183
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7923194465434154e-05,
+      "loss": 0.3662,
+      "step": 6184
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7922858630127785e-05,
+      "loss": 0.3397,
+      "step": 6185
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7922522794821423e-05,
+      "loss": 0.3989,
+      "step": 6186
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7922186959515054e-05,
+      "loss": 0.5004,
+      "step": 6187
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.792185112420869e-05,
+      "loss": 0.3427,
+      "step": 6188
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7921515288902324e-05,
+      "loss": 0.3201,
+      "step": 6189
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.792117945359596e-05,
+      "loss": 0.2879,
+      "step": 6190
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7920843618289593e-05,
+      "loss": 0.4724,
+      "step": 6191
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7920507782983228e-05,
+      "loss": 0.2326,
+      "step": 6192
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7920171947676862e-05,
+      "loss": 0.3132,
+      "step": 6193
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7919836112370494e-05,
+      "loss": 0.4143,
+      "step": 6194
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.791950027706413e-05,
+      "loss": 0.1673,
+      "step": 6195
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7919164441757763e-05,
+      "loss": 0.2911,
+      "step": 6196
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7918828606451397e-05,
+      "loss": 0.226,
+      "step": 6197
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7918492771145032e-05,
+      "loss": 0.3621,
+      "step": 6198
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7918156935838667e-05,
+      "loss": 0.3054,
+      "step": 6199
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.79178211005323e-05,
+      "loss": 0.4316,
+      "step": 6200
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7917485265225936e-05,
+      "loss": 0.4775,
+      "step": 6201
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.791714942991957e-05,
+      "loss": 0.2948,
+      "step": 6202
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7916813594613202e-05,
+      "loss": 0.5554,
+      "step": 6203
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.791647775930684e-05,
+      "loss": 0.4073,
+      "step": 6204
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.791614192400047e-05,
+      "loss": 0.3331,
+      "step": 6205
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7915806088694106e-05,
+      "loss": 0.2991,
+      "step": 6206
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.791547025338774e-05,
+      "loss": 0.1976,
+      "step": 6207
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7915134418081375e-05,
+      "loss": 0.4126,
+      "step": 6208
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7914798582775006e-05,
+      "loss": 0.193,
+      "step": 6209
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7914462747468645e-05,
+      "loss": 0.2609,
+      "step": 6210
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7914126912162276e-05,
+      "loss": 0.271,
+      "step": 6211
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.791379107685591e-05,
+      "loss": 0.2334,
+      "step": 6212
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7913455241549545e-05,
+      "loss": 0.5256,
+      "step": 6213
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.791311940624318e-05,
+      "loss": 0.3673,
+      "step": 6214
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7912783570936814e-05,
+      "loss": 0.2878,
+      "step": 6215
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.791244773563045e-05,
+      "loss": 0.4168,
+      "step": 6216
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7912111900324084e-05,
+      "loss": 0.3162,
+      "step": 6217
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7911776065017715e-05,
+      "loss": 0.3894,
+      "step": 6218
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7911440229711353e-05,
+      "loss": 0.361,
+      "step": 6219
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7911104394404984e-05,
+      "loss": 0.3348,
+      "step": 6220
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.791076855909862e-05,
+      "loss": 0.2774,
+      "step": 6221
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7910432723792254e-05,
+      "loss": 0.2243,
+      "step": 6222
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7910096888485888e-05,
+      "loss": 0.4422,
+      "step": 6223
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7909761053179523e-05,
+      "loss": 0.2152,
+      "step": 6224
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7909425217873157e-05,
+      "loss": 0.3489,
+      "step": 6225
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7909089382566792e-05,
+      "loss": 0.3157,
+      "step": 6226
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7908753547260423e-05,
+      "loss": 0.523,
+      "step": 6227
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7908417711954058e-05,
+      "loss": 0.3999,
+      "step": 6228
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7908081876647693e-05,
+      "loss": 0.3103,
+      "step": 6229
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7907746041341327e-05,
+      "loss": 0.2812,
+      "step": 6230
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7907410206034962e-05,
+      "loss": 0.5315,
+      "step": 6231
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7907074370728597e-05,
+      "loss": 0.3484,
+      "step": 6232
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.790673853542223e-05,
+      "loss": 0.3221,
+      "step": 6233
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7906402700115863e-05,
+      "loss": 0.4174,
+      "step": 6234
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.79060668648095e-05,
+      "loss": 0.2806,
+      "step": 6235
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7905731029503132e-05,
+      "loss": 0.2993,
+      "step": 6236
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7905395194196766e-05,
+      "loss": 0.3961,
+      "step": 6237
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.79050593588904e-05,
+      "loss": 0.266,
+      "step": 6238
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7904723523584036e-05,
+      "loss": 0.41,
+      "step": 6239
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.790438768827767e-05,
+      "loss": 0.4125,
+      "step": 6240
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7904051852971305e-05,
+      "loss": 0.2923,
+      "step": 6241
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.790371601766494e-05,
+      "loss": 0.4006,
+      "step": 6242
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.790338018235857e-05,
+      "loss": 0.2893,
+      "step": 6243
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.790304434705221e-05,
+      "loss": 0.2744,
+      "step": 6244
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.790270851174584e-05,
+      "loss": 0.3356,
+      "step": 6245
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7902372676439475e-05,
+      "loss": 0.219,
+      "step": 6246
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.790203684113311e-05,
+      "loss": 0.3733,
+      "step": 6247
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7901701005826744e-05,
+      "loss": 0.3589,
+      "step": 6248
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.790136517052038e-05,
+      "loss": 0.2943,
+      "step": 6249
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7901029335214013e-05,
+      "loss": 0.4391,
+      "step": 6250
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7900693499907648e-05,
+      "loss": 0.2864,
+      "step": 6251
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.790035766460128e-05,
+      "loss": 0.4661,
+      "step": 6252
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7900021829294914e-05,
+      "loss": 0.4102,
+      "step": 6253
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.789968599398855e-05,
+      "loss": 0.329,
+      "step": 6254
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7899350158682183e-05,
+      "loss": 0.2638,
+      "step": 6255
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7899014323375818e-05,
+      "loss": 0.4185,
+      "step": 6256
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7898678488069453e-05,
+      "loss": 0.3269,
+      "step": 6257
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7898342652763084e-05,
+      "loss": 0.4963,
+      "step": 6258
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7898006817456722e-05,
+      "loss": 0.4247,
+      "step": 6259
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7897670982150353e-05,
+      "loss": 0.3336,
+      "step": 6260
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7897335146843988e-05,
+      "loss": 0.4195,
+      "step": 6261
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7896999311537622e-05,
+      "loss": 0.3437,
+      "step": 6262
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7896663476231257e-05,
+      "loss": 0.3244,
+      "step": 6263
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7896327640924892e-05,
+      "loss": 0.3103,
+      "step": 6264
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7895991805618526e-05,
+      "loss": 0.3186,
+      "step": 6265
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.789565597031216e-05,
+      "loss": 0.2258,
+      "step": 6266
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7895320135005792e-05,
+      "loss": 0.4373,
+      "step": 6267
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.789498429969943e-05,
+      "loss": 0.2798,
+      "step": 6268
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.789464846439306e-05,
+      "loss": 0.3326,
+      "step": 6269
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7894312629086696e-05,
+      "loss": 0.4077,
+      "step": 6270
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.789397679378033e-05,
+      "loss": 0.2504,
+      "step": 6271
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7893640958473966e-05,
+      "loss": 0.2309,
+      "step": 6272
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.78933051231676e-05,
+      "loss": 0.3998,
+      "step": 6273
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7892969287861235e-05,
+      "loss": 0.2839,
+      "step": 6274
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.789263345255487e-05,
+      "loss": 0.227,
+      "step": 6275
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.78922976172485e-05,
+      "loss": 0.3173,
+      "step": 6276
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.789196178194214e-05,
+      "loss": 0.2937,
+      "step": 6277
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.789162594663577e-05,
+      "loss": 0.2049,
+      "step": 6278
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7891290111329405e-05,
+      "loss": 0.3133,
+      "step": 6279
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.789095427602304e-05,
+      "loss": 0.3623,
+      "step": 6280
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7890618440716674e-05,
+      "loss": 0.3568,
+      "step": 6281
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.789028260541031e-05,
+      "loss": 0.3801,
+      "step": 6282
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7889946770103943e-05,
+      "loss": 0.2911,
+      "step": 6283
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7889610934797578e-05,
+      "loss": 0.333,
+      "step": 6284
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.788927509949121e-05,
+      "loss": 0.2791,
+      "step": 6285
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7888939264184847e-05,
+      "loss": 0.1704,
+      "step": 6286
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.788860342887848e-05,
+      "loss": 0.2413,
+      "step": 6287
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7888267593572113e-05,
+      "loss": 0.235,
+      "step": 6288
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7887931758265748e-05,
+      "loss": 0.3759,
+      "step": 6289
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7887595922959382e-05,
+      "loss": 0.2972,
+      "step": 6290
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7887260087653017e-05,
+      "loss": 0.2535,
+      "step": 6291
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7886924252346652e-05,
+      "loss": 0.2371,
+      "step": 6292
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7886588417040286e-05,
+      "loss": 0.188,
+      "step": 6293
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7886252581733918e-05,
+      "loss": 0.3097,
+      "step": 6294
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7885916746427556e-05,
+      "loss": 0.3468,
+      "step": 6295
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7885580911121187e-05,
+      "loss": 0.3884,
+      "step": 6296
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.788524507581482e-05,
+      "loss": 0.3079,
+      "step": 6297
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7884909240508456e-05,
+      "loss": 0.4469,
+      "step": 6298
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.788457340520209e-05,
+      "loss": 0.1849,
+      "step": 6299
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7884237569895722e-05,
+      "loss": 0.3621,
+      "step": 6300
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.788390173458936e-05,
+      "loss": 0.2524,
+      "step": 6301
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.788356589928299e-05,
+      "loss": 0.2649,
+      "step": 6302
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7883230063976626e-05,
+      "loss": 0.5337,
+      "step": 6303
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.788289422867026e-05,
+      "loss": 0.2817,
+      "step": 6304
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7882558393363895e-05,
+      "loss": 0.2136,
+      "step": 6305
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.788222255805753e-05,
+      "loss": 0.4312,
+      "step": 6306
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7881886722751165e-05,
+      "loss": 0.248,
+      "step": 6307
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.78815508874448e-05,
+      "loss": 0.2878,
+      "step": 6308
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.788121505213843e-05,
+      "loss": 0.4047,
+      "step": 6309
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.788087921683207e-05,
+      "loss": 0.3491,
+      "step": 6310
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.78805433815257e-05,
+      "loss": 0.4136,
+      "step": 6311
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7880207546219335e-05,
+      "loss": 0.4424,
+      "step": 6312
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.787987171091297e-05,
+      "loss": 0.2203,
+      "step": 6313
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7879535875606604e-05,
+      "loss": 0.2623,
+      "step": 6314
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.787920004030024e-05,
+      "loss": 0.2579,
+      "step": 6315
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7878864204993873e-05,
+      "loss": 0.3076,
+      "step": 6316
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7878528369687508e-05,
+      "loss": 0.2755,
+      "step": 6317
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.787819253438114e-05,
+      "loss": 0.384,
+      "step": 6318
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7877856699074777e-05,
+      "loss": 0.3105,
+      "step": 6319
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7877520863768408e-05,
+      "loss": 0.4188,
+      "step": 6320
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7877185028462043e-05,
+      "loss": 0.4609,
+      "step": 6321
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7876849193155678e-05,
+      "loss": 0.3598,
+      "step": 6322
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7876513357849312e-05,
+      "loss": 0.4177,
+      "step": 6323
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7876177522542947e-05,
+      "loss": 0.3576,
+      "step": 6324
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.787584168723658e-05,
+      "loss": 0.4055,
+      "step": 6325
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7875505851930216e-05,
+      "loss": 0.2064,
+      "step": 6326
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7875170016623847e-05,
+      "loss": 0.3683,
+      "step": 6327
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7874834181317485e-05,
+      "loss": 0.2857,
+      "step": 6328
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7874498346011117e-05,
+      "loss": 0.3141,
+      "step": 6329
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.787416251070475e-05,
+      "loss": 0.4231,
+      "step": 6330
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7873826675398386e-05,
+      "loss": 0.3537,
+      "step": 6331
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.787349084009202e-05,
+      "loss": 0.311,
+      "step": 6332
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7873155004785655e-05,
+      "loss": 0.3988,
+      "step": 6333
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.787281916947929e-05,
+      "loss": 0.3662,
+      "step": 6334
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7872483334172925e-05,
+      "loss": 0.3479,
+      "step": 6335
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7872147498866556e-05,
+      "loss": 0.381,
+      "step": 6336
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7871811663560194e-05,
+      "loss": 0.2472,
+      "step": 6337
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7871475828253825e-05,
+      "loss": 0.3978,
+      "step": 6338
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.787113999294746e-05,
+      "loss": 0.4257,
+      "step": 6339
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7870804157641094e-05,
+      "loss": 0.4648,
+      "step": 6340
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.787046832233473e-05,
+      "loss": 0.2618,
+      "step": 6341
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7870132487028364e-05,
+      "loss": 0.2959,
+      "step": 6342
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7869796651722e-05,
+      "loss": 0.3987,
+      "step": 6343
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.786946081641563e-05,
+      "loss": 0.272,
+      "step": 6344
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7869124981109264e-05,
+      "loss": 0.3377,
+      "step": 6345
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.78687891458029e-05,
+      "loss": 0.4446,
+      "step": 6346
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7868453310496534e-05,
+      "loss": 0.2541,
+      "step": 6347
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7868117475190168e-05,
+      "loss": 0.2971,
+      "step": 6348
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7867781639883803e-05,
+      "loss": 0.39,
+      "step": 6349
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7867445804577438e-05,
+      "loss": 0.3391,
+      "step": 6350
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.786710996927107e-05,
+      "loss": 0.3616,
+      "step": 6351
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7866774133964707e-05,
+      "loss": 0.3248,
+      "step": 6352
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7866438298658338e-05,
+      "loss": 0.3047,
+      "step": 6353
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7866102463351973e-05,
+      "loss": 0.1871,
+      "step": 6354
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7865766628045607e-05,
+      "loss": 0.3393,
+      "step": 6355
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7865430792739242e-05,
+      "loss": 0.2959,
+      "step": 6356
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7865094957432877e-05,
+      "loss": 0.5104,
+      "step": 6357
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.786475912212651e-05,
+      "loss": 0.341,
+      "step": 6358
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7864423286820146e-05,
+      "loss": 0.4215,
+      "step": 6359
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7864087451513777e-05,
+      "loss": 0.2619,
+      "step": 6360
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7863751616207415e-05,
+      "loss": 0.3787,
+      "step": 6361
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7863415780901047e-05,
+      "loss": 0.3458,
+      "step": 6362
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.786307994559468e-05,
+      "loss": 0.2911,
+      "step": 6363
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7862744110288316e-05,
+      "loss": 0.3398,
+      "step": 6364
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.786240827498195e-05,
+      "loss": 0.2884,
+      "step": 6365
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7862072439675585e-05,
+      "loss": 0.378,
+      "step": 6366
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.786173660436922e-05,
+      "loss": 0.4653,
+      "step": 6367
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7861400769062854e-05,
+      "loss": 0.2176,
+      "step": 6368
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7861064933756486e-05,
+      "loss": 0.3152,
+      "step": 6369
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7860729098450124e-05,
+      "loss": 0.2617,
+      "step": 6370
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7860393263143755e-05,
+      "loss": 0.3018,
+      "step": 6371
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.786005742783739e-05,
+      "loss": 0.2932,
+      "step": 6372
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7859721592531024e-05,
+      "loss": 0.2541,
+      "step": 6373
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.785938575722466e-05,
+      "loss": 0.4241,
+      "step": 6374
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7859049921918294e-05,
+      "loss": 0.3453,
+      "step": 6375
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7858714086611928e-05,
+      "loss": 0.5472,
+      "step": 6376
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7858378251305563e-05,
+      "loss": 0.2922,
+      "step": 6377
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7858042415999194e-05,
+      "loss": 0.3431,
+      "step": 6378
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7857706580692832e-05,
+      "loss": 0.2921,
+      "step": 6379
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7857370745386463e-05,
+      "loss": 0.3688,
+      "step": 6380
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7857034910080098e-05,
+      "loss": 0.3448,
+      "step": 6381
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7856699074773733e-05,
+      "loss": 0.3092,
+      "step": 6382
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7856363239467367e-05,
+      "loss": 0.3494,
+      "step": 6383
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7856027404161002e-05,
+      "loss": 0.2741,
+      "step": 6384
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7855691568854637e-05,
+      "loss": 0.3995,
+      "step": 6385
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.785535573354827e-05,
+      "loss": 0.3234,
+      "step": 6386
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7855019898241903e-05,
+      "loss": 0.2835,
+      "step": 6387
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.785468406293554e-05,
+      "loss": 0.2959,
+      "step": 6388
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7854348227629172e-05,
+      "loss": 0.3686,
+      "step": 6389
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7854012392322807e-05,
+      "loss": 0.3872,
+      "step": 6390
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.785367655701644e-05,
+      "loss": 0.4604,
+      "step": 6391
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7853340721710076e-05,
+      "loss": 0.3934,
+      "step": 6392
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7853004886403707e-05,
+      "loss": 0.3925,
+      "step": 6393
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7852669051097345e-05,
+      "loss": 0.2532,
+      "step": 6394
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7852333215790976e-05,
+      "loss": 0.3883,
+      "step": 6395
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.785199738048461e-05,
+      "loss": 0.2769,
+      "step": 6396
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7851661545178246e-05,
+      "loss": 0.2126,
+      "step": 6397
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.785132570987188e-05,
+      "loss": 0.2732,
+      "step": 6398
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7850989874565515e-05,
+      "loss": 0.519,
+      "step": 6399
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.785065403925915e-05,
+      "loss": 0.2776,
+      "step": 6400
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7850318203952784e-05,
+      "loss": 0.4847,
+      "step": 6401
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7849982368646416e-05,
+      "loss": 0.2302,
+      "step": 6402
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7849646533340054e-05,
+      "loss": 0.4507,
+      "step": 6403
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7849310698033685e-05,
+      "loss": 0.358,
+      "step": 6404
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.784897486272732e-05,
+      "loss": 0.3155,
+      "step": 6405
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7848639027420954e-05,
+      "loss": 0.3903,
+      "step": 6406
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.784830319211459e-05,
+      "loss": 0.3244,
+      "step": 6407
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7847967356808223e-05,
+      "loss": 0.3214,
+      "step": 6408
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7847631521501855e-05,
+      "loss": 0.2774,
+      "step": 6409
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7847295686195493e-05,
+      "loss": 0.2694,
+      "step": 6410
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7846959850889124e-05,
+      "loss": 0.3085,
+      "step": 6411
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.784662401558276e-05,
+      "loss": 0.2207,
+      "step": 6412
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7846288180276393e-05,
+      "loss": 0.2777,
+      "step": 6413
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7845952344970028e-05,
+      "loss": 0.1968,
+      "step": 6414
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7845616509663663e-05,
+      "loss": 0.2639,
+      "step": 6415
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7845280674357297e-05,
+      "loss": 0.2644,
+      "step": 6416
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7844944839050932e-05,
+      "loss": 0.1376,
+      "step": 6417
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7844609003744563e-05,
+      "loss": 0.3001,
+      "step": 6418
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.78442731684382e-05,
+      "loss": 0.4655,
+      "step": 6419
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7843937333131832e-05,
+      "loss": 0.2528,
+      "step": 6420
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7843601497825467e-05,
+      "loss": 0.3377,
+      "step": 6421
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.78432656625191e-05,
+      "loss": 0.1489,
+      "step": 6422
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7842929827212736e-05,
+      "loss": 0.544,
+      "step": 6423
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.784259399190637e-05,
+      "loss": 0.4401,
+      "step": 6424
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7842258156600006e-05,
+      "loss": 0.3397,
+      "step": 6425
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.784192232129364e-05,
+      "loss": 0.3522,
+      "step": 6426
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.784158648598727e-05,
+      "loss": 0.3631,
+      "step": 6427
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.784125065068091e-05,
+      "loss": 0.3761,
+      "step": 6428
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.784091481537454e-05,
+      "loss": 0.3528,
+      "step": 6429
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7840578980068175e-05,
+      "loss": 0.2098,
+      "step": 6430
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.784024314476181e-05,
+      "loss": 0.3652,
+      "step": 6431
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7839907309455445e-05,
+      "loss": 0.4464,
+      "step": 6432
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.783957147414908e-05,
+      "loss": 0.2164,
+      "step": 6433
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7839235638842714e-05,
+      "loss": 0.2954,
+      "step": 6434
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7838899803536345e-05,
+      "loss": 0.2857,
+      "step": 6435
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.783856396822998e-05,
+      "loss": 0.3,
+      "step": 6436
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7838228132923615e-05,
+      "loss": 0.4609,
+      "step": 6437
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.783789229761725e-05,
+      "loss": 0.3079,
+      "step": 6438
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7837556462310884e-05,
+      "loss": 0.2971,
+      "step": 6439
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.783722062700452e-05,
+      "loss": 0.3577,
+      "step": 6440
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7836884791698153e-05,
+      "loss": 0.239,
+      "step": 6441
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7836548956391784e-05,
+      "loss": 0.2404,
+      "step": 6442
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7836213121085422e-05,
+      "loss": 0.4181,
+      "step": 6443
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7835877285779054e-05,
+      "loss": 0.3985,
+      "step": 6444
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.783554145047269e-05,
+      "loss": 0.261,
+      "step": 6445
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7835205615166323e-05,
+      "loss": 0.3927,
+      "step": 6446
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7834869779859958e-05,
+      "loss": 0.4406,
+      "step": 6447
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7834533944553592e-05,
+      "loss": 0.3377,
+      "step": 6448
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7834198109247227e-05,
+      "loss": 0.3148,
+      "step": 6449
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.783386227394086e-05,
+      "loss": 0.3984,
+      "step": 6450
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7833526438634493e-05,
+      "loss": 0.4131,
+      "step": 6451
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.783319060332813e-05,
+      "loss": 0.1741,
+      "step": 6452
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7832854768021762e-05,
+      "loss": 0.2502,
+      "step": 6453
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7832518932715397e-05,
+      "loss": 0.4725,
+      "step": 6454
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.783218309740903e-05,
+      "loss": 0.3148,
+      "step": 6455
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7831847262102666e-05,
+      "loss": 0.4341,
+      "step": 6456
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.78315114267963e-05,
+      "loss": 0.304,
+      "step": 6457
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7831175591489935e-05,
+      "loss": 0.414,
+      "step": 6458
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.783083975618357e-05,
+      "loss": 0.4382,
+      "step": 6459
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.78305039208772e-05,
+      "loss": 0.2943,
+      "step": 6460
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.783016808557084e-05,
+      "loss": 0.2853,
+      "step": 6461
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.782983225026447e-05,
+      "loss": 0.2481,
+      "step": 6462
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7829496414958105e-05,
+      "loss": 0.4588,
+      "step": 6463
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.782916057965174e-05,
+      "loss": 0.2532,
+      "step": 6464
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7828824744345375e-05,
+      "loss": 0.2521,
+      "step": 6465
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.782848890903901e-05,
+      "loss": 0.3182,
+      "step": 6466
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7828153073732644e-05,
+      "loss": 0.4701,
+      "step": 6467
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.782781723842628e-05,
+      "loss": 0.3462,
+      "step": 6468
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.782748140311991e-05,
+      "loss": 0.3608,
+      "step": 6469
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7827145567813548e-05,
+      "loss": 0.3918,
+      "step": 6470
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.782680973250718e-05,
+      "loss": 0.3087,
+      "step": 6471
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7826473897200814e-05,
+      "loss": 0.186,
+      "step": 6472
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.782613806189445e-05,
+      "loss": 0.3111,
+      "step": 6473
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7825802226588083e-05,
+      "loss": 0.3971,
+      "step": 6474
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7825466391281718e-05,
+      "loss": 0.2441,
+      "step": 6475
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7825130555975352e-05,
+      "loss": 0.379,
+      "step": 6476
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7824794720668987e-05,
+      "loss": 0.4364,
+      "step": 6477
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7824458885362618e-05,
+      "loss": 0.5361,
+      "step": 6478
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7824123050056253e-05,
+      "loss": 0.3158,
+      "step": 6479
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7823787214749888e-05,
+      "loss": 0.337,
+      "step": 6480
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7823451379443522e-05,
+      "loss": 0.3102,
+      "step": 6481
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7823115544137157e-05,
+      "loss": 0.2243,
+      "step": 6482
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.782277970883079e-05,
+      "loss": 0.566,
+      "step": 6483
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7822443873524423e-05,
+      "loss": 0.1839,
+      "step": 6484
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.782210803821806e-05,
+      "loss": 0.4749,
+      "step": 6485
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7821772202911692e-05,
+      "loss": 0.3501,
+      "step": 6486
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7821436367605327e-05,
+      "loss": 0.2702,
+      "step": 6487
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.782110053229896e-05,
+      "loss": 0.3104,
+      "step": 6488
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7820764696992596e-05,
+      "loss": 0.441,
+      "step": 6489
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.782042886168623e-05,
+      "loss": 0.2407,
+      "step": 6490
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7820093026379865e-05,
+      "loss": 0.2865,
+      "step": 6491
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.78197571910735e-05,
+      "loss": 0.3013,
+      "step": 6492
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.781942135576713e-05,
+      "loss": 0.4335,
+      "step": 6493
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.781908552046077e-05,
+      "loss": 0.2891,
+      "step": 6494
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.78187496851544e-05,
+      "loss": 0.283,
+      "step": 6495
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7818413849848035e-05,
+      "loss": 0.3664,
+      "step": 6496
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.781807801454167e-05,
+      "loss": 0.3618,
+      "step": 6497
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7817742179235304e-05,
+      "loss": 0.2876,
+      "step": 6498
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.781740634392894e-05,
+      "loss": 0.2267,
+      "step": 6499
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7817070508622574e-05,
+      "loss": 0.4894,
+      "step": 6500
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.781673467331621e-05,
+      "loss": 0.2508,
+      "step": 6501
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.781639883800984e-05,
+      "loss": 0.3285,
+      "step": 6502
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7816063002703478e-05,
+      "loss": 0.319,
+      "step": 6503
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.781572716739711e-05,
+      "loss": 0.3762,
+      "step": 6504
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7815391332090744e-05,
+      "loss": 0.4191,
+      "step": 6505
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7815055496784378e-05,
+      "loss": 0.2747,
+      "step": 6506
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7814719661478013e-05,
+      "loss": 0.3779,
+      "step": 6507
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7814383826171647e-05,
+      "loss": 0.2926,
+      "step": 6508
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7814047990865282e-05,
+      "loss": 0.3395,
+      "step": 6509
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7813712155558917e-05,
+      "loss": 0.2578,
+      "step": 6510
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7813376320252548e-05,
+      "loss": 0.2975,
+      "step": 6511
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7813040484946186e-05,
+      "loss": 0.2765,
+      "step": 6512
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7812704649639817e-05,
+      "loss": 0.318,
+      "step": 6513
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7812368814333452e-05,
+      "loss": 0.6924,
+      "step": 6514
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7812032979027087e-05,
+      "loss": 0.3396,
+      "step": 6515
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.781169714372072e-05,
+      "loss": 0.3978,
+      "step": 6516
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7811361308414356e-05,
+      "loss": 0.297,
+      "step": 6517
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.781102547310799e-05,
+      "loss": 0.4151,
+      "step": 6518
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7810689637801625e-05,
+      "loss": 0.3977,
+      "step": 6519
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7810353802495256e-05,
+      "loss": 0.4003,
+      "step": 6520
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7810017967188894e-05,
+      "loss": 0.2715,
+      "step": 6521
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7809682131882526e-05,
+      "loss": 0.2389,
+      "step": 6522
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.780934629657616e-05,
+      "loss": 0.3047,
+      "step": 6523
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7809010461269795e-05,
+      "loss": 0.2592,
+      "step": 6524
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.780867462596343e-05,
+      "loss": 0.4771,
+      "step": 6525
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.780833879065706e-05,
+      "loss": 0.2523,
+      "step": 6526
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.78080029553507e-05,
+      "loss": 0.317,
+      "step": 6527
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.780766712004433e-05,
+      "loss": 0.2842,
+      "step": 6528
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7807331284737965e-05,
+      "loss": 0.5117,
+      "step": 6529
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.78069954494316e-05,
+      "loss": 0.4745,
+      "step": 6530
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7806659614125234e-05,
+      "loss": 0.3483,
+      "step": 6531
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.780632377881887e-05,
+      "loss": 0.2713,
+      "step": 6532
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7805987943512503e-05,
+      "loss": 0.363,
+      "step": 6533
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7805652108206138e-05,
+      "loss": 0.3342,
+      "step": 6534
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.780531627289977e-05,
+      "loss": 0.3482,
+      "step": 6535
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7804980437593407e-05,
+      "loss": 0.4286,
+      "step": 6536
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.780464460228704e-05,
+      "loss": 0.3557,
+      "step": 6537
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7804308766980673e-05,
+      "loss": 0.4666,
+      "step": 6538
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7803972931674308e-05,
+      "loss": 0.3788,
+      "step": 6539
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7803637096367943e-05,
+      "loss": 0.4436,
+      "step": 6540
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7803301261061577e-05,
+      "loss": 0.3233,
+      "step": 6541
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7802965425755212e-05,
+      "loss": 0.2812,
+      "step": 6542
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7802629590448847e-05,
+      "loss": 0.2582,
+      "step": 6543
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7802293755142478e-05,
+      "loss": 0.3129,
+      "step": 6544
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7801957919836116e-05,
+      "loss": 0.4009,
+      "step": 6545
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7801622084529747e-05,
+      "loss": 0.4236,
+      "step": 6546
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7801286249223382e-05,
+      "loss": 0.4262,
+      "step": 6547
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7800950413917016e-05,
+      "loss": 0.2268,
+      "step": 6548
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.780061457861065e-05,
+      "loss": 0.4109,
+      "step": 6549
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7800278743304286e-05,
+      "loss": 0.1308,
+      "step": 6550
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.779994290799792e-05,
+      "loss": 0.402,
+      "step": 6551
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7799607072691555e-05,
+      "loss": 0.4789,
+      "step": 6552
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7799271237385186e-05,
+      "loss": 0.2205,
+      "step": 6553
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7798935402078824e-05,
+      "loss": 0.2189,
+      "step": 6554
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7798599566772456e-05,
+      "loss": 0.2211,
+      "step": 6555
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.779826373146609e-05,
+      "loss": 0.3361,
+      "step": 6556
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7797927896159725e-05,
+      "loss": 0.2072,
+      "step": 6557
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.779759206085336e-05,
+      "loss": 0.183,
+      "step": 6558
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7797256225546994e-05,
+      "loss": 0.2656,
+      "step": 6559
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.779692039024063e-05,
+      "loss": 0.2085,
+      "step": 6560
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7796584554934263e-05,
+      "loss": 0.4017,
+      "step": 6561
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7796248719627895e-05,
+      "loss": 0.3715,
+      "step": 6562
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7795912884321533e-05,
+      "loss": 0.3909,
+      "step": 6563
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7795577049015164e-05,
+      "loss": 0.4909,
+      "step": 6564
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.77952412137088e-05,
+      "loss": 0.2751,
+      "step": 6565
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7794905378402433e-05,
+      "loss": 0.3398,
+      "step": 6566
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7794569543096068e-05,
+      "loss": 0.2554,
+      "step": 6567
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7794233707789703e-05,
+      "loss": 0.2611,
+      "step": 6568
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7793897872483337e-05,
+      "loss": 0.2671,
+      "step": 6569
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.779356203717697e-05,
+      "loss": 0.4073,
+      "step": 6570
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7793226201870603e-05,
+      "loss": 0.1661,
+      "step": 6571
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7792890366564238e-05,
+      "loss": 0.2696,
+      "step": 6572
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7792554531257872e-05,
+      "loss": 0.3337,
+      "step": 6573
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7792218695951507e-05,
+      "loss": 0.3792,
+      "step": 6574
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7791882860645142e-05,
+      "loss": 0.4219,
+      "step": 6575
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7791547025338776e-05,
+      "loss": 0.3908,
+      "step": 6576
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7791211190032408e-05,
+      "loss": 0.3235,
+      "step": 6577
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7790875354726046e-05,
+      "loss": 0.3819,
+      "step": 6578
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7790539519419677e-05,
+      "loss": 0.4336,
+      "step": 6579
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.779020368411331e-05,
+      "loss": 0.282,
+      "step": 6580
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7789867848806946e-05,
+      "loss": 0.375,
+      "step": 6581
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.778953201350058e-05,
+      "loss": 0.3762,
+      "step": 6582
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7789196178194216e-05,
+      "loss": 0.2993,
+      "step": 6583
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.778886034288785e-05,
+      "loss": 0.2765,
+      "step": 6584
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7788524507581485e-05,
+      "loss": 0.2831,
+      "step": 6585
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7788188672275116e-05,
+      "loss": 0.1432,
+      "step": 6586
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.778785283696875e-05,
+      "loss": 0.3274,
+      "step": 6587
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7787517001662385e-05,
+      "loss": 0.2291,
+      "step": 6588
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.778718116635602e-05,
+      "loss": 0.32,
+      "step": 6589
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7786845331049655e-05,
+      "loss": 0.1565,
+      "step": 6590
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.778650949574329e-05,
+      "loss": 0.3815,
+      "step": 6591
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7786173660436924e-05,
+      "loss": 0.2604,
+      "step": 6592
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7785837825130555e-05,
+      "loss": 0.3846,
+      "step": 6593
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7785501989824193e-05,
+      "loss": 0.2272,
+      "step": 6594
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7785166154517825e-05,
+      "loss": 0.3469,
+      "step": 6595
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.778483031921146e-05,
+      "loss": 0.2319,
+      "step": 6596
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7784494483905094e-05,
+      "loss": 0.3877,
+      "step": 6597
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.778415864859873e-05,
+      "loss": 0.317,
+      "step": 6598
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7783822813292363e-05,
+      "loss": 0.3371,
+      "step": 6599
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7783486977985998e-05,
+      "loss": 0.3667,
+      "step": 6600
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7783151142679632e-05,
+      "loss": 0.3641,
+      "step": 6601
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7782815307373264e-05,
+      "loss": 0.5067,
+      "step": 6602
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7782479472066902e-05,
+      "loss": 0.2073,
+      "step": 6603
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7782143636760533e-05,
+      "loss": 0.3102,
+      "step": 6604
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7781807801454168e-05,
+      "loss": 0.3602,
+      "step": 6605
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7781471966147802e-05,
+      "loss": 0.3083,
+      "step": 6606
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7781136130841437e-05,
+      "loss": 0.3573,
+      "step": 6607
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.778080029553507e-05,
+      "loss": 0.3033,
+      "step": 6608
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7780464460228706e-05,
+      "loss": 0.3074,
+      "step": 6609
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.778012862492234e-05,
+      "loss": 0.358,
+      "step": 6610
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7779792789615972e-05,
+      "loss": 0.5123,
+      "step": 6611
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.777945695430961e-05,
+      "loss": 0.3583,
+      "step": 6612
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.777912111900324e-05,
+      "loss": 0.3619,
+      "step": 6613
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7778785283696876e-05,
+      "loss": 0.4285,
+      "step": 6614
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.777844944839051e-05,
+      "loss": 0.3393,
+      "step": 6615
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7778113613084145e-05,
+      "loss": 0.3486,
+      "step": 6616
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7777777777777777e-05,
+      "loss": 0.3754,
+      "step": 6617
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7777441942471415e-05,
+      "loss": 0.2754,
+      "step": 6618
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7777106107165046e-05,
+      "loss": 0.2856,
+      "step": 6619
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.777677027185868e-05,
+      "loss": 0.2944,
+      "step": 6620
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7776434436552315e-05,
+      "loss": 0.2395,
+      "step": 6621
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.777609860124595e-05,
+      "loss": 0.5072,
+      "step": 6622
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7775762765939584e-05,
+      "loss": 0.4654,
+      "step": 6623
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.777542693063322e-05,
+      "loss": 0.3898,
+      "step": 6624
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7775091095326854e-05,
+      "loss": 0.5362,
+      "step": 6625
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7774755260020485e-05,
+      "loss": 0.2872,
+      "step": 6626
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7774419424714123e-05,
+      "loss": 0.2642,
+      "step": 6627
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7774083589407754e-05,
+      "loss": 0.5126,
+      "step": 6628
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.777374775410139e-05,
+      "loss": 0.2726,
+      "step": 6629
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7773411918795024e-05,
+      "loss": 0.5025,
+      "step": 6630
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7773076083488658e-05,
+      "loss": 0.4646,
+      "step": 6631
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7772740248182293e-05,
+      "loss": 0.2831,
+      "step": 6632
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7772404412875928e-05,
+      "loss": 0.4329,
+      "step": 6633
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7772068577569562e-05,
+      "loss": 0.2725,
+      "step": 6634
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7771732742263193e-05,
+      "loss": 0.2242,
+      "step": 6635
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.777139690695683e-05,
+      "loss": 0.3486,
+      "step": 6636
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7771061071650463e-05,
+      "loss": 0.2529,
+      "step": 6637
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7770725236344097e-05,
+      "loss": 0.341,
+      "step": 6638
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7770389401037732e-05,
+      "loss": 0.2217,
+      "step": 6639
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7770053565731367e-05,
+      "loss": 0.305,
+      "step": 6640
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7769717730425e-05,
+      "loss": 0.2992,
+      "step": 6641
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7769381895118636e-05,
+      "loss": 0.3571,
+      "step": 6642
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.776904605981227e-05,
+      "loss": 0.2684,
+      "step": 6643
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7768710224505902e-05,
+      "loss": 0.3223,
+      "step": 6644
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.776837438919954e-05,
+      "loss": 0.3744,
+      "step": 6645
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.776803855389317e-05,
+      "loss": 0.3281,
+      "step": 6646
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7767702718586806e-05,
+      "loss": 0.3634,
+      "step": 6647
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.776736688328044e-05,
+      "loss": 0.2229,
+      "step": 6648
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7767031047974075e-05,
+      "loss": 0.2445,
+      "step": 6649
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.776669521266771e-05,
+      "loss": 0.3238,
+      "step": 6650
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7766359377361344e-05,
+      "loss": 0.344,
+      "step": 6651
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.776602354205498e-05,
+      "loss": 0.359,
+      "step": 6652
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.776568770674861e-05,
+      "loss": 0.1936,
+      "step": 6653
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.776535187144225e-05,
+      "loss": 0.2631,
+      "step": 6654
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.776501603613588e-05,
+      "loss": 0.5651,
+      "step": 6655
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7764680200829514e-05,
+      "loss": 0.3572,
+      "step": 6656
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.776434436552315e-05,
+      "loss": 0.3172,
+      "step": 6657
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7764008530216784e-05,
+      "loss": 0.4005,
+      "step": 6658
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7763672694910418e-05,
+      "loss": 0.3476,
+      "step": 6659
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7763336859604053e-05,
+      "loss": 0.1711,
+      "step": 6660
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7763001024297684e-05,
+      "loss": 0.4831,
+      "step": 6661
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.776266518899132e-05,
+      "loss": 0.3376,
+      "step": 6662
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7762329353684953e-05,
+      "loss": 0.2377,
+      "step": 6663
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7761993518378588e-05,
+      "loss": 0.4914,
+      "step": 6664
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7761657683072223e-05,
+      "loss": 0.3947,
+      "step": 6665
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7761321847765857e-05,
+      "loss": 0.2804,
+      "step": 6666
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7760986012459492e-05,
+      "loss": 0.2191,
+      "step": 6667
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7760650177153123e-05,
+      "loss": 0.193,
+      "step": 6668
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.776031434184676e-05,
+      "loss": 0.31,
+      "step": 6669
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7759978506540393e-05,
+      "loss": 0.4064,
+      "step": 6670
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7759642671234027e-05,
+      "loss": 0.3519,
+      "step": 6671
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7759306835927662e-05,
+      "loss": 0.1699,
+      "step": 6672
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7758971000621297e-05,
+      "loss": 0.5524,
+      "step": 6673
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.775863516531493e-05,
+      "loss": 0.2985,
+      "step": 6674
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7758299330008566e-05,
+      "loss": 0.2352,
+      "step": 6675
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.77579634947022e-05,
+      "loss": 0.5914,
+      "step": 6676
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7757627659395832e-05,
+      "loss": 0.2838,
+      "step": 6677
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.775729182408947e-05,
+      "loss": 0.3136,
+      "step": 6678
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.77569559887831e-05,
+      "loss": 0.387,
+      "step": 6679
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7756620153476736e-05,
+      "loss": 0.2948,
+      "step": 6680
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.775628431817037e-05,
+      "loss": 0.2239,
+      "step": 6681
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7755948482864005e-05,
+      "loss": 0.1933,
+      "step": 6682
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.775561264755764e-05,
+      "loss": 0.4009,
+      "step": 6683
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7755276812251274e-05,
+      "loss": 0.5083,
+      "step": 6684
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.775494097694491e-05,
+      "loss": 0.4688,
+      "step": 6685
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.775460514163854e-05,
+      "loss": 0.3069,
+      "step": 6686
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7754269306332178e-05,
+      "loss": 0.2705,
+      "step": 6687
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.775393347102581e-05,
+      "loss": 0.2487,
+      "step": 6688
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7753597635719444e-05,
+      "loss": 0.271,
+      "step": 6689
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.775326180041308e-05,
+      "loss": 0.3078,
+      "step": 6690
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7752925965106713e-05,
+      "loss": 0.2996,
+      "step": 6691
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7752590129800348e-05,
+      "loss": 0.2872,
+      "step": 6692
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7752254294493983e-05,
+      "loss": 0.2992,
+      "step": 6693
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7751918459187617e-05,
+      "loss": 0.5084,
+      "step": 6694
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.775158262388125e-05,
+      "loss": 0.2374,
+      "step": 6695
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7751246788574887e-05,
+      "loss": 0.4412,
+      "step": 6696
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7750910953268518e-05,
+      "loss": 0.3714,
+      "step": 6697
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7750575117962153e-05,
+      "loss": 0.4147,
+      "step": 6698
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7750239282655787e-05,
+      "loss": 0.4099,
+      "step": 6699
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7749903447349422e-05,
+      "loss": 0.3486,
+      "step": 6700
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7749567612043056e-05,
+      "loss": 0.3422,
+      "step": 6701
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.774923177673669e-05,
+      "loss": 0.1538,
+      "step": 6702
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7748895941430326e-05,
+      "loss": 0.2369,
+      "step": 6703
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7748560106123957e-05,
+      "loss": 0.3191,
+      "step": 6704
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.774822427081759e-05,
+      "loss": 0.3914,
+      "step": 6705
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7747888435511226e-05,
+      "loss": 0.2463,
+      "step": 6706
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.774755260020486e-05,
+      "loss": 0.3519,
+      "step": 6707
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7747216764898496e-05,
+      "loss": 0.3825,
+      "step": 6708
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.774688092959213e-05,
+      "loss": 0.2543,
+      "step": 6709
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.774654509428576e-05,
+      "loss": 0.5144,
+      "step": 6710
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.77462092589794e-05,
+      "loss": 0.7597,
+      "step": 6711
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.774587342367303e-05,
+      "loss": 0.1971,
+      "step": 6712
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7745537588366665e-05,
+      "loss": 0.4015,
+      "step": 6713
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.77452017530603e-05,
+      "loss": 0.5131,
+      "step": 6714
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7744865917753935e-05,
+      "loss": 0.3929,
+      "step": 6715
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.774453008244757e-05,
+      "loss": 0.2875,
+      "step": 6716
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7744194247141204e-05,
+      "loss": 0.3137,
+      "step": 6717
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.774385841183484e-05,
+      "loss": 0.3937,
+      "step": 6718
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.774352257652847e-05,
+      "loss": 0.8197,
+      "step": 6719
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7743186741222108e-05,
+      "loss": 0.2293,
+      "step": 6720
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.774285090591574e-05,
+      "loss": 0.4244,
+      "step": 6721
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7742515070609374e-05,
+      "loss": 0.3279,
+      "step": 6722
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.774217923530301e-05,
+      "loss": 0.303,
+      "step": 6723
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7741843399996643e-05,
+      "loss": 0.229,
+      "step": 6724
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7741507564690278e-05,
+      "loss": 0.4289,
+      "step": 6725
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7741171729383913e-05,
+      "loss": 0.503,
+      "step": 6726
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7740835894077547e-05,
+      "loss": 0.3938,
+      "step": 6727
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.774050005877118e-05,
+      "loss": 0.2625,
+      "step": 6728
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7740164223464816e-05,
+      "loss": 0.3868,
+      "step": 6729
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7739828388158448e-05,
+      "loss": 0.2243,
+      "step": 6730
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7739492552852082e-05,
+      "loss": 0.2741,
+      "step": 6731
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7739156717545717e-05,
+      "loss": 0.2624,
+      "step": 6732
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.773882088223935e-05,
+      "loss": 0.2998,
+      "step": 6733
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7738485046932986e-05,
+      "loss": 0.3462,
+      "step": 6734
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.773814921162662e-05,
+      "loss": 0.3055,
+      "step": 6735
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7737813376320256e-05,
+      "loss": 0.3942,
+      "step": 6736
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7737477541013887e-05,
+      "loss": 0.2725,
+      "step": 6737
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7737141705707525e-05,
+      "loss": 0.2706,
+      "step": 6738
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7736805870401156e-05,
+      "loss": 0.409,
+      "step": 6739
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.773647003509479e-05,
+      "loss": 0.4205,
+      "step": 6740
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7736134199788425e-05,
+      "loss": 0.2891,
+      "step": 6741
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.773579836448206e-05,
+      "loss": 0.2206,
+      "step": 6742
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7735462529175695e-05,
+      "loss": 0.3978,
+      "step": 6743
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.773512669386933e-05,
+      "loss": 0.5853,
+      "step": 6744
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7734790858562964e-05,
+      "loss": 0.3579,
+      "step": 6745
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7734455023256595e-05,
+      "loss": 0.3647,
+      "step": 6746
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7734119187950233e-05,
+      "loss": 0.6864,
+      "step": 6747
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7733783352643865e-05,
+      "loss": 0.3066,
+      "step": 6748
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.77334475173375e-05,
+      "loss": 0.2204,
+      "step": 6749
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7733111682031134e-05,
+      "loss": 0.2894,
+      "step": 6750
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.773277584672477e-05,
+      "loss": 0.1985,
+      "step": 6751
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.77324400114184e-05,
+      "loss": 0.3308,
+      "step": 6752
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7732104176112038e-05,
+      "loss": 0.3874,
+      "step": 6753
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.773176834080567e-05,
+      "loss": 0.3111,
+      "step": 6754
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7731432505499304e-05,
+      "loss": 0.388,
+      "step": 6755
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.773109667019294e-05,
+      "loss": 0.2363,
+      "step": 6756
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7730760834886573e-05,
+      "loss": 0.4027,
+      "step": 6757
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7730424999580208e-05,
+      "loss": 0.3819,
+      "step": 6758
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7730089164273842e-05,
+      "loss": 0.333,
+      "step": 6759
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7729753328967477e-05,
+      "loss": 0.2707,
+      "step": 6760
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7729417493661108e-05,
+      "loss": 0.5424,
+      "step": 6761
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7729081658354746e-05,
+      "loss": 0.3383,
+      "step": 6762
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7728745823048378e-05,
+      "loss": 0.5896,
+      "step": 6763
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7728409987742012e-05,
+      "loss": 0.1806,
+      "step": 6764
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7728074152435647e-05,
+      "loss": 0.4253,
+      "step": 6765
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.772773831712928e-05,
+      "loss": 0.3476,
+      "step": 6766
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7727402481822916e-05,
+      "loss": 0.3675,
+      "step": 6767
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7727066646516547e-05,
+      "loss": 0.3306,
+      "step": 6768
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7726730811210185e-05,
+      "loss": 0.2759,
+      "step": 6769
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7726394975903817e-05,
+      "loss": 0.5769,
+      "step": 6770
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.772605914059745e-05,
+      "loss": 0.4716,
+      "step": 6771
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7725723305291086e-05,
+      "loss": 0.2637,
+      "step": 6772
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.772538746998472e-05,
+      "loss": 0.3207,
+      "step": 6773
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7725051634678355e-05,
+      "loss": 0.2953,
+      "step": 6774
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.772471579937199e-05,
+      "loss": 0.2618,
+      "step": 6775
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7724379964065625e-05,
+      "loss": 0.3201,
+      "step": 6776
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7724044128759256e-05,
+      "loss": 0.4466,
+      "step": 6777
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7723708293452894e-05,
+      "loss": 0.4369,
+      "step": 6778
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7723372458146525e-05,
+      "loss": 0.2815,
+      "step": 6779
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.772303662284016e-05,
+      "loss": 0.3578,
+      "step": 6780
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7722700787533794e-05,
+      "loss": 0.3846,
+      "step": 6781
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.772236495222743e-05,
+      "loss": 0.2481,
+      "step": 6782
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7722029116921064e-05,
+      "loss": 0.28,
+      "step": 6783
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.77216932816147e-05,
+      "loss": 0.3586,
+      "step": 6784
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7721357446308333e-05,
+      "loss": 0.2257,
+      "step": 6785
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7721021611001964e-05,
+      "loss": 0.3235,
+      "step": 6786
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7720685775695602e-05,
+      "loss": 0.4249,
+      "step": 6787
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7720349940389234e-05,
+      "loss": 0.4131,
+      "step": 6788
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7720014105082868e-05,
+      "loss": 0.2779,
+      "step": 6789
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7719678269776503e-05,
+      "loss": 0.2124,
+      "step": 6790
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7719342434470137e-05,
+      "loss": 0.2478,
+      "step": 6791
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7719006599163772e-05,
+      "loss": 0.4071,
+      "step": 6792
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7718670763857407e-05,
+      "loss": 0.3245,
+      "step": 6793
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.771833492855104e-05,
+      "loss": 0.3024,
+      "step": 6794
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7717999093244673e-05,
+      "loss": 0.4016,
+      "step": 6795
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7717663257938307e-05,
+      "loss": 0.2838,
+      "step": 6796
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7717327422631942e-05,
+      "loss": 0.2387,
+      "step": 6797
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7716991587325577e-05,
+      "loss": 0.3575,
+      "step": 6798
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.771665575201921e-05,
+      "loss": 0.5819,
+      "step": 6799
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7716319916712846e-05,
+      "loss": 0.4632,
+      "step": 6800
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7715984081406477e-05,
+      "loss": 0.4726,
+      "step": 6801
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7715648246100115e-05,
+      "loss": 0.2601,
+      "step": 6802
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7715312410793746e-05,
+      "loss": 0.3023,
+      "step": 6803
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.771497657548738e-05,
+      "loss": 0.2352,
+      "step": 6804
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7714640740181016e-05,
+      "loss": 0.405,
+      "step": 6805
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.771430490487465e-05,
+      "loss": 0.3578,
+      "step": 6806
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7713969069568285e-05,
+      "loss": 0.332,
+      "step": 6807
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.771363323426192e-05,
+      "loss": 0.2573,
+      "step": 6808
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7713297398955554e-05,
+      "loss": 0.4091,
+      "step": 6809
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7712961563649186e-05,
+      "loss": 0.3119,
+      "step": 6810
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7712625728342824e-05,
+      "loss": 0.3267,
+      "step": 6811
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7712289893036455e-05,
+      "loss": 0.3616,
+      "step": 6812
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.771195405773009e-05,
+      "loss": 0.1377,
+      "step": 6813
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7711618222423724e-05,
+      "loss": 0.4453,
+      "step": 6814
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.771128238711736e-05,
+      "loss": 0.354,
+      "step": 6815
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7710946551810993e-05,
+      "loss": 0.3762,
+      "step": 6816
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7710610716504628e-05,
+      "loss": 0.2353,
+      "step": 6817
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7710274881198263e-05,
+      "loss": 0.3346,
+      "step": 6818
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7709939045891894e-05,
+      "loss": 0.2953,
+      "step": 6819
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7709603210585532e-05,
+      "loss": 0.3943,
+      "step": 6820
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7709267375279163e-05,
+      "loss": 0.4982,
+      "step": 6821
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7708931539972798e-05,
+      "loss": 0.3057,
+      "step": 6822
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7708595704666433e-05,
+      "loss": 0.2459,
+      "step": 6823
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7708259869360067e-05,
+      "loss": 0.2382,
+      "step": 6824
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7707924034053702e-05,
+      "loss": 0.2248,
+      "step": 6825
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7707588198747337e-05,
+      "loss": 0.2861,
+      "step": 6826
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.770725236344097e-05,
+      "loss": 0.4287,
+      "step": 6827
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7706916528134602e-05,
+      "loss": 0.4489,
+      "step": 6828
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.770658069282824e-05,
+      "loss": 0.3218,
+      "step": 6829
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7706244857521872e-05,
+      "loss": 0.2709,
+      "step": 6830
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7705909022215506e-05,
+      "loss": 0.3966,
+      "step": 6831
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.770557318690914e-05,
+      "loss": 0.4864,
+      "step": 6832
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7705237351602776e-05,
+      "loss": 0.2968,
+      "step": 6833
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.770490151629641e-05,
+      "loss": 0.3383,
+      "step": 6834
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7704565680990045e-05,
+      "loss": 0.1734,
+      "step": 6835
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.770422984568368e-05,
+      "loss": 0.4976,
+      "step": 6836
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.770389401037731e-05,
+      "loss": 0.3556,
+      "step": 6837
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.770355817507095e-05,
+      "loss": 0.2302,
+      "step": 6838
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.770322233976458e-05,
+      "loss": 0.228,
+      "step": 6839
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7702886504458215e-05,
+      "loss": 0.3889,
+      "step": 6840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.770255066915185e-05,
+      "loss": 0.4653,
+      "step": 6841
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7702214833845484e-05,
+      "loss": 0.3664,
+      "step": 6842
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7701878998539115e-05,
+      "loss": 0.232,
+      "step": 6843
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7701543163232753e-05,
+      "loss": 0.3096,
+      "step": 6844
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7701207327926385e-05,
+      "loss": 0.3237,
+      "step": 6845
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.770087149262002e-05,
+      "loss": 0.3655,
+      "step": 6846
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7700535657313654e-05,
+      "loss": 0.232,
+      "step": 6847
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.770019982200729e-05,
+      "loss": 0.4221,
+      "step": 6848
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7699863986700923e-05,
+      "loss": 0.4926,
+      "step": 6849
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7699528151394558e-05,
+      "loss": 0.2513,
+      "step": 6850
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7699192316088193e-05,
+      "loss": 0.2533,
+      "step": 6851
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7698856480781824e-05,
+      "loss": 0.5928,
+      "step": 6852
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7698520645475462e-05,
+      "loss": 0.3362,
+      "step": 6853
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7698184810169093e-05,
+      "loss": 0.4001,
+      "step": 6854
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7697848974862728e-05,
+      "loss": 0.3285,
+      "step": 6855
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7697513139556362e-05,
+      "loss": 0.2941,
+      "step": 6856
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7697177304249997e-05,
+      "loss": 0.2321,
+      "step": 6857
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7696841468943632e-05,
+      "loss": 0.3398,
+      "step": 6858
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7696505633637266e-05,
+      "loss": 0.2215,
+      "step": 6859
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.76961697983309e-05,
+      "loss": 0.435,
+      "step": 6860
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7695833963024532e-05,
+      "loss": 0.2934,
+      "step": 6861
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.769549812771817e-05,
+      "loss": 0.3005,
+      "step": 6862
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.76951622924118e-05,
+      "loss": 0.46,
+      "step": 6863
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7694826457105436e-05,
+      "loss": 0.4588,
+      "step": 6864
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.769449062179907e-05,
+      "loss": 0.3322,
+      "step": 6865
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7694154786492706e-05,
+      "loss": 0.4388,
+      "step": 6866
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.769381895118634e-05,
+      "loss": 0.272,
+      "step": 6867
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7693483115879975e-05,
+      "loss": 0.2728,
+      "step": 6868
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.769314728057361e-05,
+      "loss": 0.4396,
+      "step": 6869
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.769281144526724e-05,
+      "loss": 0.6711,
+      "step": 6870
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.769247560996088e-05,
+      "loss": 0.3105,
+      "step": 6871
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.769213977465451e-05,
+      "loss": 0.2255,
+      "step": 6872
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7691803939348145e-05,
+      "loss": 0.4636,
+      "step": 6873
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.769146810404178e-05,
+      "loss": 0.3379,
+      "step": 6874
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7691132268735414e-05,
+      "loss": 0.322,
+      "step": 6875
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.769079643342905e-05,
+      "loss": 0.4972,
+      "step": 6876
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7690460598122683e-05,
+      "loss": 0.4869,
+      "step": 6877
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7690124762816318e-05,
+      "loss": 0.4805,
+      "step": 6878
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.768978892750995e-05,
+      "loss": 0.5127,
+      "step": 6879
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7689453092203587e-05,
+      "loss": 0.3631,
+      "step": 6880
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.768911725689722e-05,
+      "loss": 0.428,
+      "step": 6881
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7688781421590853e-05,
+      "loss": 0.2529,
+      "step": 6882
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7688445586284488e-05,
+      "loss": 0.2349,
+      "step": 6883
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7688109750978122e-05,
+      "loss": 0.2672,
+      "step": 6884
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7687773915671757e-05,
+      "loss": 0.3502,
+      "step": 6885
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7687438080365392e-05,
+      "loss": 0.3144,
+      "step": 6886
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7687102245059023e-05,
+      "loss": 0.3574,
+      "step": 6887
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7686766409752658e-05,
+      "loss": 0.2461,
+      "step": 6888
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7686430574446292e-05,
+      "loss": 0.2609,
+      "step": 6889
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7686094739139927e-05,
+      "loss": 0.2878,
+      "step": 6890
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.768575890383356e-05,
+      "loss": 0.4583,
+      "step": 6891
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7685423068527196e-05,
+      "loss": 0.4758,
+      "step": 6892
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.768508723322083e-05,
+      "loss": 0.368,
+      "step": 6893
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7684751397914462e-05,
+      "loss": 0.3043,
+      "step": 6894
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.76844155626081e-05,
+      "loss": 0.4625,
+      "step": 6895
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.768407972730173e-05,
+      "loss": 0.2392,
+      "step": 6896
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7683743891995366e-05,
+      "loss": 0.2878,
+      "step": 6897
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7683408056689e-05,
+      "loss": 0.3297,
+      "step": 6898
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7683072221382635e-05,
+      "loss": 0.4356,
+      "step": 6899
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.768273638607627e-05,
+      "loss": 0.3646,
+      "step": 6900
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7682400550769905e-05,
+      "loss": 0.2623,
+      "step": 6901
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.768206471546354e-05,
+      "loss": 0.5229,
+      "step": 6902
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.768172888015717e-05,
+      "loss": 0.2172,
+      "step": 6903
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.768139304485081e-05,
+      "loss": 0.3059,
+      "step": 6904
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.768105720954444e-05,
+      "loss": 0.2603,
+      "step": 6905
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7680721374238074e-05,
+      "loss": 0.3343,
+      "step": 6906
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.768038553893171e-05,
+      "loss": 0.381,
+      "step": 6907
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7680049703625344e-05,
+      "loss": 0.4587,
+      "step": 6908
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.767971386831898e-05,
+      "loss": 0.2989,
+      "step": 6909
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7679378033012613e-05,
+      "loss": 0.4764,
+      "step": 6910
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7679042197706248e-05,
+      "loss": 0.3178,
+      "step": 6911
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.767870636239988e-05,
+      "loss": 0.3088,
+      "step": 6912
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7678370527093517e-05,
+      "loss": 0.3251,
+      "step": 6913
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7678034691787148e-05,
+      "loss": 0.2766,
+      "step": 6914
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7677698856480783e-05,
+      "loss": 0.2155,
+      "step": 6915
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7677363021174418e-05,
+      "loss": 0.3611,
+      "step": 6916
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7677027185868052e-05,
+      "loss": 0.369,
+      "step": 6917
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7676691350561687e-05,
+      "loss": 0.3696,
+      "step": 6918
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.767635551525532e-05,
+      "loss": 0.397,
+      "step": 6919
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7676019679948956e-05,
+      "loss": 0.2706,
+      "step": 6920
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7675683844642587e-05,
+      "loss": 0.3089,
+      "step": 6921
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7675348009336225e-05,
+      "loss": 0.222,
+      "step": 6922
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7675012174029857e-05,
+      "loss": 0.4104,
+      "step": 6923
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.767467633872349e-05,
+      "loss": 0.301,
+      "step": 6924
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7674340503417126e-05,
+      "loss": 0.2514,
+      "step": 6925
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.767400466811076e-05,
+      "loss": 0.3389,
+      "step": 6926
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7673668832804395e-05,
+      "loss": 0.271,
+      "step": 6927
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.767333299749803e-05,
+      "loss": 0.2254,
+      "step": 6928
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7672997162191665e-05,
+      "loss": 0.1889,
+      "step": 6929
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7672661326885296e-05,
+      "loss": 0.3806,
+      "step": 6930
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.767232549157893e-05,
+      "loss": 0.2895,
+      "step": 6931
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7671989656272565e-05,
+      "loss": 0.4419,
+      "step": 6932
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.76716538209662e-05,
+      "loss": 0.3441,
+      "step": 6933
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7671317985659834e-05,
+      "loss": 0.5564,
+      "step": 6934
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.767098215035347e-05,
+      "loss": 0.625,
+      "step": 6935
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.76706463150471e-05,
+      "loss": 0.2274,
+      "step": 6936
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.767031047974074e-05,
+      "loss": 0.3755,
+      "step": 6937
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.766997464443437e-05,
+      "loss": 0.3008,
+      "step": 6938
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7669638809128004e-05,
+      "loss": 0.3949,
+      "step": 6939
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.766930297382164e-05,
+      "loss": 0.4325,
+      "step": 6940
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7668967138515274e-05,
+      "loss": 0.4403,
+      "step": 6941
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7668631303208908e-05,
+      "loss": 0.3204,
+      "step": 6942
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7668295467902543e-05,
+      "loss": 0.4578,
+      "step": 6943
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7667959632596178e-05,
+      "loss": 0.3838,
+      "step": 6944
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.766762379728981e-05,
+      "loss": 0.5045,
+      "step": 6945
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7667287961983447e-05,
+      "loss": 0.3671,
+      "step": 6946
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7666952126677078e-05,
+      "loss": 0.3549,
+      "step": 6947
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7666616291370713e-05,
+      "loss": 0.3561,
+      "step": 6948
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7666280456064347e-05,
+      "loss": 0.2676,
+      "step": 6949
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7665944620757982e-05,
+      "loss": 0.3597,
+      "step": 6950
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7665608785451617e-05,
+      "loss": 0.4763,
+      "step": 6951
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7665272950145248e-05,
+      "loss": 0.3797,
+      "step": 6952
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7664937114838886e-05,
+      "loss": 0.3342,
+      "step": 6953
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7664601279532517e-05,
+      "loss": 0.4177,
+      "step": 6954
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7664265444226152e-05,
+      "loss": 0.3137,
+      "step": 6955
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7663929608919787e-05,
+      "loss": 0.4508,
+      "step": 6956
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.766359377361342e-05,
+      "loss": 0.4086,
+      "step": 6957
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7663257938307056e-05,
+      "loss": 0.2855,
+      "step": 6958
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.766292210300069e-05,
+      "loss": 0.4637,
+      "step": 6959
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7662586267694325e-05,
+      "loss": 0.3479,
+      "step": 6960
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7662250432387956e-05,
+      "loss": 0.4412,
+      "step": 6961
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7661914597081594e-05,
+      "loss": 0.3498,
+      "step": 6962
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7661578761775226e-05,
+      "loss": 0.3712,
+      "step": 6963
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.766124292646886e-05,
+      "loss": 0.2991,
+      "step": 6964
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7660907091162495e-05,
+      "loss": 0.5812,
+      "step": 6965
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.766057125585613e-05,
+      "loss": 0.4138,
+      "step": 6966
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7660235420549764e-05,
+      "loss": 0.2366,
+      "step": 6967
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.76598995852434e-05,
+      "loss": 0.3674,
+      "step": 6968
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7659563749937034e-05,
+      "loss": 0.5271,
+      "step": 6969
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7659227914630665e-05,
+      "loss": 0.2021,
+      "step": 6970
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7658892079324303e-05,
+      "loss": 0.4548,
+      "step": 6971
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7658556244017934e-05,
+      "loss": 0.453,
+      "step": 6972
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.765822040871157e-05,
+      "loss": 0.2637,
+      "step": 6973
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7657884573405203e-05,
+      "loss": 0.2152,
+      "step": 6974
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7657548738098838e-05,
+      "loss": 0.3796,
+      "step": 6975
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7657212902792473e-05,
+      "loss": 0.2413,
+      "step": 6976
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7656877067486107e-05,
+      "loss": 0.4288,
+      "step": 6977
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.765654123217974e-05,
+      "loss": 0.2619,
+      "step": 6978
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7656205396873373e-05,
+      "loss": 0.2404,
+      "step": 6979
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7655869561567008e-05,
+      "loss": 0.3504,
+      "step": 6980
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7655533726260643e-05,
+      "loss": 0.2108,
+      "step": 6981
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7655197890954277e-05,
+      "loss": 0.287,
+      "step": 6982
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7654862055647912e-05,
+      "loss": 0.5265,
+      "step": 6983
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7654526220341546e-05,
+      "loss": 0.2953,
+      "step": 6984
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7654190385035178e-05,
+      "loss": 0.4035,
+      "step": 6985
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7653854549728816e-05,
+      "loss": 0.2033,
+      "step": 6986
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7653518714422447e-05,
+      "loss": 0.3573,
+      "step": 6987
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7653182879116082e-05,
+      "loss": 0.2516,
+      "step": 6988
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7652847043809716e-05,
+      "loss": 0.6343,
+      "step": 6989
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.765251120850335e-05,
+      "loss": 0.1321,
+      "step": 6990
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7652175373196986e-05,
+      "loss": 0.2864,
+      "step": 6991
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.765183953789062e-05,
+      "loss": 0.3228,
+      "step": 6992
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7651503702584255e-05,
+      "loss": 0.2008,
+      "step": 6993
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7651167867277886e-05,
+      "loss": 0.2545,
+      "step": 6994
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7650832031971524e-05,
+      "loss": 0.4467,
+      "step": 6995
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7650496196665155e-05,
+      "loss": 0.2356,
+      "step": 6996
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.765016036135879e-05,
+      "loss": 0.279,
+      "step": 6997
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7649824526052425e-05,
+      "loss": 0.2969,
+      "step": 6998
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.764948869074606e-05,
+      "loss": 0.2657,
+      "step": 6999
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7649152855439694e-05,
+      "loss": 0.4359,
+      "step": 7000
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.764881702013333e-05,
+      "loss": 0.2807,
+      "step": 7001
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7648481184826963e-05,
+      "loss": 0.428,
+      "step": 7002
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7648145349520595e-05,
+      "loss": 0.3994,
+      "step": 7003
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7647809514214233e-05,
+      "loss": 0.3978,
+      "step": 7004
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7647473678907864e-05,
+      "loss": 0.4208,
+      "step": 7005
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.76471378436015e-05,
+      "loss": 0.2644,
+      "step": 7006
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7646802008295133e-05,
+      "loss": 0.4109,
+      "step": 7007
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7646466172988768e-05,
+      "loss": 0.2814,
+      "step": 7008
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7646130337682403e-05,
+      "loss": 0.3649,
+      "step": 7009
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7645794502376037e-05,
+      "loss": 0.3868,
+      "step": 7010
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7645458667069672e-05,
+      "loss": 0.389,
+      "step": 7011
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7645122831763303e-05,
+      "loss": 0.3206,
+      "step": 7012
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.764478699645694e-05,
+      "loss": 0.3989,
+      "step": 7013
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7644451161150572e-05,
+      "loss": 0.3237,
+      "step": 7014
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7644115325844207e-05,
+      "loss": 0.3611,
+      "step": 7015
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.764377949053784e-05,
+      "loss": 0.3141,
+      "step": 7016
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7643443655231476e-05,
+      "loss": 0.2936,
+      "step": 7017
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.764310781992511e-05,
+      "loss": 0.4971,
+      "step": 7018
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7642771984618746e-05,
+      "loss": 0.3064,
+      "step": 7019
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.764243614931238e-05,
+      "loss": 0.2982,
+      "step": 7020
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.764210031400601e-05,
+      "loss": 0.2542,
+      "step": 7021
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7641764478699646e-05,
+      "loss": 0.2442,
+      "step": 7022
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.764142864339328e-05,
+      "loss": 0.2496,
+      "step": 7023
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7641092808086915e-05,
+      "loss": 0.4098,
+      "step": 7024
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.764075697278055e-05,
+      "loss": 0.2833,
+      "step": 7025
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7640421137474185e-05,
+      "loss": 0.5065,
+      "step": 7026
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7640085302167816e-05,
+      "loss": 0.3471,
+      "step": 7027
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7639749466861454e-05,
+      "loss": 0.29,
+      "step": 7028
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7639413631555085e-05,
+      "loss": 0.3761,
+      "step": 7029
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.763907779624872e-05,
+      "loss": 0.3376,
+      "step": 7030
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7638741960942355e-05,
+      "loss": 0.2679,
+      "step": 7031
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.763840612563599e-05,
+      "loss": 0.266,
+      "step": 7032
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7638070290329624e-05,
+      "loss": 0.392,
+      "step": 7033
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.763773445502326e-05,
+      "loss": 0.5419,
+      "step": 7034
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7637398619716893e-05,
+      "loss": 0.351,
+      "step": 7035
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7637062784410524e-05,
+      "loss": 0.4254,
+      "step": 7036
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7636726949104162e-05,
+      "loss": 0.3472,
+      "step": 7037
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7636391113797794e-05,
+      "loss": 0.3031,
+      "step": 7038
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.763605527849143e-05,
+      "loss": 0.4381,
+      "step": 7039
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7635719443185063e-05,
+      "loss": 0.3352,
+      "step": 7040
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7635383607878698e-05,
+      "loss": 0.2677,
+      "step": 7041
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7635047772572332e-05,
+      "loss": 0.3599,
+      "step": 7042
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7634711937265967e-05,
+      "loss": 0.2591,
+      "step": 7043
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.76343761019596e-05,
+      "loss": 0.1837,
+      "step": 7044
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7634040266653233e-05,
+      "loss": 0.2606,
+      "step": 7045
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.763370443134687e-05,
+      "loss": 0.2975,
+      "step": 7046
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7633368596040502e-05,
+      "loss": 0.4513,
+      "step": 7047
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7633032760734137e-05,
+      "loss": 0.3375,
+      "step": 7048
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.763269692542777e-05,
+      "loss": 0.344,
+      "step": 7049
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7632361090121406e-05,
+      "loss": 0.3744,
+      "step": 7050
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.763202525481504e-05,
+      "loss": 0.4517,
+      "step": 7051
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7631689419508675e-05,
+      "loss": 0.2811,
+      "step": 7052
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.763135358420231e-05,
+      "loss": 0.304,
+      "step": 7053
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.763101774889594e-05,
+      "loss": 0.3204,
+      "step": 7054
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.763068191358958e-05,
+      "loss": 0.4266,
+      "step": 7055
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.763034607828321e-05,
+      "loss": 0.3097,
+      "step": 7056
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7630010242976845e-05,
+      "loss": 0.288,
+      "step": 7057
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.762967440767048e-05,
+      "loss": 0.3817,
+      "step": 7058
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7629338572364115e-05,
+      "loss": 0.4263,
+      "step": 7059
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.762900273705775e-05,
+      "loss": 0.203,
+      "step": 7060
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7628666901751384e-05,
+      "loss": 0.3513,
+      "step": 7061
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.762833106644502e-05,
+      "loss": 0.2421,
+      "step": 7062
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.762799523113865e-05,
+      "loss": 0.3839,
+      "step": 7063
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7627659395832288e-05,
+      "loss": 0.3603,
+      "step": 7064
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.762732356052592e-05,
+      "loss": 0.1257,
+      "step": 7065
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7626987725219554e-05,
+      "loss": 0.2273,
+      "step": 7066
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.762665188991319e-05,
+      "loss": 0.5336,
+      "step": 7067
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7626316054606823e-05,
+      "loss": 0.3278,
+      "step": 7068
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7625980219300454e-05,
+      "loss": 0.4149,
+      "step": 7069
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7625644383994092e-05,
+      "loss": 0.3985,
+      "step": 7070
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7625308548687724e-05,
+      "loss": 0.3565,
+      "step": 7071
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7624972713381358e-05,
+      "loss": 0.3785,
+      "step": 7072
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7624636878074993e-05,
+      "loss": 0.3343,
+      "step": 7073
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7624301042768627e-05,
+      "loss": 0.3535,
+      "step": 7074
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7623965207462262e-05,
+      "loss": 0.2744,
+      "step": 7075
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7623629372155897e-05,
+      "loss": 0.3223,
+      "step": 7076
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.762329353684953e-05,
+      "loss": 0.3461,
+      "step": 7077
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7622957701543163e-05,
+      "loss": 0.4263,
+      "step": 7078
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.76226218662368e-05,
+      "loss": 0.4869,
+      "step": 7079
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7622286030930432e-05,
+      "loss": 0.2649,
+      "step": 7080
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7621950195624067e-05,
+      "loss": 0.3935,
+      "step": 7081
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.76216143603177e-05,
+      "loss": 0.4225,
+      "step": 7082
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7621278525011336e-05,
+      "loss": 0.2036,
+      "step": 7083
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.762094268970497e-05,
+      "loss": 0.2273,
+      "step": 7084
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7620606854398605e-05,
+      "loss": 0.352,
+      "step": 7085
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.762027101909224e-05,
+      "loss": 0.2468,
+      "step": 7086
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.761993518378587e-05,
+      "loss": 0.3873,
+      "step": 7087
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.761959934847951e-05,
+      "loss": 0.4443,
+      "step": 7088
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.761926351317314e-05,
+      "loss": 0.2644,
+      "step": 7089
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7618927677866775e-05,
+      "loss": 0.2986,
+      "step": 7090
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.761859184256041e-05,
+      "loss": 0.3085,
+      "step": 7091
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7618256007254044e-05,
+      "loss": 0.2091,
+      "step": 7092
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.761792017194768e-05,
+      "loss": 0.2196,
+      "step": 7093
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7617584336641314e-05,
+      "loss": 0.413,
+      "step": 7094
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.761724850133495e-05,
+      "loss": 0.4217,
+      "step": 7095
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.761691266602858e-05,
+      "loss": 0.3042,
+      "step": 7096
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7616576830722218e-05,
+      "loss": 0.3679,
+      "step": 7097
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.761624099541585e-05,
+      "loss": 0.3028,
+      "step": 7098
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7615905160109484e-05,
+      "loss": 0.286,
+      "step": 7099
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7615569324803118e-05,
+      "loss": 0.2492,
+      "step": 7100
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7615233489496753e-05,
+      "loss": 0.3178,
+      "step": 7101
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7614897654190387e-05,
+      "loss": 0.349,
+      "step": 7102
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7614561818884022e-05,
+      "loss": 0.3933,
+      "step": 7103
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7614225983577657e-05,
+      "loss": 0.539,
+      "step": 7104
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7613890148271288e-05,
+      "loss": 0.3133,
+      "step": 7105
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7613554312964926e-05,
+      "loss": 0.2568,
+      "step": 7106
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7613218477658557e-05,
+      "loss": 0.3794,
+      "step": 7107
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7612882642352192e-05,
+      "loss": 0.4194,
+      "step": 7108
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7612546807045827e-05,
+      "loss": 0.2911,
+      "step": 7109
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.761221097173946e-05,
+      "loss": 0.4093,
+      "step": 7110
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7611875136433096e-05,
+      "loss": 0.3487,
+      "step": 7111
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.761153930112673e-05,
+      "loss": 0.3678,
+      "step": 7112
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7611203465820362e-05,
+      "loss": 0.6562,
+      "step": 7113
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7610867630513996e-05,
+      "loss": 0.327,
+      "step": 7114
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.761053179520763e-05,
+      "loss": 0.3431,
+      "step": 7115
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7610195959901266e-05,
+      "loss": 0.4505,
+      "step": 7116
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.76098601245949e-05,
+      "loss": 0.3515,
+      "step": 7117
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7609524289288535e-05,
+      "loss": 0.3593,
+      "step": 7118
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.760918845398217e-05,
+      "loss": 0.2621,
+      "step": 7119
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.76088526186758e-05,
+      "loss": 0.1716,
+      "step": 7120
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.760851678336944e-05,
+      "loss": 0.3844,
+      "step": 7121
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.760818094806307e-05,
+      "loss": 0.3092,
+      "step": 7122
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7607845112756705e-05,
+      "loss": 0.2975,
+      "step": 7123
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.760750927745034e-05,
+      "loss": 0.3291,
+      "step": 7124
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7607173442143974e-05,
+      "loss": 0.3494,
+      "step": 7125
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.760683760683761e-05,
+      "loss": 0.3612,
+      "step": 7126
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7606501771531243e-05,
+      "loss": 0.3342,
+      "step": 7127
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7606165936224878e-05,
+      "loss": 0.1739,
+      "step": 7128
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.760583010091851e-05,
+      "loss": 0.4455,
+      "step": 7129
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7605494265612144e-05,
+      "loss": 0.3689,
+      "step": 7130
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.760515843030578e-05,
+      "loss": 0.1756,
+      "step": 7131
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7604822594999413e-05,
+      "loss": 0.3628,
+      "step": 7132
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7604486759693048e-05,
+      "loss": 0.3377,
+      "step": 7133
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7604150924386683e-05,
+      "loss": 0.3354,
+      "step": 7134
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7603815089080317e-05,
+      "loss": 0.4117,
+      "step": 7135
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.760347925377395e-05,
+      "loss": 0.3646,
+      "step": 7136
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7603143418467587e-05,
+      "loss": 0.2766,
+      "step": 7137
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7602807583161218e-05,
+      "loss": 0.3318,
+      "step": 7138
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7602471747854852e-05,
+      "loss": 0.2986,
+      "step": 7139
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7602135912548487e-05,
+      "loss": 0.1864,
+      "step": 7140
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7601800077242122e-05,
+      "loss": 0.2695,
+      "step": 7141
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7601464241935756e-05,
+      "loss": 0.2585,
+      "step": 7142
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.760112840662939e-05,
+      "loss": 0.6152,
+      "step": 7143
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7600792571323026e-05,
+      "loss": 0.1767,
+      "step": 7144
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7600456736016657e-05,
+      "loss": 0.4133,
+      "step": 7145
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7600120900710295e-05,
+      "loss": 0.6099,
+      "step": 7146
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7599785065403926e-05,
+      "loss": 0.3099,
+      "step": 7147
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.759944923009756e-05,
+      "loss": 0.2084,
+      "step": 7148
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7599113394791196e-05,
+      "loss": 0.3071,
+      "step": 7149
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.759877755948483e-05,
+      "loss": 0.3359,
+      "step": 7150
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7598441724178465e-05,
+      "loss": 0.1707,
+      "step": 7151
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.75981058888721e-05,
+      "loss": 0.3286,
+      "step": 7152
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7597770053565734e-05,
+      "loss": 0.3419,
+      "step": 7153
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7597434218259365e-05,
+      "loss": 0.3983,
+      "step": 7154
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7597098382953003e-05,
+      "loss": 0.2743,
+      "step": 7155
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7596762547646635e-05,
+      "loss": 0.2275,
+      "step": 7156
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.759642671234027e-05,
+      "loss": 0.3927,
+      "step": 7157
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7596090877033904e-05,
+      "loss": 0.228,
+      "step": 7158
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.759575504172754e-05,
+      "loss": 0.4772,
+      "step": 7159
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.759541920642117e-05,
+      "loss": 0.4147,
+      "step": 7160
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7595083371114808e-05,
+      "loss": 0.391,
+      "step": 7161
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.759474753580844e-05,
+      "loss": 0.1396,
+      "step": 7162
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7594411700502074e-05,
+      "loss": 0.5761,
+      "step": 7163
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.759407586519571e-05,
+      "loss": 0.2844,
+      "step": 7164
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7593740029889343e-05,
+      "loss": 0.4157,
+      "step": 7165
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7593404194582978e-05,
+      "loss": 0.1879,
+      "step": 7166
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7593068359276612e-05,
+      "loss": 0.2078,
+      "step": 7167
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7592732523970247e-05,
+      "loss": 0.2735,
+      "step": 7168
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.759239668866388e-05,
+      "loss": 0.3261,
+      "step": 7169
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7592060853357516e-05,
+      "loss": 0.341,
+      "step": 7170
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7591725018051148e-05,
+      "loss": 0.3873,
+      "step": 7171
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7591389182744782e-05,
+      "loss": 0.2791,
+      "step": 7172
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7591053347438417e-05,
+      "loss": 0.2555,
+      "step": 7173
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.759071751213205e-05,
+      "loss": 0.196,
+      "step": 7174
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7590381676825686e-05,
+      "loss": 0.4959,
+      "step": 7175
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.759004584151932e-05,
+      "loss": 0.3718,
+      "step": 7176
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7589710006212956e-05,
+      "loss": 0.3724,
+      "step": 7177
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7589374170906587e-05,
+      "loss": 0.4082,
+      "step": 7178
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7589038335600225e-05,
+      "loss": 0.3299,
+      "step": 7179
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7588702500293856e-05,
+      "loss": 0.3361,
+      "step": 7180
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.758836666498749e-05,
+      "loss": 0.2746,
+      "step": 7181
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7588030829681125e-05,
+      "loss": 0.3137,
+      "step": 7182
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.758769499437476e-05,
+      "loss": 0.3827,
+      "step": 7183
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7587359159068395e-05,
+      "loss": 0.2877,
+      "step": 7184
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.758702332376203e-05,
+      "loss": 0.2945,
+      "step": 7185
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7586687488455664e-05,
+      "loss": 0.2488,
+      "step": 7186
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7586351653149295e-05,
+      "loss": 0.3298,
+      "step": 7187
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7586015817842933e-05,
+      "loss": 0.2375,
+      "step": 7188
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7585679982536565e-05,
+      "loss": 0.3488,
+      "step": 7189
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.75853441472302e-05,
+      "loss": 0.3495,
+      "step": 7190
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7585008311923834e-05,
+      "loss": 0.2884,
+      "step": 7191
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.758467247661747e-05,
+      "loss": 0.3188,
+      "step": 7192
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7584336641311103e-05,
+      "loss": 0.268,
+      "step": 7193
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7584000806004738e-05,
+      "loss": 0.2695,
+      "step": 7194
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7583664970698372e-05,
+      "loss": 0.5074,
+      "step": 7195
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7583329135392004e-05,
+      "loss": 0.3789,
+      "step": 7196
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.758299330008564e-05,
+      "loss": 0.273,
+      "step": 7197
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7582657464779273e-05,
+      "loss": 0.3706,
+      "step": 7198
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7582321629472908e-05,
+      "loss": 0.3926,
+      "step": 7199
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7581985794166542e-05,
+      "loss": 0.2521,
+      "step": 7200
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7581649958860177e-05,
+      "loss": 0.5868,
+      "step": 7201
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.758131412355381e-05,
+      "loss": 0.3403,
+      "step": 7202
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7580978288247446e-05,
+      "loss": 0.5076,
+      "step": 7203
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7580642452941077e-05,
+      "loss": 0.4415,
+      "step": 7204
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7580306617634712e-05,
+      "loss": 0.4191,
+      "step": 7205
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7579970782328347e-05,
+      "loss": 0.2621,
+      "step": 7206
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.757963494702198e-05,
+      "loss": 0.3639,
+      "step": 7207
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7579299111715616e-05,
+      "loss": 0.4207,
+      "step": 7208
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.757896327640925e-05,
+      "loss": 0.3598,
+      "step": 7209
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7578627441102885e-05,
+      "loss": 0.3473,
+      "step": 7210
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7578291605796517e-05,
+      "loss": 0.5184,
+      "step": 7211
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7577955770490155e-05,
+      "loss": 0.4725,
+      "step": 7212
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7577619935183786e-05,
+      "loss": 0.3221,
+      "step": 7213
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.757728409987742e-05,
+      "loss": 0.2718,
+      "step": 7214
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7576948264571055e-05,
+      "loss": 0.4456,
+      "step": 7215
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.757661242926469e-05,
+      "loss": 0.2914,
+      "step": 7216
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7576276593958324e-05,
+      "loss": 0.2979,
+      "step": 7217
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.757594075865196e-05,
+      "loss": 0.526,
+      "step": 7218
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7575604923345594e-05,
+      "loss": 0.3593,
+      "step": 7219
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7575269088039225e-05,
+      "loss": 0.2036,
+      "step": 7220
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7574933252732863e-05,
+      "loss": 0.4497,
+      "step": 7221
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7574597417426494e-05,
+      "loss": 0.3233,
+      "step": 7222
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.757426158212013e-05,
+      "loss": 0.3261,
+      "step": 7223
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7573925746813764e-05,
+      "loss": 0.3172,
+      "step": 7224
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7573589911507398e-05,
+      "loss": 0.3726,
+      "step": 7225
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7573254076201033e-05,
+      "loss": 0.3073,
+      "step": 7226
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7572918240894668e-05,
+      "loss": 0.2512,
+      "step": 7227
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7572582405588302e-05,
+      "loss": 0.1799,
+      "step": 7228
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7572246570281933e-05,
+      "loss": 0.3218,
+      "step": 7229
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.757191073497557e-05,
+      "loss": 0.369,
+      "step": 7230
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7571574899669203e-05,
+      "loss": 0.2734,
+      "step": 7231
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7571239064362837e-05,
+      "loss": 0.3905,
+      "step": 7232
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7570903229056472e-05,
+      "loss": 0.442,
+      "step": 7233
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7570567393750107e-05,
+      "loss": 0.3949,
+      "step": 7234
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.757023155844374e-05,
+      "loss": 0.2449,
+      "step": 7235
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7569895723137376e-05,
+      "loss": 0.3624,
+      "step": 7236
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.756955988783101e-05,
+      "loss": 0.3636,
+      "step": 7237
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7569224052524642e-05,
+      "loss": 0.3564,
+      "step": 7238
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.756888821721828e-05,
+      "loss": 0.4909,
+      "step": 7239
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.756855238191191e-05,
+      "loss": 0.3784,
+      "step": 7240
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7568216546605546e-05,
+      "loss": 0.2618,
+      "step": 7241
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.756788071129918e-05,
+      "loss": 0.3967,
+      "step": 7242
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7567544875992815e-05,
+      "loss": 0.2764,
+      "step": 7243
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.756720904068645e-05,
+      "loss": 0.4527,
+      "step": 7244
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.7566873205380084e-05,
+      "loss": 0.2927,
+      "step": 7245
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.756653737007372e-05,
+      "loss": 0.4264,
+      "step": 7246
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.756620153476735e-05,
+      "loss": 0.2681,
+      "step": 7247
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7565865699460985e-05,
+      "loss": 0.2248,
+      "step": 7248
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.756552986415462e-05,
+      "loss": 0.4381,
+      "step": 7249
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7565194028848254e-05,
+      "loss": 0.3043,
+      "step": 7250
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.756485819354189e-05,
+      "loss": 0.4652,
+      "step": 7251
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7564522358235524e-05,
+      "loss": 0.373,
+      "step": 7252
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7564186522929155e-05,
+      "loss": 0.3397,
+      "step": 7253
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7563850687622793e-05,
+      "loss": 0.652,
+      "step": 7254
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7563514852316424e-05,
+      "loss": 0.2592,
+      "step": 7255
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.756317901701006e-05,
+      "loss": 0.402,
+      "step": 7256
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7562843181703693e-05,
+      "loss": 0.2861,
+      "step": 7257
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7562507346397328e-05,
+      "loss": 0.2729,
+      "step": 7258
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7562171511090963e-05,
+      "loss": 0.1995,
+      "step": 7259
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7561835675784597e-05,
+      "loss": 0.2888,
+      "step": 7260
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7561499840478232e-05,
+      "loss": 0.4158,
+      "step": 7261
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7561164005171863e-05,
+      "loss": 0.2323,
+      "step": 7262
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.75608281698655e-05,
+      "loss": 0.2241,
+      "step": 7263
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7560492334559133e-05,
+      "loss": 0.4733,
+      "step": 7264
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7560156499252767e-05,
+      "loss": 0.3224,
+      "step": 7265
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7559820663946402e-05,
+      "loss": 0.1663,
+      "step": 7266
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7559484828640037e-05,
+      "loss": 0.2646,
+      "step": 7267
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.755914899333367e-05,
+      "loss": 0.5071,
+      "step": 7268
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7558813158027306e-05,
+      "loss": 0.404,
+      "step": 7269
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.755847732272094e-05,
+      "loss": 0.2701,
+      "step": 7270
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7558141487414572e-05,
+      "loss": 0.4181,
+      "step": 7271
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.755780565210821e-05,
+      "loss": 0.3122,
+      "step": 7272
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.755746981680184e-05,
+      "loss": 0.2071,
+      "step": 7273
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7557133981495476e-05,
+      "loss": 0.2238,
+      "step": 7274
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.755679814618911e-05,
+      "loss": 0.2775,
+      "step": 7275
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7556462310882745e-05,
+      "loss": 0.215,
+      "step": 7276
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.755612647557638e-05,
+      "loss": 0.3986,
+      "step": 7277
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7555790640270014e-05,
+      "loss": 0.3844,
+      "step": 7278
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.755545480496365e-05,
+      "loss": 0.3746,
+      "step": 7279
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.755511896965728e-05,
+      "loss": 0.3517,
+      "step": 7280
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7554783134350918e-05,
+      "loss": 0.2333,
+      "step": 7281
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.755444729904455e-05,
+      "loss": 0.3445,
+      "step": 7282
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7554111463738184e-05,
+      "loss": 0.2838,
+      "step": 7283
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.755377562843182e-05,
+      "loss": 0.1768,
+      "step": 7284
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7553439793125453e-05,
+      "loss": 0.3112,
+      "step": 7285
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7553103957819088e-05,
+      "loss": 0.3341,
+      "step": 7286
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7552768122512723e-05,
+      "loss": 0.5107,
+      "step": 7287
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7552432287206357e-05,
+      "loss": 0.352,
+      "step": 7288
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.755209645189999e-05,
+      "loss": 0.1996,
+      "step": 7289
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7551760616593627e-05,
+      "loss": 0.1699,
+      "step": 7290
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7551424781287258e-05,
+      "loss": 0.4012,
+      "step": 7291
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7551088945980893e-05,
+      "loss": 0.361,
+      "step": 7292
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7550753110674527e-05,
+      "loss": 0.2633,
+      "step": 7293
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7550417275368162e-05,
+      "loss": 0.3976,
+      "step": 7294
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7550081440061793e-05,
+      "loss": 0.372,
+      "step": 7295
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.754974560475543e-05,
+      "loss": 0.3002,
+      "step": 7296
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7549409769449062e-05,
+      "loss": 0.2962,
+      "step": 7297
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7549073934142697e-05,
+      "loss": 0.3222,
+      "step": 7298
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.754873809883633e-05,
+      "loss": 0.3029,
+      "step": 7299
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7548402263529966e-05,
+      "loss": 0.4513,
+      "step": 7300
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.75480664282236e-05,
+      "loss": 0.2925,
+      "step": 7301
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7547730592917236e-05,
+      "loss": 0.3774,
+      "step": 7302
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.754739475761087e-05,
+      "loss": 0.2716,
+      "step": 7303
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.75470589223045e-05,
+      "loss": 0.4666,
+      "step": 7304
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.754672308699814e-05,
+      "loss": 0.204,
+      "step": 7305
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.754638725169177e-05,
+      "loss": 0.3408,
+      "step": 7306
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7546051416385405e-05,
+      "loss": 0.3263,
+      "step": 7307
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.754571558107904e-05,
+      "loss": 0.3754,
+      "step": 7308
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7545379745772675e-05,
+      "loss": 0.2971,
+      "step": 7309
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.754504391046631e-05,
+      "loss": 0.4974,
+      "step": 7310
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.754470807515994e-05,
+      "loss": 0.3825,
+      "step": 7311
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.754437223985358e-05,
+      "loss": 0.1663,
+      "step": 7312
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.754403640454721e-05,
+      "loss": 0.263,
+      "step": 7313
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7543700569240845e-05,
+      "loss": 0.2496,
+      "step": 7314
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.754336473393448e-05,
+      "loss": 0.3308,
+      "step": 7315
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7543028898628114e-05,
+      "loss": 0.3117,
+      "step": 7316
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.754269306332175e-05,
+      "loss": 0.2689,
+      "step": 7317
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7542357228015383e-05,
+      "loss": 0.4134,
+      "step": 7318
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7542021392709018e-05,
+      "loss": 0.1892,
+      "step": 7319
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.754168555740265e-05,
+      "loss": 0.409,
+      "step": 7320
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7541349722096287e-05,
+      "loss": 0.4275,
+      "step": 7321
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.754101388678992e-05,
+      "loss": 0.2955,
+      "step": 7322
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7540678051483553e-05,
+      "loss": 0.3895,
+      "step": 7323
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7540342216177188e-05,
+      "loss": 0.4204,
+      "step": 7324
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7540006380870822e-05,
+      "loss": 0.3077,
+      "step": 7325
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7539670545564457e-05,
+      "loss": 0.2449,
+      "step": 7326
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.753933471025809e-05,
+      "loss": 0.3437,
+      "step": 7327
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7538998874951726e-05,
+      "loss": 0.3543,
+      "step": 7328
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7538663039645358e-05,
+      "loss": 0.2065,
+      "step": 7329
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7538327204338996e-05,
+      "loss": 0.2056,
+      "step": 7330
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7537991369032627e-05,
+      "loss": 0.5658,
+      "step": 7331
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.753765553372626e-05,
+      "loss": 0.2867,
+      "step": 7332
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7537319698419896e-05,
+      "loss": 0.2888,
+      "step": 7333
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.753698386311353e-05,
+      "loss": 0.3952,
+      "step": 7334
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7536648027807165e-05,
+      "loss": 0.4223,
+      "step": 7335
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.75363121925008e-05,
+      "loss": 0.3498,
+      "step": 7336
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7535976357194435e-05,
+      "loss": 0.3197,
+      "step": 7337
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7535640521888066e-05,
+      "loss": 0.29,
+      "step": 7338
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.75353046865817e-05,
+      "loss": 0.3378,
+      "step": 7339
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7534968851275335e-05,
+      "loss": 0.2307,
+      "step": 7340
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.753463301596897e-05,
+      "loss": 0.4571,
+      "step": 7341
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7534297180662605e-05,
+      "loss": 0.228,
+      "step": 7342
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.753396134535624e-05,
+      "loss": 0.3093,
+      "step": 7343
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.753362551004987e-05,
+      "loss": 0.2386,
+      "step": 7344
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.753328967474351e-05,
+      "loss": 0.2864,
+      "step": 7345
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.753295383943714e-05,
+      "loss": 0.5462,
+      "step": 7346
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7532618004130774e-05,
+      "loss": 0.2805,
+      "step": 7347
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.753228216882441e-05,
+      "loss": 0.3364,
+      "step": 7348
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7531946333518044e-05,
+      "loss": 0.4775,
+      "step": 7349
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.753161049821168e-05,
+      "loss": 0.2545,
+      "step": 7350
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7531274662905313e-05,
+      "loss": 0.3218,
+      "step": 7351
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7530938827598948e-05,
+      "loss": 0.296,
+      "step": 7352
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.753060299229258e-05,
+      "loss": 0.2151,
+      "step": 7353
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7530267156986217e-05,
+      "loss": 0.5591,
+      "step": 7354
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7529931321679848e-05,
+      "loss": 0.3326,
+      "step": 7355
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7529595486373483e-05,
+      "loss": 0.5002,
+      "step": 7356
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7529259651067118e-05,
+      "loss": 0.2554,
+      "step": 7357
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7528923815760752e-05,
+      "loss": 0.441,
+      "step": 7358
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7528587980454387e-05,
+      "loss": 0.2427,
+      "step": 7359
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.752825214514802e-05,
+      "loss": 0.5106,
+      "step": 7360
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7527916309841656e-05,
+      "loss": 0.238,
+      "step": 7361
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7527580474535287e-05,
+      "loss": 0.4663,
+      "step": 7362
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7527244639228925e-05,
+      "loss": 0.2846,
+      "step": 7363
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7526908803922557e-05,
+      "loss": 0.4405,
+      "step": 7364
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.752657296861619e-05,
+      "loss": 0.3929,
+      "step": 7365
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7526237133309826e-05,
+      "loss": 0.1669,
+      "step": 7366
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.752590129800346e-05,
+      "loss": 0.3222,
+      "step": 7367
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7525565462697095e-05,
+      "loss": 0.2557,
+      "step": 7368
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.752522962739073e-05,
+      "loss": 0.2751,
+      "step": 7369
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7524893792084365e-05,
+      "loss": 0.3677,
+      "step": 7370
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7524557956777996e-05,
+      "loss": 0.1689,
+      "step": 7371
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7524222121471634e-05,
+      "loss": 0.2304,
+      "step": 7372
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7523886286165265e-05,
+      "loss": 0.2855,
+      "step": 7373
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.75235504508589e-05,
+      "loss": 0.3404,
+      "step": 7374
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7523214615552534e-05,
+      "loss": 0.3185,
+      "step": 7375
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.752287878024617e-05,
+      "loss": 0.4335,
+      "step": 7376
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7522542944939804e-05,
+      "loss": 0.3671,
+      "step": 7377
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.752220710963344e-05,
+      "loss": 0.5305,
+      "step": 7378
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7521871274327073e-05,
+      "loss": 0.372,
+      "step": 7379
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7521535439020704e-05,
+      "loss": 0.4467,
+      "step": 7380
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7521199603714342e-05,
+      "loss": 0.2446,
+      "step": 7381
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7520863768407974e-05,
+      "loss": 0.2366,
+      "step": 7382
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7520527933101608e-05,
+      "loss": 0.3304,
+      "step": 7383
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7520192097795243e-05,
+      "loss": 0.3946,
+      "step": 7384
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7519856262488877e-05,
+      "loss": 0.301,
+      "step": 7385
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.751952042718251e-05,
+      "loss": 0.4011,
+      "step": 7386
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7519184591876147e-05,
+      "loss": 0.4564,
+      "step": 7387
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7518848756569778e-05,
+      "loss": 0.376,
+      "step": 7388
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7518512921263413e-05,
+      "loss": 0.4519,
+      "step": 7389
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7518177085957047e-05,
+      "loss": 0.3304,
+      "step": 7390
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7517841250650682e-05,
+      "loss": 0.3591,
+      "step": 7391
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7517505415344317e-05,
+      "loss": 0.3051,
+      "step": 7392
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.751716958003795e-05,
+      "loss": 0.2724,
+      "step": 7393
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7516833744731586e-05,
+      "loss": 0.3376,
+      "step": 7394
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7516497909425217e-05,
+      "loss": 0.3651,
+      "step": 7395
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7516162074118855e-05,
+      "loss": 0.2019,
+      "step": 7396
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7515826238812486e-05,
+      "loss": 0.3687,
+      "step": 7397
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.751549040350612e-05,
+      "loss": 0.299,
+      "step": 7398
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7515154568199756e-05,
+      "loss": 0.3257,
+      "step": 7399
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.751481873289339e-05,
+      "loss": 0.2125,
+      "step": 7400
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7514482897587025e-05,
+      "loss": 0.3028,
+      "step": 7401
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.751414706228066e-05,
+      "loss": 0.3106,
+      "step": 7402
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7513811226974294e-05,
+      "loss": 0.2496,
+      "step": 7403
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7513475391667926e-05,
+      "loss": 0.2454,
+      "step": 7404
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7513139556361564e-05,
+      "loss": 0.3817,
+      "step": 7405
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7512803721055195e-05,
+      "loss": 0.2477,
+      "step": 7406
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.751246788574883e-05,
+      "loss": 0.3619,
+      "step": 7407
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7512132050442464e-05,
+      "loss": 0.2412,
+      "step": 7408
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.75117962151361e-05,
+      "loss": 0.342,
+      "step": 7409
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7511460379829733e-05,
+      "loss": 0.474,
+      "step": 7410
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7511124544523368e-05,
+      "loss": 0.3206,
+      "step": 7411
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7510788709217003e-05,
+      "loss": 0.3953,
+      "step": 7412
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7510452873910634e-05,
+      "loss": 0.3438,
+      "step": 7413
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7510117038604272e-05,
+      "loss": 0.2701,
+      "step": 7414
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7509781203297903e-05,
+      "loss": 0.442,
+      "step": 7415
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7509445367991538e-05,
+      "loss": 0.2644,
+      "step": 7416
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7509109532685173e-05,
+      "loss": 0.3832,
+      "step": 7417
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7508773697378807e-05,
+      "loss": 0.2871,
+      "step": 7418
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7508437862072442e-05,
+      "loss": 0.39,
+      "step": 7419
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7508102026766077e-05,
+      "loss": 0.6451,
+      "step": 7420
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.750776619145971e-05,
+      "loss": 0.4273,
+      "step": 7421
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7507430356153342e-05,
+      "loss": 0.2851,
+      "step": 7422
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.750709452084698e-05,
+      "loss": 0.3458,
+      "step": 7423
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7506758685540612e-05,
+      "loss": 0.3996,
+      "step": 7424
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7506422850234246e-05,
+      "loss": 0.2783,
+      "step": 7425
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.750608701492788e-05,
+      "loss": 0.5671,
+      "step": 7426
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7505751179621516e-05,
+      "loss": 0.321,
+      "step": 7427
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7505415344315147e-05,
+      "loss": 0.1844,
+      "step": 7428
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7505079509008785e-05,
+      "loss": 0.1611,
+      "step": 7429
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7504743673702416e-05,
+      "loss": 0.3522,
+      "step": 7430
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.750440783839605e-05,
+      "loss": 0.2386,
+      "step": 7431
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7504072003089686e-05,
+      "loss": 0.4107,
+      "step": 7432
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.750373616778332e-05,
+      "loss": 0.2545,
+      "step": 7433
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7503400332476955e-05,
+      "loss": 0.2947,
+      "step": 7434
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.750306449717059e-05,
+      "loss": 0.4706,
+      "step": 7435
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7502728661864224e-05,
+      "loss": 0.1941,
+      "step": 7436
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7502392826557855e-05,
+      "loss": 0.2689,
+      "step": 7437
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7502056991251493e-05,
+      "loss": 0.215,
+      "step": 7438
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7501721155945125e-05,
+      "loss": 0.1664,
+      "step": 7439
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.750138532063876e-05,
+      "loss": 0.2808,
+      "step": 7440
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7501049485332394e-05,
+      "loss": 0.4853,
+      "step": 7441
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.750071365002603e-05,
+      "loss": 0.3661,
+      "step": 7442
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7500377814719663e-05,
+      "loss": 0.3957,
+      "step": 7443
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.7500041979413298e-05,
+      "loss": 0.4907,
+      "step": 7444
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7499706144106933e-05,
+      "loss": 0.3105,
+      "step": 7445
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7499370308800564e-05,
+      "loss": 0.3297,
+      "step": 7446
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7499034473494202e-05,
+      "loss": 0.3428,
+      "step": 7447
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7498698638187833e-05,
+      "loss": 0.3997,
+      "step": 7448
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7498362802881468e-05,
+      "loss": 0.1914,
+      "step": 7449
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7498026967575102e-05,
+      "loss": 0.432,
+      "step": 7450
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7497691132268737e-05,
+      "loss": 0.3134,
+      "step": 7451
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7497355296962372e-05,
+      "loss": 0.3504,
+      "step": 7452
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7497019461656006e-05,
+      "loss": 0.4157,
+      "step": 7453
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.749668362634964e-05,
+      "loss": 0.3036,
+      "step": 7454
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7496347791043272e-05,
+      "loss": 0.4896,
+      "step": 7455
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.749601195573691e-05,
+      "loss": 0.334,
+      "step": 7456
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.749567612043054e-05,
+      "loss": 0.3036,
+      "step": 7457
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7495340285124176e-05,
+      "loss": 0.1565,
+      "step": 7458
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.749500444981781e-05,
+      "loss": 0.2815,
+      "step": 7459
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7494668614511446e-05,
+      "loss": 0.4148,
+      "step": 7460
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.749433277920508e-05,
+      "loss": 0.2781,
+      "step": 7461
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7493996943898715e-05,
+      "loss": 0.1321,
+      "step": 7462
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.749366110859235e-05,
+      "loss": 0.5327,
+      "step": 7463
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.749332527328598e-05,
+      "loss": 0.3589,
+      "step": 7464
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.749298943797962e-05,
+      "loss": 0.2792,
+      "step": 7465
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.749265360267325e-05,
+      "loss": 0.4871,
+      "step": 7466
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7492317767366885e-05,
+      "loss": 0.332,
+      "step": 7467
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.749198193206052e-05,
+      "loss": 0.3922,
+      "step": 7468
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7491646096754154e-05,
+      "loss": 0.1827,
+      "step": 7469
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.749131026144779e-05,
+      "loss": 0.1516,
+      "step": 7470
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7490974426141423e-05,
+      "loss": 0.5558,
+      "step": 7471
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7490638590835058e-05,
+      "loss": 0.3552,
+      "step": 7472
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.749030275552869e-05,
+      "loss": 0.4292,
+      "step": 7473
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7489966920222324e-05,
+      "loss": 0.1352,
+      "step": 7474
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.748963108491596e-05,
+      "loss": 0.4131,
+      "step": 7475
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7489295249609593e-05,
+      "loss": 0.3251,
+      "step": 7476
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7488959414303228e-05,
+      "loss": 0.3928,
+      "step": 7477
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7488623578996862e-05,
+      "loss": 0.2734,
+      "step": 7478
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7488287743690494e-05,
+      "loss": 0.2029,
+      "step": 7479
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7487951908384132e-05,
+      "loss": 0.404,
+      "step": 7480
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7487616073077763e-05,
+      "loss": 0.3833,
+      "step": 7481
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7487280237771398e-05,
+      "loss": 0.3072,
+      "step": 7482
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7486944402465032e-05,
+      "loss": 0.4173,
+      "step": 7483
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7486608567158667e-05,
+      "loss": 0.3202,
+      "step": 7484
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.74862727318523e-05,
+      "loss": 0.338,
+      "step": 7485
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7485936896545936e-05,
+      "loss": 0.2717,
+      "step": 7486
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.748560106123957e-05,
+      "loss": 0.2945,
+      "step": 7487
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7485265225933202e-05,
+      "loss": 0.2237,
+      "step": 7488
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.748492939062684e-05,
+      "loss": 0.4287,
+      "step": 7489
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.748459355532047e-05,
+      "loss": 0.3097,
+      "step": 7490
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7484257720014106e-05,
+      "loss": 0.3048,
+      "step": 7491
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.748392188470774e-05,
+      "loss": 0.4478,
+      "step": 7492
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7483586049401375e-05,
+      "loss": 0.3521,
+      "step": 7493
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.748325021409501e-05,
+      "loss": 0.3164,
+      "step": 7494
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.748291437878864e-05,
+      "loss": 0.3514,
+      "step": 7495
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.748257854348228e-05,
+      "loss": 0.354,
+      "step": 7496
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.748224270817591e-05,
+      "loss": 0.463,
+      "step": 7497
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7481906872869545e-05,
+      "loss": 0.4971,
+      "step": 7498
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.748157103756318e-05,
+      "loss": 0.3482,
+      "step": 7499
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7481235202256814e-05,
+      "loss": 0.3461,
+      "step": 7500
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.748089936695045e-05,
+      "loss": 0.4214,
+      "step": 7501
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7480563531644084e-05,
+      "loss": 0.2573,
+      "step": 7502
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.748022769633772e-05,
+      "loss": 0.2694,
+      "step": 7503
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.747989186103135e-05,
+      "loss": 0.4004,
+      "step": 7504
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7479556025724988e-05,
+      "loss": 0.2167,
+      "step": 7505
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.747922019041862e-05,
+      "loss": 0.3975,
+      "step": 7506
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7478884355112254e-05,
+      "loss": 0.4873,
+      "step": 7507
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7478548519805888e-05,
+      "loss": 0.4322,
+      "step": 7508
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7478212684499523e-05,
+      "loss": 0.279,
+      "step": 7509
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7477876849193158e-05,
+      "loss": 0.3206,
+      "step": 7510
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7477541013886792e-05,
+      "loss": 0.4251,
+      "step": 7511
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7477205178580427e-05,
+      "loss": 0.2326,
+      "step": 7512
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7476869343274058e-05,
+      "loss": 0.2394,
+      "step": 7513
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7476533507967696e-05,
+      "loss": 0.4391,
+      "step": 7514
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7476197672661327e-05,
+      "loss": 0.2777,
+      "step": 7515
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7475861837354962e-05,
+      "loss": 0.2745,
+      "step": 7516
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7475526002048597e-05,
+      "loss": 0.3065,
+      "step": 7517
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.747519016674223e-05,
+      "loss": 0.3277,
+      "step": 7518
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7474854331435863e-05,
+      "loss": 0.3226,
+      "step": 7519
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.74745184961295e-05,
+      "loss": 0.3542,
+      "step": 7520
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7474182660823132e-05,
+      "loss": 0.3048,
+      "step": 7521
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7473846825516767e-05,
+      "loss": 0.2804,
+      "step": 7522
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.74735109902104e-05,
+      "loss": 0.2416,
+      "step": 7523
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7473175154904036e-05,
+      "loss": 0.3075,
+      "step": 7524
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.747283931959767e-05,
+      "loss": 0.2361,
+      "step": 7525
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7472503484291305e-05,
+      "loss": 0.2684,
+      "step": 7526
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.747216764898494e-05,
+      "loss": 0.3972,
+      "step": 7527
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.747183181367857e-05,
+      "loss": 0.4965,
+      "step": 7528
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.747149597837221e-05,
+      "loss": 0.2901,
+      "step": 7529
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.747116014306584e-05,
+      "loss": 0.4367,
+      "step": 7530
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7470824307759475e-05,
+      "loss": 0.2497,
+      "step": 7531
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.747048847245311e-05,
+      "loss": 0.5089,
+      "step": 7532
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7470152637146744e-05,
+      "loss": 0.4547,
+      "step": 7533
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.746981680184038e-05,
+      "loss": 0.1823,
+      "step": 7534
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7469480966534014e-05,
+      "loss": 0.3478,
+      "step": 7535
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7469145131227648e-05,
+      "loss": 0.3359,
+      "step": 7536
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.746880929592128e-05,
+      "loss": 0.1761,
+      "step": 7537
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7468473460614918e-05,
+      "loss": 0.3587,
+      "step": 7538
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.746813762530855e-05,
+      "loss": 0.2638,
+      "step": 7539
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7467801790002183e-05,
+      "loss": 0.2785,
+      "step": 7540
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7467465954695818e-05,
+      "loss": 0.382,
+      "step": 7541
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7467130119389453e-05,
+      "loss": 0.3838,
+      "step": 7542
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7466794284083087e-05,
+      "loss": 0.3155,
+      "step": 7543
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7466458448776722e-05,
+      "loss": 0.1616,
+      "step": 7544
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7466122613470357e-05,
+      "loss": 0.1959,
+      "step": 7545
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7465786778163988e-05,
+      "loss": 0.3695,
+      "step": 7546
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7465450942857626e-05,
+      "loss": 0.317,
+      "step": 7547
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7465115107551257e-05,
+      "loss": 0.2996,
+      "step": 7548
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7464779272244892e-05,
+      "loss": 0.4049,
+      "step": 7549
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7464443436938527e-05,
+      "loss": 0.4852,
+      "step": 7550
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.746410760163216e-05,
+      "loss": 0.2438,
+      "step": 7551
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7463771766325796e-05,
+      "loss": 0.2873,
+      "step": 7552
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.746343593101943e-05,
+      "loss": 0.3091,
+      "step": 7553
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7463100095713065e-05,
+      "loss": 0.3666,
+      "step": 7554
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7462764260406696e-05,
+      "loss": 0.2126,
+      "step": 7555
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7462428425100334e-05,
+      "loss": 0.278,
+      "step": 7556
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7462092589793966e-05,
+      "loss": 0.3569,
+      "step": 7557
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.74617567544876e-05,
+      "loss": 0.2938,
+      "step": 7558
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7461420919181235e-05,
+      "loss": 0.337,
+      "step": 7559
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.746108508387487e-05,
+      "loss": 0.3795,
+      "step": 7560
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7460749248568504e-05,
+      "loss": 0.2892,
+      "step": 7561
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.746041341326214e-05,
+      "loss": 0.3118,
+      "step": 7562
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.746007757795577e-05,
+      "loss": 0.4124,
+      "step": 7563
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7459741742649405e-05,
+      "loss": 0.3965,
+      "step": 7564
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.745940590734304e-05,
+      "loss": 0.4162,
+      "step": 7565
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7459070072036674e-05,
+      "loss": 0.3228,
+      "step": 7566
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.745873423673031e-05,
+      "loss": 0.3947,
+      "step": 7567
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7458398401423943e-05,
+      "loss": 0.3336,
+      "step": 7568
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7458062566117578e-05,
+      "loss": 0.2468,
+      "step": 7569
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.745772673081121e-05,
+      "loss": 0.4144,
+      "step": 7570
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7457390895504847e-05,
+      "loss": 0.4337,
+      "step": 7571
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.745705506019848e-05,
+      "loss": 0.2512,
+      "step": 7572
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7456719224892113e-05,
+      "loss": 0.5259,
+      "step": 7573
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7456383389585748e-05,
+      "loss": 0.3249,
+      "step": 7574
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7456047554279383e-05,
+      "loss": 0.4201,
+      "step": 7575
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7455711718973017e-05,
+      "loss": 0.3012,
+      "step": 7576
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7455375883666652e-05,
+      "loss": 0.3676,
+      "step": 7577
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7455040048360286e-05,
+      "loss": 0.3932,
+      "step": 7578
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7454704213053918e-05,
+      "loss": 0.5059,
+      "step": 7579
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7454368377747556e-05,
+      "loss": 0.2835,
+      "step": 7580
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7454032542441187e-05,
+      "loss": 0.4737,
+      "step": 7581
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.745369670713482e-05,
+      "loss": 0.7066,
+      "step": 7582
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7453360871828456e-05,
+      "loss": 0.4906,
+      "step": 7583
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.745302503652209e-05,
+      "loss": 0.5006,
+      "step": 7584
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7452689201215726e-05,
+      "loss": 0.4055,
+      "step": 7585
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.745235336590936e-05,
+      "loss": 0.2907,
+      "step": 7586
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7452017530602995e-05,
+      "loss": 0.3858,
+      "step": 7587
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7451681695296626e-05,
+      "loss": 0.3575,
+      "step": 7588
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7451345859990264e-05,
+      "loss": 0.2329,
+      "step": 7589
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7451010024683895e-05,
+      "loss": 0.4002,
+      "step": 7590
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.745067418937753e-05,
+      "loss": 0.307,
+      "step": 7591
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7450338354071165e-05,
+      "loss": 0.2265,
+      "step": 7592
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.74500025187648e-05,
+      "loss": 0.4933,
+      "step": 7593
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7449666683458434e-05,
+      "loss": 0.2899,
+      "step": 7594
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.744933084815207e-05,
+      "loss": 0.2949,
+      "step": 7595
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7448995012845703e-05,
+      "loss": 0.4522,
+      "step": 7596
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7448659177539335e-05,
+      "loss": 0.2496,
+      "step": 7597
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7448323342232973e-05,
+      "loss": 0.4124,
+      "step": 7598
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7447987506926604e-05,
+      "loss": 0.3297,
+      "step": 7599
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.744765167162024e-05,
+      "loss": 0.2106,
+      "step": 7600
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7447315836313873e-05,
+      "loss": 0.3983,
+      "step": 7601
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7446980001007508e-05,
+      "loss": 0.3019,
+      "step": 7602
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7446644165701143e-05,
+      "loss": 0.4177,
+      "step": 7603
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7446308330394777e-05,
+      "loss": 0.2707,
+      "step": 7604
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7445972495088412e-05,
+      "loss": 0.2514,
+      "step": 7605
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7445636659782043e-05,
+      "loss": 0.4089,
+      "step": 7606
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.744530082447568e-05,
+      "loss": 0.5262,
+      "step": 7607
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7444964989169312e-05,
+      "loss": 0.3064,
+      "step": 7608
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7444629153862947e-05,
+      "loss": 0.4303,
+      "step": 7609
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.744429331855658e-05,
+      "loss": 0.2386,
+      "step": 7610
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7443957483250216e-05,
+      "loss": 0.2879,
+      "step": 7611
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7443621647943848e-05,
+      "loss": 0.3329,
+      "step": 7612
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7443285812637486e-05,
+      "loss": 0.3663,
+      "step": 7613
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7442949977331117e-05,
+      "loss": 0.3559,
+      "step": 7614
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.744261414202475e-05,
+      "loss": 0.4442,
+      "step": 7615
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7442278306718386e-05,
+      "loss": 0.2181,
+      "step": 7616
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.744194247141202e-05,
+      "loss": 0.274,
+      "step": 7617
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7441606636105655e-05,
+      "loss": 0.3511,
+      "step": 7618
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.744127080079929e-05,
+      "loss": 0.3442,
+      "step": 7619
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7440934965492925e-05,
+      "loss": 0.37,
+      "step": 7620
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7440599130186556e-05,
+      "loss": 0.2692,
+      "step": 7621
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7440263294880194e-05,
+      "loss": 0.1783,
+      "step": 7622
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7439927459573825e-05,
+      "loss": 0.2167,
+      "step": 7623
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.743959162426746e-05,
+      "loss": 0.2403,
+      "step": 7624
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7439255788961095e-05,
+      "loss": 0.3668,
+      "step": 7625
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.743891995365473e-05,
+      "loss": 0.2086,
+      "step": 7626
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7438584118348364e-05,
+      "loss": 0.4014,
+      "step": 7627
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7438248283042e-05,
+      "loss": 0.3083,
+      "step": 7628
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7437912447735633e-05,
+      "loss": 0.6242,
+      "step": 7629
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7437576612429264e-05,
+      "loss": 0.2972,
+      "step": 7630
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7437240777122902e-05,
+      "loss": 0.498,
+      "step": 7631
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7436904941816534e-05,
+      "loss": 0.288,
+      "step": 7632
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.743656910651017e-05,
+      "loss": 0.2219,
+      "step": 7633
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7436233271203803e-05,
+      "loss": 0.537,
+      "step": 7634
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7435897435897438e-05,
+      "loss": 0.3677,
+      "step": 7635
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7435561600591072e-05,
+      "loss": 0.2473,
+      "step": 7636
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7435225765284707e-05,
+      "loss": 0.3977,
+      "step": 7637
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.743488992997834e-05,
+      "loss": 0.3503,
+      "step": 7638
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7434554094671973e-05,
+      "loss": 0.3792,
+      "step": 7639
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.743421825936561e-05,
+      "loss": 0.1953,
+      "step": 7640
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7433882424059242e-05,
+      "loss": 0.5191,
+      "step": 7641
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.7433546588752877e-05,
+      "loss": 0.3745,
+      "step": 7642
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.743321075344651e-05,
+      "loss": 0.2697,
+      "step": 7643
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7432874918140146e-05,
+      "loss": 0.3836,
+      "step": 7644
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.743253908283378e-05,
+      "loss": 0.2198,
+      "step": 7645
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7432203247527415e-05,
+      "loss": 0.3732,
+      "step": 7646
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.743186741222105e-05,
+      "loss": 0.232,
+      "step": 7647
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.743153157691468e-05,
+      "loss": 0.3883,
+      "step": 7648
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.743119574160832e-05,
+      "loss": 0.3181,
+      "step": 7649
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.743085990630195e-05,
+      "loss": 0.3192,
+      "step": 7650
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7430524070995585e-05,
+      "loss": 0.405,
+      "step": 7651
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.743018823568922e-05,
+      "loss": 0.3259,
+      "step": 7652
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7429852400382855e-05,
+      "loss": 0.3278,
+      "step": 7653
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7429516565076486e-05,
+      "loss": 0.2644,
+      "step": 7654
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7429180729770124e-05,
+      "loss": 0.2691,
+      "step": 7655
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7428844894463755e-05,
+      "loss": 0.2631,
+      "step": 7656
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.742850905915739e-05,
+      "loss": 0.2957,
+      "step": 7657
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7428173223851024e-05,
+      "loss": 0.3253,
+      "step": 7658
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.742783738854466e-05,
+      "loss": 0.2936,
+      "step": 7659
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7427501553238294e-05,
+      "loss": 0.3574,
+      "step": 7660
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.742716571793193e-05,
+      "loss": 0.2587,
+      "step": 7661
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7426829882625563e-05,
+      "loss": 0.2766,
+      "step": 7662
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7426494047319194e-05,
+      "loss": 0.2669,
+      "step": 7663
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7426158212012832e-05,
+      "loss": 0.3084,
+      "step": 7664
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7425822376706464e-05,
+      "loss": 0.5718,
+      "step": 7665
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7425486541400098e-05,
+      "loss": 0.504,
+      "step": 7666
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7425150706093733e-05,
+      "loss": 0.2089,
+      "step": 7667
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7424814870787367e-05,
+      "loss": 0.4964,
+      "step": 7668
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7424479035481002e-05,
+      "loss": 0.2688,
+      "step": 7669
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7424143200174633e-05,
+      "loss": 0.502,
+      "step": 7670
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.742380736486827e-05,
+      "loss": 0.2809,
+      "step": 7671
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7423471529561903e-05,
+      "loss": 0.3151,
+      "step": 7672
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7423135694255537e-05,
+      "loss": 0.2322,
+      "step": 7673
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7422799858949172e-05,
+      "loss": 0.3267,
+      "step": 7674
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7422464023642807e-05,
+      "loss": 0.2272,
+      "step": 7675
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.742212818833644e-05,
+      "loss": 0.2734,
+      "step": 7676
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7421792353030076e-05,
+      "loss": 0.1999,
+      "step": 7677
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.742145651772371e-05,
+      "loss": 0.6005,
+      "step": 7678
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7421120682417342e-05,
+      "loss": 0.3562,
+      "step": 7679
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.742078484711098e-05,
+      "loss": 0.2593,
+      "step": 7680
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.742044901180461e-05,
+      "loss": 0.4693,
+      "step": 7681
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7420113176498246e-05,
+      "loss": 0.464,
+      "step": 7682
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.741977734119188e-05,
+      "loss": 0.2841,
+      "step": 7683
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7419441505885515e-05,
+      "loss": 0.3746,
+      "step": 7684
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.741910567057915e-05,
+      "loss": 0.4855,
+      "step": 7685
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7418769835272784e-05,
+      "loss": 0.3044,
+      "step": 7686
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.741843399996642e-05,
+      "loss": 0.3888,
+      "step": 7687
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.741809816466005e-05,
+      "loss": 0.311,
+      "step": 7688
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7417762329353688e-05,
+      "loss": 0.2586,
+      "step": 7689
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.741742649404732e-05,
+      "loss": 0.3033,
+      "step": 7690
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7417090658740954e-05,
+      "loss": 0.3639,
+      "step": 7691
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.741675482343459e-05,
+      "loss": 0.4722,
+      "step": 7692
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7416418988128224e-05,
+      "loss": 0.3881,
+      "step": 7693
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7416083152821858e-05,
+      "loss": 0.5552,
+      "step": 7694
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7415747317515493e-05,
+      "loss": 0.3111,
+      "step": 7695
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7415411482209127e-05,
+      "loss": 0.3619,
+      "step": 7696
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.741507564690276e-05,
+      "loss": 0.2969,
+      "step": 7697
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7414739811596393e-05,
+      "loss": 0.3068,
+      "step": 7698
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7414403976290028e-05,
+      "loss": 0.4519,
+      "step": 7699
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7414068140983663e-05,
+      "loss": 0.26,
+      "step": 7700
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7413732305677297e-05,
+      "loss": 0.2136,
+      "step": 7701
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7413396470370932e-05,
+      "loss": 0.3331,
+      "step": 7702
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7413060635064563e-05,
+      "loss": 0.3727,
+      "step": 7703
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.74127247997582e-05,
+      "loss": 0.2549,
+      "step": 7704
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7412388964451832e-05,
+      "loss": 0.1952,
+      "step": 7705
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7412053129145467e-05,
+      "loss": 0.2926,
+      "step": 7706
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7411717293839102e-05,
+      "loss": 0.3571,
+      "step": 7707
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7411381458532736e-05,
+      "loss": 0.4346,
+      "step": 7708
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.741104562322637e-05,
+      "loss": 0.3979,
+      "step": 7709
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7410709787920006e-05,
+      "loss": 0.2855,
+      "step": 7710
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.741037395261364e-05,
+      "loss": 0.1828,
+      "step": 7711
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.741003811730727e-05,
+      "loss": 0.5191,
+      "step": 7712
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.740970228200091e-05,
+      "loss": 0.3247,
+      "step": 7713
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.740936644669454e-05,
+      "loss": 0.4267,
+      "step": 7714
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7409030611388176e-05,
+      "loss": 0.3941,
+      "step": 7715
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.740869477608181e-05,
+      "loss": 0.5286,
+      "step": 7716
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7408358940775445e-05,
+      "loss": 0.4318,
+      "step": 7717
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.740802310546908e-05,
+      "loss": 0.4819,
+      "step": 7718
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7407687270162714e-05,
+      "loss": 0.2103,
+      "step": 7719
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.740735143485635e-05,
+      "loss": 0.2124,
+      "step": 7720
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.740701559954998e-05,
+      "loss": 0.2449,
+      "step": 7721
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7406679764243618e-05,
+      "loss": 0.4488,
+      "step": 7722
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.740634392893725e-05,
+      "loss": 0.5335,
+      "step": 7723
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7406008093630884e-05,
+      "loss": 0.2632,
+      "step": 7724
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.740567225832452e-05,
+      "loss": 0.4998,
+      "step": 7725
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7405336423018153e-05,
+      "loss": 0.3426,
+      "step": 7726
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7405000587711788e-05,
+      "loss": 0.4265,
+      "step": 7727
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7404664752405423e-05,
+      "loss": 0.3717,
+      "step": 7728
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7404328917099057e-05,
+      "loss": 0.3295,
+      "step": 7729
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.740399308179269e-05,
+      "loss": 0.3252,
+      "step": 7730
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7403657246486327e-05,
+      "loss": 0.216,
+      "step": 7731
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7403321411179958e-05,
+      "loss": 0.3611,
+      "step": 7732
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7402985575873592e-05,
+      "loss": 0.3847,
+      "step": 7733
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7402649740567227e-05,
+      "loss": 0.3886,
+      "step": 7734
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7402313905260862e-05,
+      "loss": 0.4655,
+      "step": 7735
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7401978069954496e-05,
+      "loss": 0.4058,
+      "step": 7736
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.740164223464813e-05,
+      "loss": 0.5313,
+      "step": 7737
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7401306399341766e-05,
+      "loss": 0.3864,
+      "step": 7738
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7400970564035397e-05,
+      "loss": 0.5343,
+      "step": 7739
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7400634728729035e-05,
+      "loss": 0.3553,
+      "step": 7740
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7400298893422666e-05,
+      "loss": 0.3206,
+      "step": 7741
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.73999630581163e-05,
+      "loss": 0.5434,
+      "step": 7742
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7399627222809936e-05,
+      "loss": 0.3999,
+      "step": 7743
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.739929138750357e-05,
+      "loss": 0.3615,
+      "step": 7744
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.73989555521972e-05,
+      "loss": 0.2759,
+      "step": 7745
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.739861971689084e-05,
+      "loss": 0.2728,
+      "step": 7746
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.739828388158447e-05,
+      "loss": 0.3172,
+      "step": 7747
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7397948046278105e-05,
+      "loss": 0.3076,
+      "step": 7748
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.739761221097174e-05,
+      "loss": 0.356,
+      "step": 7749
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7397276375665375e-05,
+      "loss": 0.2764,
+      "step": 7750
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.739694054035901e-05,
+      "loss": 0.1959,
+      "step": 7751
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7396604705052644e-05,
+      "loss": 0.3073,
+      "step": 7752
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.739626886974628e-05,
+      "loss": 0.3719,
+      "step": 7753
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.739593303443991e-05,
+      "loss": 0.3343,
+      "step": 7754
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7395597199133548e-05,
+      "loss": 0.3531,
+      "step": 7755
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.739526136382718e-05,
+      "loss": 0.4775,
+      "step": 7756
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7394925528520814e-05,
+      "loss": 0.1753,
+      "step": 7757
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.739458969321445e-05,
+      "loss": 0.2518,
+      "step": 7758
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7394253857908083e-05,
+      "loss": 0.3266,
+      "step": 7759
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7393918022601718e-05,
+      "loss": 0.1766,
+      "step": 7760
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7393582187295352e-05,
+      "loss": 0.4852,
+      "step": 7761
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7393246351988987e-05,
+      "loss": 0.5194,
+      "step": 7762
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.739291051668262e-05,
+      "loss": 0.2894,
+      "step": 7763
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7392574681376256e-05,
+      "loss": 0.4191,
+      "step": 7764
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7392238846069888e-05,
+      "loss": 0.3708,
+      "step": 7765
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7391903010763522e-05,
+      "loss": 0.3403,
+      "step": 7766
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7391567175457157e-05,
+      "loss": 0.2813,
+      "step": 7767
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.739123134015079e-05,
+      "loss": 0.4821,
+      "step": 7768
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7390895504844426e-05,
+      "loss": 0.3285,
+      "step": 7769
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.739055966953806e-05,
+      "loss": 0.4846,
+      "step": 7770
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7390223834231696e-05,
+      "loss": 0.1737,
+      "step": 7771
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7389887998925327e-05,
+      "loss": 0.3898,
+      "step": 7772
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7389552163618965e-05,
+      "loss": 0.5057,
+      "step": 7773
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7389216328312596e-05,
+      "loss": 0.274,
+      "step": 7774
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.738888049300623e-05,
+      "loss": 0.3499,
+      "step": 7775
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7388544657699865e-05,
+      "loss": 0.2508,
+      "step": 7776
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.73882088223935e-05,
+      "loss": 0.4157,
+      "step": 7777
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7387872987087135e-05,
+      "loss": 0.3785,
+      "step": 7778
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.738753715178077e-05,
+      "loss": 0.3135,
+      "step": 7779
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7387201316474404e-05,
+      "loss": 0.2972,
+      "step": 7780
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7386865481168035e-05,
+      "loss": 0.3227,
+      "step": 7781
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7386529645861673e-05,
+      "loss": 0.3097,
+      "step": 7782
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7386193810555304e-05,
+      "loss": 0.3643,
+      "step": 7783
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.738585797524894e-05,
+      "loss": 0.2322,
+      "step": 7784
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7385522139942574e-05,
+      "loss": 0.3753,
+      "step": 7785
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.738518630463621e-05,
+      "loss": 0.2634,
+      "step": 7786
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7384850469329843e-05,
+      "loss": 0.2036,
+      "step": 7787
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7384514634023478e-05,
+      "loss": 0.3103,
+      "step": 7788
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.738417879871711e-05,
+      "loss": 0.3603,
+      "step": 7789
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7383842963410744e-05,
+      "loss": 0.3304,
+      "step": 7790
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7383507128104378e-05,
+      "loss": 0.3567,
+      "step": 7791
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7383171292798013e-05,
+      "loss": 0.4198,
+      "step": 7792
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7382835457491648e-05,
+      "loss": 0.3664,
+      "step": 7793
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7382499622185282e-05,
+      "loss": 0.3375,
+      "step": 7794
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7382163786878917e-05,
+      "loss": 0.3352,
+      "step": 7795
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7381827951572548e-05,
+      "loss": 0.2918,
+      "step": 7796
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7381492116266186e-05,
+      "loss": 0.3071,
+      "step": 7797
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7381156280959817e-05,
+      "loss": 0.2755,
+      "step": 7798
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7380820445653452e-05,
+      "loss": 0.3157,
+      "step": 7799
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7380484610347087e-05,
+      "loss": 0.3998,
+      "step": 7800
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.738014877504072e-05,
+      "loss": 0.3245,
+      "step": 7801
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7379812939734356e-05,
+      "loss": 0.3042,
+      "step": 7802
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.737947710442799e-05,
+      "loss": 0.4836,
+      "step": 7803
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7379141269121625e-05,
+      "loss": 0.4763,
+      "step": 7804
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7378805433815257e-05,
+      "loss": 0.2944,
+      "step": 7805
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7378469598508895e-05,
+      "loss": 0.5068,
+      "step": 7806
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7378133763202526e-05,
+      "loss": 0.3991,
+      "step": 7807
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.737779792789616e-05,
+      "loss": 0.3097,
+      "step": 7808
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7377462092589795e-05,
+      "loss": 0.232,
+      "step": 7809
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.737712625728343e-05,
+      "loss": 0.4228,
+      "step": 7810
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7376790421977064e-05,
+      "loss": 0.286,
+      "step": 7811
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.73764545866707e-05,
+      "loss": 0.3133,
+      "step": 7812
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7376118751364334e-05,
+      "loss": 0.3639,
+      "step": 7813
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7375782916057965e-05,
+      "loss": 0.2155,
+      "step": 7814
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7375447080751603e-05,
+      "loss": 0.4593,
+      "step": 7815
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7375111245445234e-05,
+      "loss": 0.359,
+      "step": 7816
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.737477541013887e-05,
+      "loss": 0.2564,
+      "step": 7817
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7374439574832504e-05,
+      "loss": 0.4628,
+      "step": 7818
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7374103739526138e-05,
+      "loss": 0.5431,
+      "step": 7819
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7373767904219773e-05,
+      "loss": 0.3733,
+      "step": 7820
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7373432068913408e-05,
+      "loss": 0.3652,
+      "step": 7821
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7373096233607042e-05,
+      "loss": 0.3581,
+      "step": 7822
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7372760398300673e-05,
+      "loss": 0.2589,
+      "step": 7823
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.737242456299431e-05,
+      "loss": 0.3604,
+      "step": 7824
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7372088727687943e-05,
+      "loss": 0.4307,
+      "step": 7825
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7371752892381577e-05,
+      "loss": 0.2577,
+      "step": 7826
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7371417057075212e-05,
+      "loss": 0.2928,
+      "step": 7827
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7371081221768847e-05,
+      "loss": 0.3846,
+      "step": 7828
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.737074538646248e-05,
+      "loss": 0.3204,
+      "step": 7829
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7370409551156116e-05,
+      "loss": 0.3317,
+      "step": 7830
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.737007371584975e-05,
+      "loss": 0.2419,
+      "step": 7831
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7369737880543382e-05,
+      "loss": 0.5356,
+      "step": 7832
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7369402045237017e-05,
+      "loss": 0.6417,
+      "step": 7833
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.736906620993065e-05,
+      "loss": 0.4028,
+      "step": 7834
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7368730374624286e-05,
+      "loss": 0.2292,
+      "step": 7835
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.736839453931792e-05,
+      "loss": 0.3658,
+      "step": 7836
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7368058704011555e-05,
+      "loss": 0.3232,
+      "step": 7837
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7367722868705186e-05,
+      "loss": 0.4843,
+      "step": 7838
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7367387033398824e-05,
+      "loss": 0.2053,
+      "step": 7839
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.7367051198092456e-05,
+      "loss": 0.2953,
+      "step": 7840
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.736671536278609e-05,
+      "loss": 0.3216,
+      "step": 7841
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7366379527479725e-05,
+      "loss": 0.3361,
+      "step": 7842
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.736604369217336e-05,
+      "loss": 0.3814,
+      "step": 7843
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7365707856866994e-05,
+      "loss": 0.397,
+      "step": 7844
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.736537202156063e-05,
+      "loss": 0.3617,
+      "step": 7845
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7365036186254264e-05,
+      "loss": 0.303,
+      "step": 7846
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7364700350947895e-05,
+      "loss": 0.3882,
+      "step": 7847
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7364364515641533e-05,
+      "loss": 0.2282,
+      "step": 7848
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7364028680335164e-05,
+      "loss": 0.3865,
+      "step": 7849
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.73636928450288e-05,
+      "loss": 0.2281,
+      "step": 7850
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7363357009722433e-05,
+      "loss": 0.2392,
+      "step": 7851
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7363021174416068e-05,
+      "loss": 0.2397,
+      "step": 7852
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7362685339109703e-05,
+      "loss": 0.2133,
+      "step": 7853
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7362349503803334e-05,
+      "loss": 0.3956,
+      "step": 7854
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7362013668496972e-05,
+      "loss": 0.3429,
+      "step": 7855
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7361677833190603e-05,
+      "loss": 0.3647,
+      "step": 7856
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7361341997884238e-05,
+      "loss": 0.6244,
+      "step": 7857
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7361006162577873e-05,
+      "loss": 0.6256,
+      "step": 7858
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7360670327271507e-05,
+      "loss": 0.3704,
+      "step": 7859
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7360334491965142e-05,
+      "loss": 0.2596,
+      "step": 7860
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7359998656658777e-05,
+      "loss": 0.2837,
+      "step": 7861
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.735966282135241e-05,
+      "loss": 0.3872,
+      "step": 7862
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7359326986046042e-05,
+      "loss": 0.394,
+      "step": 7863
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.735899115073968e-05,
+      "loss": 0.3177,
+      "step": 7864
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7358655315433312e-05,
+      "loss": 0.477,
+      "step": 7865
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7358319480126946e-05,
+      "loss": 0.4007,
+      "step": 7866
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.735798364482058e-05,
+      "loss": 0.3481,
+      "step": 7867
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7357647809514216e-05,
+      "loss": 0.2928,
+      "step": 7868
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.735731197420785e-05,
+      "loss": 0.3128,
+      "step": 7869
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7356976138901485e-05,
+      "loss": 0.4224,
+      "step": 7870
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.735664030359512e-05,
+      "loss": 0.5399,
+      "step": 7871
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.735630446828875e-05,
+      "loss": 0.4265,
+      "step": 7872
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.735596863298239e-05,
+      "loss": 0.3395,
+      "step": 7873
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.735563279767602e-05,
+      "loss": 0.335,
+      "step": 7874
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7355296962369655e-05,
+      "loss": 0.1782,
+      "step": 7875
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.735496112706329e-05,
+      "loss": 0.2759,
+      "step": 7876
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7354625291756924e-05,
+      "loss": 0.3221,
+      "step": 7877
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.735428945645056e-05,
+      "loss": 0.2471,
+      "step": 7878
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7353953621144193e-05,
+      "loss": 0.3493,
+      "step": 7879
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7353617785837825e-05,
+      "loss": 0.3058,
+      "step": 7880
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.735328195053146e-05,
+      "loss": 0.3362,
+      "step": 7881
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7352946115225094e-05,
+      "loss": 0.3152,
+      "step": 7882
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.735261027991873e-05,
+      "loss": 0.2077,
+      "step": 7883
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7352274444612363e-05,
+      "loss": 0.2456,
+      "step": 7884
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7351938609305998e-05,
+      "loss": 0.32,
+      "step": 7885
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7351602773999633e-05,
+      "loss": 0.2621,
+      "step": 7886
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7351266938693264e-05,
+      "loss": 0.3185,
+      "step": 7887
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7350931103386902e-05,
+      "loss": 0.2981,
+      "step": 7888
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7350595268080533e-05,
+      "loss": 0.2113,
+      "step": 7889
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7350259432774168e-05,
+      "loss": 0.3675,
+      "step": 7890
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7349923597467802e-05,
+      "loss": 0.2592,
+      "step": 7891
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7349587762161437e-05,
+      "loss": 0.3433,
+      "step": 7892
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.734925192685507e-05,
+      "loss": 0.3265,
+      "step": 7893
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7348916091548706e-05,
+      "loss": 0.3196,
+      "step": 7894
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.734858025624234e-05,
+      "loss": 0.3615,
+      "step": 7895
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7348244420935972e-05,
+      "loss": 0.3678,
+      "step": 7896
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.734790858562961e-05,
+      "loss": 0.4069,
+      "step": 7897
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.734757275032324e-05,
+      "loss": 0.2785,
+      "step": 7898
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7347236915016876e-05,
+      "loss": 0.2318,
+      "step": 7899
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.734690107971051e-05,
+      "loss": 0.3238,
+      "step": 7900
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7346565244404145e-05,
+      "loss": 0.439,
+      "step": 7901
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.734622940909778e-05,
+      "loss": 0.4496,
+      "step": 7902
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7345893573791415e-05,
+      "loss": 0.2954,
+      "step": 7903
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.734555773848505e-05,
+      "loss": 0.4312,
+      "step": 7904
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.734522190317868e-05,
+      "loss": 0.3257,
+      "step": 7905
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.734488606787232e-05,
+      "loss": 0.3138,
+      "step": 7906
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.734455023256595e-05,
+      "loss": 0.5727,
+      "step": 7907
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7344214397259585e-05,
+      "loss": 0.6068,
+      "step": 7908
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.734387856195322e-05,
+      "loss": 0.2506,
+      "step": 7909
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7343542726646854e-05,
+      "loss": 0.3312,
+      "step": 7910
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.734320689134049e-05,
+      "loss": 0.3977,
+      "step": 7911
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7342871056034123e-05,
+      "loss": 0.3893,
+      "step": 7912
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7342535220727758e-05,
+      "loss": 0.3591,
+      "step": 7913
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.734219938542139e-05,
+      "loss": 0.3056,
+      "step": 7914
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7341863550115027e-05,
+      "loss": 0.2163,
+      "step": 7915
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.734152771480866e-05,
+      "loss": 0.2701,
+      "step": 7916
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7341191879502293e-05,
+      "loss": 0.3018,
+      "step": 7917
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7340856044195928e-05,
+      "loss": 0.3879,
+      "step": 7918
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7340520208889562e-05,
+      "loss": 0.3752,
+      "step": 7919
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7340184373583197e-05,
+      "loss": 0.4222,
+      "step": 7920
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.733984853827683e-05,
+      "loss": 0.2825,
+      "step": 7921
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7339512702970466e-05,
+      "loss": 0.3871,
+      "step": 7922
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7339176867664098e-05,
+      "loss": 0.3295,
+      "step": 7923
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7338841032357732e-05,
+      "loss": 0.3694,
+      "step": 7924
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7338505197051367e-05,
+      "loss": 0.3613,
+      "step": 7925
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7338169361745e-05,
+      "loss": 0.2493,
+      "step": 7926
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7337833526438636e-05,
+      "loss": 0.5374,
+      "step": 7927
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.733749769113227e-05,
+      "loss": 0.389,
+      "step": 7928
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7337161855825902e-05,
+      "loss": 0.3901,
+      "step": 7929
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.733682602051954e-05,
+      "loss": 0.2757,
+      "step": 7930
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.733649018521317e-05,
+      "loss": 0.4306,
+      "step": 7931
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7336154349906806e-05,
+      "loss": 0.4532,
+      "step": 7932
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.733581851460044e-05,
+      "loss": 0.3159,
+      "step": 7933
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7335482679294075e-05,
+      "loss": 0.5027,
+      "step": 7934
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.733514684398771e-05,
+      "loss": 0.2521,
+      "step": 7935
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7334811008681345e-05,
+      "loss": 0.311,
+      "step": 7936
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.733447517337498e-05,
+      "loss": 0.3147,
+      "step": 7937
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.733413933806861e-05,
+      "loss": 0.3203,
+      "step": 7938
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.733380350276225e-05,
+      "loss": 0.3279,
+      "step": 7939
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.733346766745588e-05,
+      "loss": 0.165,
+      "step": 7940
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7333131832149514e-05,
+      "loss": 0.4221,
+      "step": 7941
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.733279599684315e-05,
+      "loss": 0.2239,
+      "step": 7942
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7332460161536784e-05,
+      "loss": 0.2275,
+      "step": 7943
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.733212432623042e-05,
+      "loss": 0.1932,
+      "step": 7944
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7331788490924053e-05,
+      "loss": 0.3167,
+      "step": 7945
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7331452655617688e-05,
+      "loss": 0.2711,
+      "step": 7946
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.733111682031132e-05,
+      "loss": 0.3127,
+      "step": 7947
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7330780985004957e-05,
+      "loss": 0.354,
+      "step": 7948
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7330445149698588e-05,
+      "loss": 0.4154,
+      "step": 7949
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7330109314392223e-05,
+      "loss": 0.3287,
+      "step": 7950
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7329773479085857e-05,
+      "loss": 0.2726,
+      "step": 7951
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7329437643779492e-05,
+      "loss": 0.3628,
+      "step": 7952
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7329101808473127e-05,
+      "loss": 0.251,
+      "step": 7953
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.732876597316676e-05,
+      "loss": 0.3748,
+      "step": 7954
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7328430137860396e-05,
+      "loss": 0.3486,
+      "step": 7955
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7328094302554027e-05,
+      "loss": 0.4279,
+      "step": 7956
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7327758467247665e-05,
+      "loss": 0.4975,
+      "step": 7957
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7327422631941297e-05,
+      "loss": 0.3407,
+      "step": 7958
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.732708679663493e-05,
+      "loss": 0.2452,
+      "step": 7959
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7326750961328566e-05,
+      "loss": 0.4543,
+      "step": 7960
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.73264151260222e-05,
+      "loss": 0.2509,
+      "step": 7961
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7326079290715835e-05,
+      "loss": 0.2723,
+      "step": 7962
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.732574345540947e-05,
+      "loss": 0.573,
+      "step": 7963
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7325407620103105e-05,
+      "loss": 0.3521,
+      "step": 7964
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7325071784796736e-05,
+      "loss": 0.3358,
+      "step": 7965
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7324735949490374e-05,
+      "loss": 0.4699,
+      "step": 7966
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7324400114184005e-05,
+      "loss": 0.2784,
+      "step": 7967
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.732406427887764e-05,
+      "loss": 0.328,
+      "step": 7968
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7323728443571274e-05,
+      "loss": 0.2289,
+      "step": 7969
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.732339260826491e-05,
+      "loss": 0.2555,
+      "step": 7970
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.732305677295854e-05,
+      "loss": 0.3168,
+      "step": 7971
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.732272093765218e-05,
+      "loss": 0.2606,
+      "step": 7972
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.732238510234581e-05,
+      "loss": 0.3543,
+      "step": 7973
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7322049267039444e-05,
+      "loss": 0.2528,
+      "step": 7974
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.732171343173308e-05,
+      "loss": 0.3096,
+      "step": 7975
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7321377596426714e-05,
+      "loss": 0.3989,
+      "step": 7976
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7321041761120348e-05,
+      "loss": 0.3161,
+      "step": 7977
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7320705925813983e-05,
+      "loss": 0.3663,
+      "step": 7978
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7320370090507617e-05,
+      "loss": 0.3197,
+      "step": 7979
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.732003425520125e-05,
+      "loss": 0.4905,
+      "step": 7980
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7319698419894887e-05,
+      "loss": 0.4307,
+      "step": 7981
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7319362584588518e-05,
+      "loss": 0.3135,
+      "step": 7982
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7319026749282153e-05,
+      "loss": 0.4881,
+      "step": 7983
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7318690913975787e-05,
+      "loss": 0.4671,
+      "step": 7984
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7318355078669422e-05,
+      "loss": 0.3185,
+      "step": 7985
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7318019243363057e-05,
+      "loss": 0.4158,
+      "step": 7986
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.731768340805669e-05,
+      "loss": 0.3138,
+      "step": 7987
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7317347572750326e-05,
+      "loss": 0.4431,
+      "step": 7988
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7317011737443957e-05,
+      "loss": 0.333,
+      "step": 7989
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7316675902137595e-05,
+      "loss": 0.5138,
+      "step": 7990
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7316340066831226e-05,
+      "loss": 0.3581,
+      "step": 7991
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.731600423152486e-05,
+      "loss": 0.2648,
+      "step": 7992
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7315668396218496e-05,
+      "loss": 0.2195,
+      "step": 7993
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.731533256091213e-05,
+      "loss": 0.2513,
+      "step": 7994
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7314996725605765e-05,
+      "loss": 0.244,
+      "step": 7995
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.73146608902994e-05,
+      "loss": 0.4977,
+      "step": 7996
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7314325054993034e-05,
+      "loss": 0.3519,
+      "step": 7997
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7313989219686666e-05,
+      "loss": 0.4263,
+      "step": 7998
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7313653384380304e-05,
+      "loss": 0.3038,
+      "step": 7999
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7313317549073935e-05,
+      "loss": 0.3293,
+      "step": 8000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.731298171376757e-05,
+      "loss": 0.4932,
+      "step": 8001
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7312645878461204e-05,
+      "loss": 0.2884,
+      "step": 8002
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.731231004315484e-05,
+      "loss": 0.1239,
+      "step": 8003
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7311974207848473e-05,
+      "loss": 0.587,
+      "step": 8004
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7311638372542108e-05,
+      "loss": 0.2704,
+      "step": 8005
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7311302537235743e-05,
+      "loss": 0.5051,
+      "step": 8006
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7310966701929374e-05,
+      "loss": 0.2651,
+      "step": 8007
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7310630866623012e-05,
+      "loss": 0.2696,
+      "step": 8008
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7310295031316643e-05,
+      "loss": 0.3957,
+      "step": 8009
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7309959196010278e-05,
+      "loss": 0.4007,
+      "step": 8010
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7309623360703913e-05,
+      "loss": 0.2667,
+      "step": 8011
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7309287525397547e-05,
+      "loss": 0.2923,
+      "step": 8012
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7308951690091182e-05,
+      "loss": 0.3128,
+      "step": 8013
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7308615854784817e-05,
+      "loss": 0.231,
+      "step": 8014
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7308280019478448e-05,
+      "loss": 0.495,
+      "step": 8015
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7307944184172082e-05,
+      "loss": 0.2979,
+      "step": 8016
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7307608348865717e-05,
+      "loss": 0.3063,
+      "step": 8017
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7307272513559352e-05,
+      "loss": 0.3494,
+      "step": 8018
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7306936678252986e-05,
+      "loss": 0.4117,
+      "step": 8019
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.730660084294662e-05,
+      "loss": 0.4643,
+      "step": 8020
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7306265007640256e-05,
+      "loss": 0.3378,
+      "step": 8021
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7305929172333887e-05,
+      "loss": 0.3157,
+      "step": 8022
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7305593337027525e-05,
+      "loss": 0.2971,
+      "step": 8023
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7305257501721156e-05,
+      "loss": 0.3391,
+      "step": 8024
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.730492166641479e-05,
+      "loss": 0.3781,
+      "step": 8025
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7304585831108426e-05,
+      "loss": 0.3871,
+      "step": 8026
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.730424999580206e-05,
+      "loss": 0.4438,
+      "step": 8027
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7303914160495695e-05,
+      "loss": 0.2299,
+      "step": 8028
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.730357832518933e-05,
+      "loss": 0.4919,
+      "step": 8029
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7303242489882964e-05,
+      "loss": 0.2956,
+      "step": 8030
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7302906654576595e-05,
+      "loss": 0.3652,
+      "step": 8031
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.730257081927023e-05,
+      "loss": 0.273,
+      "step": 8032
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7302234983963865e-05,
+      "loss": 0.376,
+      "step": 8033
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.73018991486575e-05,
+      "loss": 0.3003,
+      "step": 8034
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7301563313351134e-05,
+      "loss": 0.3957,
+      "step": 8035
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.730122747804477e-05,
+      "loss": 0.4359,
+      "step": 8036
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7300891642738403e-05,
+      "loss": 0.2858,
+      "step": 8037
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7300555807432035e-05,
+      "loss": 0.1947,
+      "step": 8038
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.7300219972125673e-05,
+      "loss": 0.3379,
+      "step": 8039
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7299884136819304e-05,
+      "loss": 0.3929,
+      "step": 8040
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.729954830151294e-05,
+      "loss": 0.2989,
+      "step": 8041
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7299212466206573e-05,
+      "loss": 0.3154,
+      "step": 8042
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7298876630900208e-05,
+      "loss": 0.1907,
+      "step": 8043
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7298540795593842e-05,
+      "loss": 0.4403,
+      "step": 8044
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7298204960287477e-05,
+      "loss": 0.2154,
+      "step": 8045
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7297869124981112e-05,
+      "loss": 0.3459,
+      "step": 8046
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7297533289674743e-05,
+      "loss": 0.2448,
+      "step": 8047
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.729719745436838e-05,
+      "loss": 0.4564,
+      "step": 8048
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7296861619062012e-05,
+      "loss": 0.1528,
+      "step": 8049
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7296525783755647e-05,
+      "loss": 0.5699,
+      "step": 8050
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.729618994844928e-05,
+      "loss": 0.2861,
+      "step": 8051
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7295854113142916e-05,
+      "loss": 0.3722,
+      "step": 8052
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.729551827783655e-05,
+      "loss": 0.356,
+      "step": 8053
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7295182442530186e-05,
+      "loss": 0.3384,
+      "step": 8054
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.729484660722382e-05,
+      "loss": 0.528,
+      "step": 8055
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.729451077191745e-05,
+      "loss": 0.5207,
+      "step": 8056
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.729417493661109e-05,
+      "loss": 0.3756,
+      "step": 8057
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.729383910130472e-05,
+      "loss": 0.2328,
+      "step": 8058
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7293503265998355e-05,
+      "loss": 0.2584,
+      "step": 8059
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.729316743069199e-05,
+      "loss": 0.3054,
+      "step": 8060
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7292831595385625e-05,
+      "loss": 0.4183,
+      "step": 8061
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7292495760079256e-05,
+      "loss": 0.3929,
+      "step": 8062
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7292159924772894e-05,
+      "loss": 0.4098,
+      "step": 8063
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7291824089466525e-05,
+      "loss": 0.3997,
+      "step": 8064
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.729148825416016e-05,
+      "loss": 0.2384,
+      "step": 8065
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7291152418853795e-05,
+      "loss": 0.3489,
+      "step": 8066
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.729081658354743e-05,
+      "loss": 0.4065,
+      "step": 8067
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7290480748241064e-05,
+      "loss": 0.4008,
+      "step": 8068
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.72901449129347e-05,
+      "loss": 0.307,
+      "step": 8069
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7289809077628333e-05,
+      "loss": 0.3348,
+      "step": 8070
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7289473242321964e-05,
+      "loss": 0.4166,
+      "step": 8071
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7289137407015602e-05,
+      "loss": 0.2297,
+      "step": 8072
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7288801571709234e-05,
+      "loss": 0.3594,
+      "step": 8073
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7288465736402868e-05,
+      "loss": 0.2309,
+      "step": 8074
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7288129901096503e-05,
+      "loss": 0.2494,
+      "step": 8075
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7287794065790138e-05,
+      "loss": 0.374,
+      "step": 8076
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7287458230483772e-05,
+      "loss": 0.4379,
+      "step": 8077
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7287122395177407e-05,
+      "loss": 0.3886,
+      "step": 8078
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.728678655987104e-05,
+      "loss": 0.2777,
+      "step": 8079
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7286450724564673e-05,
+      "loss": 0.2639,
+      "step": 8080
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.728611488925831e-05,
+      "loss": 0.5412,
+      "step": 8081
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7285779053951942e-05,
+      "loss": 0.2566,
+      "step": 8082
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7285443218645577e-05,
+      "loss": 0.3254,
+      "step": 8083
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.728510738333921e-05,
+      "loss": 0.1256,
+      "step": 8084
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7284771548032846e-05,
+      "loss": 0.3762,
+      "step": 8085
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.728443571272648e-05,
+      "loss": 0.4231,
+      "step": 8086
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7284099877420115e-05,
+      "loss": 0.2777,
+      "step": 8087
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.728376404211375e-05,
+      "loss": 0.4019,
+      "step": 8088
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.728342820680738e-05,
+      "loss": 0.4815,
+      "step": 8089
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.728309237150102e-05,
+      "loss": 0.3053,
+      "step": 8090
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.728275653619465e-05,
+      "loss": 0.2127,
+      "step": 8091
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7282420700888285e-05,
+      "loss": 0.4053,
+      "step": 8092
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.728208486558192e-05,
+      "loss": 0.2798,
+      "step": 8093
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7281749030275554e-05,
+      "loss": 0.38,
+      "step": 8094
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.728141319496919e-05,
+      "loss": 0.3646,
+      "step": 8095
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7281077359662824e-05,
+      "loss": 0.3316,
+      "step": 8096
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.728074152435646e-05,
+      "loss": 0.3059,
+      "step": 8097
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.728040568905009e-05,
+      "loss": 0.3227,
+      "step": 8098
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7280069853743728e-05,
+      "loss": 0.2266,
+      "step": 8099
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.727973401843736e-05,
+      "loss": 0.2725,
+      "step": 8100
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7279398183130994e-05,
+      "loss": 0.5019,
+      "step": 8101
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7279062347824628e-05,
+      "loss": 0.5488,
+      "step": 8102
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7278726512518263e-05,
+      "loss": 0.494,
+      "step": 8103
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7278390677211898e-05,
+      "loss": 0.3647,
+      "step": 8104
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7278054841905532e-05,
+      "loss": 0.3143,
+      "step": 8105
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7277719006599163e-05,
+      "loss": 0.3071,
+      "step": 8106
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7277383171292798e-05,
+      "loss": 0.246,
+      "step": 8107
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7277047335986433e-05,
+      "loss": 0.3417,
+      "step": 8108
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7276711500680067e-05,
+      "loss": 0.3871,
+      "step": 8109
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7276375665373702e-05,
+      "loss": 0.21,
+      "step": 8110
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7276039830067337e-05,
+      "loss": 0.3376,
+      "step": 8111
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.727570399476097e-05,
+      "loss": 0.2958,
+      "step": 8112
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7275368159454603e-05,
+      "loss": 0.4794,
+      "step": 8113
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.727503232414824e-05,
+      "loss": 0.2115,
+      "step": 8114
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7274696488841872e-05,
+      "loss": 0.3566,
+      "step": 8115
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7274360653535507e-05,
+      "loss": 0.1987,
+      "step": 8116
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.727402481822914e-05,
+      "loss": 0.3491,
+      "step": 8117
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7273688982922776e-05,
+      "loss": 0.3316,
+      "step": 8118
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.727335314761641e-05,
+      "loss": 0.3884,
+      "step": 8119
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7273017312310045e-05,
+      "loss": 0.5024,
+      "step": 8120
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.727268147700368e-05,
+      "loss": 0.4335,
+      "step": 8121
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.727234564169731e-05,
+      "loss": 0.3452,
+      "step": 8122
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.727200980639095e-05,
+      "loss": 0.3404,
+      "step": 8123
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.727167397108458e-05,
+      "loss": 0.3844,
+      "step": 8124
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7271338135778215e-05,
+      "loss": 0.2322,
+      "step": 8125
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.727100230047185e-05,
+      "loss": 0.5578,
+      "step": 8126
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7270666465165484e-05,
+      "loss": 0.2803,
+      "step": 8127
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.727033062985912e-05,
+      "loss": 0.2235,
+      "step": 8128
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7269994794552754e-05,
+      "loss": 0.1665,
+      "step": 8129
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7269658959246388e-05,
+      "loss": 0.3119,
+      "step": 8130
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.726932312394002e-05,
+      "loss": 0.3924,
+      "step": 8131
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7268987288633658e-05,
+      "loss": 0.3808,
+      "step": 8132
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.726865145332729e-05,
+      "loss": 0.282,
+      "step": 8133
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7268315618020923e-05,
+      "loss": 0.3319,
+      "step": 8134
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7267979782714558e-05,
+      "loss": 0.4096,
+      "step": 8135
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7267643947408193e-05,
+      "loss": 0.3076,
+      "step": 8136
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7267308112101827e-05,
+      "loss": 0.3978,
+      "step": 8137
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7266972276795462e-05,
+      "loss": 0.329,
+      "step": 8138
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7266636441489097e-05,
+      "loss": 0.3612,
+      "step": 8139
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7266300606182728e-05,
+      "loss": 0.3355,
+      "step": 8140
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7265964770876366e-05,
+      "loss": 0.4939,
+      "step": 8141
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7265628935569997e-05,
+      "loss": 0.3532,
+      "step": 8142
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7265293100263632e-05,
+      "loss": 0.3458,
+      "step": 8143
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7264957264957267e-05,
+      "loss": 0.3367,
+      "step": 8144
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.72646214296509e-05,
+      "loss": 0.5642,
+      "step": 8145
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7264285594344536e-05,
+      "loss": 0.3855,
+      "step": 8146
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.726394975903817e-05,
+      "loss": 0.383,
+      "step": 8147
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7263613923731805e-05,
+      "loss": 0.3294,
+      "step": 8148
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7263278088425436e-05,
+      "loss": 0.211,
+      "step": 8149
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.726294225311907e-05,
+      "loss": 0.2729,
+      "step": 8150
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7262606417812706e-05,
+      "loss": 0.3394,
+      "step": 8151
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.726227058250634e-05,
+      "loss": 0.4256,
+      "step": 8152
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7261934747199975e-05,
+      "loss": 0.3527,
+      "step": 8153
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.726159891189361e-05,
+      "loss": 0.4713,
+      "step": 8154
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.726126307658724e-05,
+      "loss": 0.2287,
+      "step": 8155
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.726092724128088e-05,
+      "loss": 0.3524,
+      "step": 8156
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.726059140597451e-05,
+      "loss": 0.3436,
+      "step": 8157
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7260255570668145e-05,
+      "loss": 0.3585,
+      "step": 8158
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.725991973536178e-05,
+      "loss": 0.3673,
+      "step": 8159
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7259583900055414e-05,
+      "loss": 0.2602,
+      "step": 8160
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.725924806474905e-05,
+      "loss": 0.227,
+      "step": 8161
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7258912229442683e-05,
+      "loss": 0.4113,
+      "step": 8162
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7258576394136318e-05,
+      "loss": 0.4592,
+      "step": 8163
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.725824055882995e-05,
+      "loss": 0.2526,
+      "step": 8164
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7257904723523587e-05,
+      "loss": 0.4454,
+      "step": 8165
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.725756888821722e-05,
+      "loss": 0.2693,
+      "step": 8166
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7257233052910853e-05,
+      "loss": 0.1864,
+      "step": 8167
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7256897217604488e-05,
+      "loss": 0.4857,
+      "step": 8168
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7256561382298123e-05,
+      "loss": 0.2252,
+      "step": 8169
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7256225546991757e-05,
+      "loss": 0.3335,
+      "step": 8170
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7255889711685392e-05,
+      "loss": 0.1857,
+      "step": 8171
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7255553876379026e-05,
+      "loss": 0.3367,
+      "step": 8172
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7255218041072658e-05,
+      "loss": 0.4177,
+      "step": 8173
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7254882205766296e-05,
+      "loss": 0.3117,
+      "step": 8174
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7254546370459927e-05,
+      "loss": 0.3231,
+      "step": 8175
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.725421053515356e-05,
+      "loss": 0.4218,
+      "step": 8176
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7253874699847196e-05,
+      "loss": 0.2746,
+      "step": 8177
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.725353886454083e-05,
+      "loss": 0.2379,
+      "step": 8178
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7253203029234466e-05,
+      "loss": 0.253,
+      "step": 8179
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.72528671939281e-05,
+      "loss": 0.24,
+      "step": 8180
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7252531358621735e-05,
+      "loss": 0.4871,
+      "step": 8181
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7252195523315366e-05,
+      "loss": 0.3041,
+      "step": 8182
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7251859688009004e-05,
+      "loss": 0.4171,
+      "step": 8183
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7251523852702635e-05,
+      "loss": 0.3697,
+      "step": 8184
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.725118801739627e-05,
+      "loss": 0.1797,
+      "step": 8185
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7250852182089905e-05,
+      "loss": 0.4023,
+      "step": 8186
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.725051634678354e-05,
+      "loss": 0.3455,
+      "step": 8187
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7250180511477174e-05,
+      "loss": 0.4426,
+      "step": 8188
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.724984467617081e-05,
+      "loss": 0.3702,
+      "step": 8189
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7249508840864443e-05,
+      "loss": 0.2997,
+      "step": 8190
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7249173005558075e-05,
+      "loss": 0.4756,
+      "step": 8191
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7248837170251713e-05,
+      "loss": 0.2368,
+      "step": 8192
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7248501334945344e-05,
+      "loss": 0.3658,
+      "step": 8193
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.724816549963898e-05,
+      "loss": 0.2499,
+      "step": 8194
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7247829664332613e-05,
+      "loss": 0.3638,
+      "step": 8195
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7247493829026248e-05,
+      "loss": 0.3221,
+      "step": 8196
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.724715799371988e-05,
+      "loss": 0.2767,
+      "step": 8197
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7246822158413517e-05,
+      "loss": 0.4772,
+      "step": 8198
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.724648632310715e-05,
+      "loss": 0.2831,
+      "step": 8199
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7246150487800783e-05,
+      "loss": 0.4183,
+      "step": 8200
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7245814652494418e-05,
+      "loss": 0.4153,
+      "step": 8201
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7245478817188052e-05,
+      "loss": 0.2893,
+      "step": 8202
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7245142981881687e-05,
+      "loss": 0.2822,
+      "step": 8203
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.724480714657532e-05,
+      "loss": 0.296,
+      "step": 8204
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7244471311268956e-05,
+      "loss": 0.347,
+      "step": 8205
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7244135475962588e-05,
+      "loss": 0.4065,
+      "step": 8206
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7243799640656226e-05,
+      "loss": 0.3765,
+      "step": 8207
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7243463805349857e-05,
+      "loss": 0.6282,
+      "step": 8208
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.724312797004349e-05,
+      "loss": 0.5165,
+      "step": 8209
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7242792134737126e-05,
+      "loss": 0.3646,
+      "step": 8210
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.724245629943076e-05,
+      "loss": 0.4292,
+      "step": 8211
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7242120464124395e-05,
+      "loss": 0.3192,
+      "step": 8212
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7241784628818027e-05,
+      "loss": 0.3847,
+      "step": 8213
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7241448793511665e-05,
+      "loss": 0.5678,
+      "step": 8214
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7241112958205296e-05,
+      "loss": 0.394,
+      "step": 8215
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.724077712289893e-05,
+      "loss": 0.2903,
+      "step": 8216
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7240441287592565e-05,
+      "loss": 0.2862,
+      "step": 8217
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.72401054522862e-05,
+      "loss": 0.1822,
+      "step": 8218
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7239769616979835e-05,
+      "loss": 0.4516,
+      "step": 8219
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.723943378167347e-05,
+      "loss": 0.2525,
+      "step": 8220
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7239097946367104e-05,
+      "loss": 0.2641,
+      "step": 8221
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7238762111060735e-05,
+      "loss": 0.3512,
+      "step": 8222
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7238426275754373e-05,
+      "loss": 0.3014,
+      "step": 8223
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7238090440448004e-05,
+      "loss": 0.4453,
+      "step": 8224
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.723775460514164e-05,
+      "loss": 0.4611,
+      "step": 8225
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7237418769835274e-05,
+      "loss": 0.3192,
+      "step": 8226
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.723708293452891e-05,
+      "loss": 0.3367,
+      "step": 8227
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7236747099222543e-05,
+      "loss": 0.4251,
+      "step": 8228
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7236411263916178e-05,
+      "loss": 0.3817,
+      "step": 8229
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7236075428609812e-05,
+      "loss": 0.2952,
+      "step": 8230
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7235739593303444e-05,
+      "loss": 0.4047,
+      "step": 8231
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.723540375799708e-05,
+      "loss": 0.2232,
+      "step": 8232
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7235067922690713e-05,
+      "loss": 0.3302,
+      "step": 8233
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7234732087384348e-05,
+      "loss": 0.3102,
+      "step": 8234
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7234396252077982e-05,
+      "loss": 0.5321,
+      "step": 8235
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7234060416771617e-05,
+      "loss": 0.3341,
+      "step": 8236
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.723372458146525e-05,
+      "loss": 0.3702,
+      "step": 8237
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.7233388746158886e-05,
+      "loss": 0.3168,
+      "step": 8238
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.723305291085252e-05,
+      "loss": 0.4965,
+      "step": 8239
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7232717075546152e-05,
+      "loss": 0.3674,
+      "step": 8240
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7232381240239787e-05,
+      "loss": 0.2251,
+      "step": 8241
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.723204540493342e-05,
+      "loss": 0.5008,
+      "step": 8242
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7231709569627056e-05,
+      "loss": 0.2929,
+      "step": 8243
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.723137373432069e-05,
+      "loss": 0.2766,
+      "step": 8244
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7231037899014325e-05,
+      "loss": 0.4337,
+      "step": 8245
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7230702063707957e-05,
+      "loss": 0.2247,
+      "step": 8246
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7230366228401595e-05,
+      "loss": 0.3287,
+      "step": 8247
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7230030393095226e-05,
+      "loss": 0.2968,
+      "step": 8248
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.722969455778886e-05,
+      "loss": 0.3224,
+      "step": 8249
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7229358722482495e-05,
+      "loss": 0.2894,
+      "step": 8250
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.722902288717613e-05,
+      "loss": 0.3103,
+      "step": 8251
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7228687051869764e-05,
+      "loss": 0.3893,
+      "step": 8252
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.72283512165634e-05,
+      "loss": 0.2453,
+      "step": 8253
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7228015381257034e-05,
+      "loss": 0.343,
+      "step": 8254
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7227679545950665e-05,
+      "loss": 0.4607,
+      "step": 8255
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7227343710644303e-05,
+      "loss": 0.174,
+      "step": 8256
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7227007875337934e-05,
+      "loss": 0.4827,
+      "step": 8257
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.722667204003157e-05,
+      "loss": 0.329,
+      "step": 8258
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7226336204725204e-05,
+      "loss": 0.3181,
+      "step": 8259
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7226000369418838e-05,
+      "loss": 0.5202,
+      "step": 8260
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7225664534112473e-05,
+      "loss": 0.4244,
+      "step": 8261
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7225328698806107e-05,
+      "loss": 0.4169,
+      "step": 8262
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7224992863499742e-05,
+      "loss": 0.282,
+      "step": 8263
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7224657028193373e-05,
+      "loss": 0.2512,
+      "step": 8264
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.722432119288701e-05,
+      "loss": 0.4228,
+      "step": 8265
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7223985357580643e-05,
+      "loss": 0.2689,
+      "step": 8266
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7223649522274277e-05,
+      "loss": 0.3744,
+      "step": 8267
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7223313686967912e-05,
+      "loss": 0.4546,
+      "step": 8268
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7222977851661547e-05,
+      "loss": 0.3686,
+      "step": 8269
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.722264201635518e-05,
+      "loss": 0.3214,
+      "step": 8270
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7222306181048816e-05,
+      "loss": 0.1747,
+      "step": 8271
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.722197034574245e-05,
+      "loss": 0.4196,
+      "step": 8272
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7221634510436082e-05,
+      "loss": 0.5426,
+      "step": 8273
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.722129867512972e-05,
+      "loss": 0.2865,
+      "step": 8274
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.722096283982335e-05,
+      "loss": 0.4036,
+      "step": 8275
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7220627004516986e-05,
+      "loss": 0.2889,
+      "step": 8276
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.722029116921062e-05,
+      "loss": 0.2942,
+      "step": 8277
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7219955333904255e-05,
+      "loss": 0.4708,
+      "step": 8278
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.721961949859789e-05,
+      "loss": 0.4118,
+      "step": 8279
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7219283663291524e-05,
+      "loss": 0.4248,
+      "step": 8280
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.721894782798516e-05,
+      "loss": 0.2496,
+      "step": 8281
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.721861199267879e-05,
+      "loss": 0.2848,
+      "step": 8282
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7218276157372428e-05,
+      "loss": 0.3075,
+      "step": 8283
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.721794032206606e-05,
+      "loss": 0.238,
+      "step": 8284
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7217604486759694e-05,
+      "loss": 0.34,
+      "step": 8285
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.721726865145333e-05,
+      "loss": 0.3295,
+      "step": 8286
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7216932816146963e-05,
+      "loss": 0.3319,
+      "step": 8287
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7216596980840595e-05,
+      "loss": 0.3235,
+      "step": 8288
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7216261145534233e-05,
+      "loss": 0.4639,
+      "step": 8289
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7215925310227864e-05,
+      "loss": 0.4029,
+      "step": 8290
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.72155894749215e-05,
+      "loss": 0.3672,
+      "step": 8291
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7215253639615133e-05,
+      "loss": 0.1563,
+      "step": 8292
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7214917804308768e-05,
+      "loss": 0.4412,
+      "step": 8293
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7214581969002403e-05,
+      "loss": 0.3231,
+      "step": 8294
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7214246133696037e-05,
+      "loss": 0.2815,
+      "step": 8295
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7213910298389672e-05,
+      "loss": 0.3824,
+      "step": 8296
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7213574463083303e-05,
+      "loss": 0.4802,
+      "step": 8297
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.721323862777694e-05,
+      "loss": 0.3923,
+      "step": 8298
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7212902792470572e-05,
+      "loss": 0.2179,
+      "step": 8299
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7212566957164207e-05,
+      "loss": 0.3573,
+      "step": 8300
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7212231121857842e-05,
+      "loss": 0.2903,
+      "step": 8301
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7211895286551476e-05,
+      "loss": 0.4135,
+      "step": 8302
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.721155945124511e-05,
+      "loss": 0.3075,
+      "step": 8303
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7211223615938746e-05,
+      "loss": 0.4396,
+      "step": 8304
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.721088778063238e-05,
+      "loss": 0.4811,
+      "step": 8305
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.721055194532601e-05,
+      "loss": 0.3674,
+      "step": 8306
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.721021611001965e-05,
+      "loss": 0.337,
+      "step": 8307
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.720988027471328e-05,
+      "loss": 0.4039,
+      "step": 8308
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7209544439406916e-05,
+      "loss": 0.3244,
+      "step": 8309
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.720920860410055e-05,
+      "loss": 0.3511,
+      "step": 8310
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7208872768794185e-05,
+      "loss": 0.2785,
+      "step": 8311
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.720853693348782e-05,
+      "loss": 0.2714,
+      "step": 8312
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7208201098181454e-05,
+      "loss": 0.3221,
+      "step": 8313
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.720786526287509e-05,
+      "loss": 0.3753,
+      "step": 8314
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.720752942756872e-05,
+      "loss": 0.2668,
+      "step": 8315
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7207193592262358e-05,
+      "loss": 0.4565,
+      "step": 8316
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.720685775695599e-05,
+      "loss": 0.3289,
+      "step": 8317
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7206521921649624e-05,
+      "loss": 0.2388,
+      "step": 8318
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.720618608634326e-05,
+      "loss": 0.3494,
+      "step": 8319
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7205850251036893e-05,
+      "loss": 0.2787,
+      "step": 8320
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7205514415730528e-05,
+      "loss": 0.3357,
+      "step": 8321
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7205178580424163e-05,
+      "loss": 0.2992,
+      "step": 8322
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7204842745117797e-05,
+      "loss": 0.3455,
+      "step": 8323
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.720450690981143e-05,
+      "loss": 0.3463,
+      "step": 8324
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7204171074505067e-05,
+      "loss": 0.3232,
+      "step": 8325
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7203835239198698e-05,
+      "loss": 0.3315,
+      "step": 8326
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7203499403892332e-05,
+      "loss": 0.338,
+      "step": 8327
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7203163568585967e-05,
+      "loss": 0.2283,
+      "step": 8328
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7202827733279602e-05,
+      "loss": 0.3977,
+      "step": 8329
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7202491897973236e-05,
+      "loss": 0.3588,
+      "step": 8330
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.720215606266687e-05,
+      "loss": 0.3478,
+      "step": 8331
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7201820227360502e-05,
+      "loss": 0.6652,
+      "step": 8332
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7201484392054137e-05,
+      "loss": 0.3708,
+      "step": 8333
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.720114855674777e-05,
+      "loss": 0.4028,
+      "step": 8334
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7200812721441406e-05,
+      "loss": 0.4305,
+      "step": 8335
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.720047688613504e-05,
+      "loss": 0.4529,
+      "step": 8336
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7200141050828676e-05,
+      "loss": 0.2349,
+      "step": 8337
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.719980521552231e-05,
+      "loss": 0.3403,
+      "step": 8338
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.719946938021594e-05,
+      "loss": 0.2437,
+      "step": 8339
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.719913354490958e-05,
+      "loss": 0.3555,
+      "step": 8340
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.719879770960321e-05,
+      "loss": 0.4642,
+      "step": 8341
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7198461874296845e-05,
+      "loss": 0.3124,
+      "step": 8342
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.719812603899048e-05,
+      "loss": 0.2511,
+      "step": 8343
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7197790203684115e-05,
+      "loss": 0.3621,
+      "step": 8344
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.719745436837775e-05,
+      "loss": 0.3493,
+      "step": 8345
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7197118533071384e-05,
+      "loss": 0.3861,
+      "step": 8346
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.719678269776502e-05,
+      "loss": 0.4019,
+      "step": 8347
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.719644686245865e-05,
+      "loss": 0.1895,
+      "step": 8348
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7196111027152288e-05,
+      "loss": 0.2575,
+      "step": 8349
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.719577519184592e-05,
+      "loss": 0.4408,
+      "step": 8350
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7195439356539554e-05,
+      "loss": 0.4683,
+      "step": 8351
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.719510352123319e-05,
+      "loss": 0.3296,
+      "step": 8352
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7194767685926823e-05,
+      "loss": 0.3307,
+      "step": 8353
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7194431850620458e-05,
+      "loss": 0.4003,
+      "step": 8354
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7194096015314092e-05,
+      "loss": 0.3205,
+      "step": 8355
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7193760180007727e-05,
+      "loss": 0.4839,
+      "step": 8356
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.719342434470136e-05,
+      "loss": 0.3187,
+      "step": 8357
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7193088509394996e-05,
+      "loss": 0.3865,
+      "step": 8358
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7192752674088628e-05,
+      "loss": 0.2905,
+      "step": 8359
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7192416838782262e-05,
+      "loss": 0.2895,
+      "step": 8360
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7192081003475897e-05,
+      "loss": 0.4428,
+      "step": 8361
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.719174516816953e-05,
+      "loss": 0.2934,
+      "step": 8362
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7191409332863166e-05,
+      "loss": 0.4415,
+      "step": 8363
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.71910734975568e-05,
+      "loss": 0.5373,
+      "step": 8364
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7190737662250435e-05,
+      "loss": 0.2896,
+      "step": 8365
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7190401826944067e-05,
+      "loss": 0.2838,
+      "step": 8366
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7190065991637705e-05,
+      "loss": 0.4373,
+      "step": 8367
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7189730156331336e-05,
+      "loss": 0.3225,
+      "step": 8368
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.718939432102497e-05,
+      "loss": 0.2955,
+      "step": 8369
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7189058485718605e-05,
+      "loss": 0.2154,
+      "step": 8370
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.718872265041224e-05,
+      "loss": 0.3701,
+      "step": 8371
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7188386815105875e-05,
+      "loss": 0.2012,
+      "step": 8372
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.718805097979951e-05,
+      "loss": 0.2917,
+      "step": 8373
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7187715144493144e-05,
+      "loss": 0.4667,
+      "step": 8374
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7187379309186775e-05,
+      "loss": 0.3586,
+      "step": 8375
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.718704347388041e-05,
+      "loss": 0.3792,
+      "step": 8376
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7186707638574044e-05,
+      "loss": 0.4433,
+      "step": 8377
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.718637180326768e-05,
+      "loss": 0.1623,
+      "step": 8378
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7186035967961314e-05,
+      "loss": 0.5675,
+      "step": 8379
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.718570013265495e-05,
+      "loss": 0.2704,
+      "step": 8380
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.718536429734858e-05,
+      "loss": 0.3373,
+      "step": 8381
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7185028462042218e-05,
+      "loss": 0.4134,
+      "step": 8382
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.718469262673585e-05,
+      "loss": 0.3316,
+      "step": 8383
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7184356791429484e-05,
+      "loss": 0.3928,
+      "step": 8384
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7184020956123118e-05,
+      "loss": 0.3899,
+      "step": 8385
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7183685120816753e-05,
+      "loss": 0.3274,
+      "step": 8386
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7183349285510388e-05,
+      "loss": 0.2224,
+      "step": 8387
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7183013450204022e-05,
+      "loss": 0.42,
+      "step": 8388
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7182677614897657e-05,
+      "loss": 0.3444,
+      "step": 8389
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7182341779591288e-05,
+      "loss": 0.2747,
+      "step": 8390
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7182005944284926e-05,
+      "loss": 0.3499,
+      "step": 8391
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7181670108978557e-05,
+      "loss": 0.503,
+      "step": 8392
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7181334273672192e-05,
+      "loss": 0.2885,
+      "step": 8393
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7180998438365827e-05,
+      "loss": 0.333,
+      "step": 8394
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.718066260305946e-05,
+      "loss": 0.4183,
+      "step": 8395
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7180326767753096e-05,
+      "loss": 0.2703,
+      "step": 8396
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7179990932446727e-05,
+      "loss": 0.3044,
+      "step": 8397
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7179655097140365e-05,
+      "loss": 0.3091,
+      "step": 8398
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7179319261833997e-05,
+      "loss": 0.3095,
+      "step": 8399
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.717898342652763e-05,
+      "loss": 0.4254,
+      "step": 8400
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7178647591221266e-05,
+      "loss": 0.3744,
+      "step": 8401
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.71783117559149e-05,
+      "loss": 0.3285,
+      "step": 8402
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7177975920608535e-05,
+      "loss": 0.2734,
+      "step": 8403
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.717764008530217e-05,
+      "loss": 0.2414,
+      "step": 8404
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7177304249995804e-05,
+      "loss": 0.1894,
+      "step": 8405
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7176968414689436e-05,
+      "loss": 0.304,
+      "step": 8406
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7176632579383074e-05,
+      "loss": 0.3657,
+      "step": 8407
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7176296744076705e-05,
+      "loss": 0.267,
+      "step": 8408
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.717596090877034e-05,
+      "loss": 0.2967,
+      "step": 8409
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7175625073463974e-05,
+      "loss": 0.2351,
+      "step": 8410
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.717528923815761e-05,
+      "loss": 0.2679,
+      "step": 8411
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7174953402851244e-05,
+      "loss": 0.2392,
+      "step": 8412
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7174617567544878e-05,
+      "loss": 0.2003,
+      "step": 8413
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7174281732238513e-05,
+      "loss": 0.3289,
+      "step": 8414
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7173945896932144e-05,
+      "loss": 0.2269,
+      "step": 8415
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7173610061625782e-05,
+      "loss": 0.2865,
+      "step": 8416
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7173274226319413e-05,
+      "loss": 0.171,
+      "step": 8417
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7172938391013048e-05,
+      "loss": 0.2383,
+      "step": 8418
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7172602555706683e-05,
+      "loss": 0.3281,
+      "step": 8419
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7172266720400317e-05,
+      "loss": 0.2506,
+      "step": 8420
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7171930885093952e-05,
+      "loss": 0.2881,
+      "step": 8421
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7171595049787587e-05,
+      "loss": 0.2805,
+      "step": 8422
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7171259214481218e-05,
+      "loss": 0.2791,
+      "step": 8423
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7170923379174853e-05,
+      "loss": 0.228,
+      "step": 8424
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7170587543868487e-05,
+      "loss": 0.4392,
+      "step": 8425
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7170251708562122e-05,
+      "loss": 0.334,
+      "step": 8426
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7169915873255757e-05,
+      "loss": 0.2588,
+      "step": 8427
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.716958003794939e-05,
+      "loss": 0.5031,
+      "step": 8428
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7169244202643026e-05,
+      "loss": 0.3419,
+      "step": 8429
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7168908367336657e-05,
+      "loss": 0.4539,
+      "step": 8430
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7168572532030295e-05,
+      "loss": 0.6697,
+      "step": 8431
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7168236696723926e-05,
+      "loss": 0.588,
+      "step": 8432
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.716790086141756e-05,
+      "loss": 0.441,
+      "step": 8433
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7167565026111196e-05,
+      "loss": 0.2266,
+      "step": 8434
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.716722919080483e-05,
+      "loss": 0.6126,
+      "step": 8435
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.7166893355498465e-05,
+      "loss": 0.5857,
+      "step": 8436
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.71665575201921e-05,
+      "loss": 0.3208,
+      "step": 8437
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7166221684885734e-05,
+      "loss": 0.3497,
+      "step": 8438
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7165885849579366e-05,
+      "loss": 0.3957,
+      "step": 8439
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7165550014273004e-05,
+      "loss": 0.3472,
+      "step": 8440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7165214178966635e-05,
+      "loss": 0.464,
+      "step": 8441
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.716487834366027e-05,
+      "loss": 0.2625,
+      "step": 8442
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7164542508353904e-05,
+      "loss": 0.4459,
+      "step": 8443
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.716420667304754e-05,
+      "loss": 0.3268,
+      "step": 8444
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7163870837741173e-05,
+      "loss": 0.2328,
+      "step": 8445
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7163535002434808e-05,
+      "loss": 0.3858,
+      "step": 8446
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7163199167128443e-05,
+      "loss": 0.2671,
+      "step": 8447
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7162863331822074e-05,
+      "loss": 0.2681,
+      "step": 8448
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7162527496515712e-05,
+      "loss": 0.3122,
+      "step": 8449
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7162191661209343e-05,
+      "loss": 0.2538,
+      "step": 8450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7161855825902978e-05,
+      "loss": 0.3598,
+      "step": 8451
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7161519990596613e-05,
+      "loss": 0.2364,
+      "step": 8452
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7161184155290247e-05,
+      "loss": 0.4902,
+      "step": 8453
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7160848319983882e-05,
+      "loss": 0.4158,
+      "step": 8454
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7160512484677516e-05,
+      "loss": 0.3532,
+      "step": 8455
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.716017664937115e-05,
+      "loss": 0.1914,
+      "step": 8456
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7159840814064782e-05,
+      "loss": 0.3701,
+      "step": 8457
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.715950497875842e-05,
+      "loss": 0.3582,
+      "step": 8458
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.715916914345205e-05,
+      "loss": 0.2678,
+      "step": 8459
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7158833308145686e-05,
+      "loss": 0.4923,
+      "step": 8460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.715849747283932e-05,
+      "loss": 0.2251,
+      "step": 8461
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7158161637532956e-05,
+      "loss": 0.191,
+      "step": 8462
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.715782580222659e-05,
+      "loss": 0.2312,
+      "step": 8463
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7157489966920225e-05,
+      "loss": 0.2419,
+      "step": 8464
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.715715413161386e-05,
+      "loss": 0.1942,
+      "step": 8465
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.715681829630749e-05,
+      "loss": 0.1817,
+      "step": 8466
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7156482461001125e-05,
+      "loss": 0.2904,
+      "step": 8467
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.715614662569476e-05,
+      "loss": 0.3566,
+      "step": 8468
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7155810790388395e-05,
+      "loss": 0.491,
+      "step": 8469
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.715547495508203e-05,
+      "loss": 0.2109,
+      "step": 8470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7155139119775664e-05,
+      "loss": 0.2663,
+      "step": 8471
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7154803284469295e-05,
+      "loss": 0.4738,
+      "step": 8472
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7154467449162933e-05,
+      "loss": 0.3445,
+      "step": 8473
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7154131613856565e-05,
+      "loss": 0.2862,
+      "step": 8474
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.71537957785502e-05,
+      "loss": 0.4604,
+      "step": 8475
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7153459943243834e-05,
+      "loss": 0.2143,
+      "step": 8476
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.715312410793747e-05,
+      "loss": 0.3401,
+      "step": 8477
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7152788272631103e-05,
+      "loss": 0.2928,
+      "step": 8478
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7152452437324738e-05,
+      "loss": 0.2847,
+      "step": 8479
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7152116602018373e-05,
+      "loss": 0.2719,
+      "step": 8480
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7151780766712004e-05,
+      "loss": 0.4339,
+      "step": 8481
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7151444931405642e-05,
+      "loss": 0.2969,
+      "step": 8482
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7151109096099273e-05,
+      "loss": 0.4513,
+      "step": 8483
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7150773260792908e-05,
+      "loss": 0.2694,
+      "step": 8484
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7150437425486542e-05,
+      "loss": 0.2424,
+      "step": 8485
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7150101590180177e-05,
+      "loss": 0.3617,
+      "step": 8486
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.714976575487381e-05,
+      "loss": 0.3956,
+      "step": 8487
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7149429919567446e-05,
+      "loss": 0.349,
+      "step": 8488
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.714909408426108e-05,
+      "loss": 0.291,
+      "step": 8489
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7148758248954712e-05,
+      "loss": 0.3629,
+      "step": 8490
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.714842241364835e-05,
+      "loss": 0.4618,
+      "step": 8491
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.714808657834198e-05,
+      "loss": 0.2537,
+      "step": 8492
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7147750743035616e-05,
+      "loss": 0.4407,
+      "step": 8493
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.714741490772925e-05,
+      "loss": 0.3224,
+      "step": 8494
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7147079072422885e-05,
+      "loss": 0.2775,
+      "step": 8495
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.714674323711652e-05,
+      "loss": 0.3633,
+      "step": 8496
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7146407401810155e-05,
+      "loss": 0.355,
+      "step": 8497
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.714607156650379e-05,
+      "loss": 0.363,
+      "step": 8498
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.714573573119742e-05,
+      "loss": 0.3455,
+      "step": 8499
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.714539989589106e-05,
+      "loss": 0.2559,
+      "step": 8500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.714506406058469e-05,
+      "loss": 0.2551,
+      "step": 8501
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7144728225278325e-05,
+      "loss": 0.1961,
+      "step": 8502
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.714439238997196e-05,
+      "loss": 0.3243,
+      "step": 8503
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7144056554665594e-05,
+      "loss": 0.375,
+      "step": 8504
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.714372071935923e-05,
+      "loss": 0.3543,
+      "step": 8505
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7143384884052863e-05,
+      "loss": 0.2776,
+      "step": 8506
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7143049048746498e-05,
+      "loss": 0.255,
+      "step": 8507
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.714271321344013e-05,
+      "loss": 0.2151,
+      "step": 8508
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7142377378133767e-05,
+      "loss": 0.2419,
+      "step": 8509
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.71420415428274e-05,
+      "loss": 0.4147,
+      "step": 8510
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7141705707521033e-05,
+      "loss": 0.4308,
+      "step": 8511
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7141369872214668e-05,
+      "loss": 0.391,
+      "step": 8512
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7141034036908302e-05,
+      "loss": 0.3563,
+      "step": 8513
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7140698201601934e-05,
+      "loss": 0.4137,
+      "step": 8514
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.714036236629557e-05,
+      "loss": 0.4918,
+      "step": 8515
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7140026530989203e-05,
+      "loss": 0.3603,
+      "step": 8516
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7139690695682838e-05,
+      "loss": 0.3545,
+      "step": 8517
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7139354860376472e-05,
+      "loss": 0.4214,
+      "step": 8518
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7139019025070107e-05,
+      "loss": 0.4058,
+      "step": 8519
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713868318976374e-05,
+      "loss": 0.4813,
+      "step": 8520
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7138347354457376e-05,
+      "loss": 0.45,
+      "step": 8521
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713801151915101e-05,
+      "loss": 0.3954,
+      "step": 8522
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7137675683844642e-05,
+      "loss": 0.2436,
+      "step": 8523
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713733984853828e-05,
+      "loss": 0.3467,
+      "step": 8524
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713700401323191e-05,
+      "loss": 0.4455,
+      "step": 8525
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7136668177925546e-05,
+      "loss": 0.2211,
+      "step": 8526
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713633234261918e-05,
+      "loss": 0.2375,
+      "step": 8527
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7135996507312815e-05,
+      "loss": 0.35,
+      "step": 8528
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713566067200645e-05,
+      "loss": 0.3364,
+      "step": 8529
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7135324836700085e-05,
+      "loss": 0.2677,
+      "step": 8530
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713498900139372e-05,
+      "loss": 0.3271,
+      "step": 8531
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713465316608735e-05,
+      "loss": 0.3058,
+      "step": 8532
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713431733078099e-05,
+      "loss": 0.4026,
+      "step": 8533
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713398149547462e-05,
+      "loss": 0.3325,
+      "step": 8534
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7133645660168254e-05,
+      "loss": 0.3884,
+      "step": 8535
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713330982486189e-05,
+      "loss": 0.3888,
+      "step": 8536
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7132973989555524e-05,
+      "loss": 0.2628,
+      "step": 8537
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713263815424916e-05,
+      "loss": 0.3936,
+      "step": 8538
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7132302318942793e-05,
+      "loss": 0.3232,
+      "step": 8539
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7131966483636428e-05,
+      "loss": 0.2886,
+      "step": 8540
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.713163064833006e-05,
+      "loss": 0.2318,
+      "step": 8541
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7131294813023697e-05,
+      "loss": 0.5603,
+      "step": 8542
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7130958977717328e-05,
+      "loss": 0.4666,
+      "step": 8543
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7130623142410963e-05,
+      "loss": 0.3117,
+      "step": 8544
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7130287307104597e-05,
+      "loss": 0.2853,
+      "step": 8545
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7129951471798232e-05,
+      "loss": 0.489,
+      "step": 8546
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7129615636491867e-05,
+      "loss": 0.2682,
+      "step": 8547
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.71292798011855e-05,
+      "loss": 0.3618,
+      "step": 8548
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7128943965879136e-05,
+      "loss": 0.4093,
+      "step": 8549
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7128608130572767e-05,
+      "loss": 0.3052,
+      "step": 8550
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7128272295266405e-05,
+      "loss": 0.2985,
+      "step": 8551
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7127936459960037e-05,
+      "loss": 0.2582,
+      "step": 8552
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.712760062465367e-05,
+      "loss": 0.5104,
+      "step": 8553
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7127264789347306e-05,
+      "loss": 0.4079,
+      "step": 8554
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.712692895404094e-05,
+      "loss": 0.2436,
+      "step": 8555
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7126593118734575e-05,
+      "loss": 0.2787,
+      "step": 8556
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.712625728342821e-05,
+      "loss": 0.2823,
+      "step": 8557
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.712592144812184e-05,
+      "loss": 0.5326,
+      "step": 8558
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7125585612815476e-05,
+      "loss": 0.3626,
+      "step": 8559
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.712524977750911e-05,
+      "loss": 0.276,
+      "step": 8560
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7124913942202745e-05,
+      "loss": 0.287,
+      "step": 8561
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.712457810689638e-05,
+      "loss": 0.2963,
+      "step": 8562
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7124242271590014e-05,
+      "loss": 0.2872,
+      "step": 8563
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.712390643628365e-05,
+      "loss": 0.2274,
+      "step": 8564
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.712357060097728e-05,
+      "loss": 0.37,
+      "step": 8565
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7123234765670918e-05,
+      "loss": 0.3025,
+      "step": 8566
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.712289893036455e-05,
+      "loss": 0.1492,
+      "step": 8567
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7122563095058184e-05,
+      "loss": 0.3834,
+      "step": 8568
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.712222725975182e-05,
+      "loss": 0.4555,
+      "step": 8569
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7121891424445454e-05,
+      "loss": 0.2088,
+      "step": 8570
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7121555589139088e-05,
+      "loss": 0.2224,
+      "step": 8571
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7121219753832723e-05,
+      "loss": 0.2557,
+      "step": 8572
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7120883918526357e-05,
+      "loss": 0.4789,
+      "step": 8573
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.712054808321999e-05,
+      "loss": 0.224,
+      "step": 8574
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7120212247913623e-05,
+      "loss": 0.496,
+      "step": 8575
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7119876412607258e-05,
+      "loss": 0.191,
+      "step": 8576
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7119540577300893e-05,
+      "loss": 0.4708,
+      "step": 8577
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7119204741994527e-05,
+      "loss": 0.4309,
+      "step": 8578
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7118868906688162e-05,
+      "loss": 0.2158,
+      "step": 8579
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7118533071381797e-05,
+      "loss": 0.3017,
+      "step": 8580
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7118197236075428e-05,
+      "loss": 0.3263,
+      "step": 8581
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7117861400769066e-05,
+      "loss": 0.2503,
+      "step": 8582
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7117525565462697e-05,
+      "loss": 0.4029,
+      "step": 8583
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7117189730156332e-05,
+      "loss": 0.3058,
+      "step": 8584
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7116853894849966e-05,
+      "loss": 0.2529,
+      "step": 8585
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.71165180595436e-05,
+      "loss": 0.3224,
+      "step": 8586
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7116182224237236e-05,
+      "loss": 0.3018,
+      "step": 8587
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.711584638893087e-05,
+      "loss": 0.2556,
+      "step": 8588
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7115510553624505e-05,
+      "loss": 0.4132,
+      "step": 8589
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7115174718318136e-05,
+      "loss": 0.382,
+      "step": 8590
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7114838883011774e-05,
+      "loss": 0.2768,
+      "step": 8591
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7114503047705406e-05,
+      "loss": 0.3488,
+      "step": 8592
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.711416721239904e-05,
+      "loss": 0.2913,
+      "step": 8593
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7113831377092675e-05,
+      "loss": 0.1605,
+      "step": 8594
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.711349554178631e-05,
+      "loss": 0.2611,
+      "step": 8595
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7113159706479944e-05,
+      "loss": 0.2467,
+      "step": 8596
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.711282387117358e-05,
+      "loss": 0.406,
+      "step": 8597
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7112488035867213e-05,
+      "loss": 0.3206,
+      "step": 8598
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7112152200560845e-05,
+      "loss": 0.3895,
+      "step": 8599
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7111816365254483e-05,
+      "loss": 0.4171,
+      "step": 8600
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7111480529948114e-05,
+      "loss": 0.3491,
+      "step": 8601
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.711114469464175e-05,
+      "loss": 0.1909,
+      "step": 8602
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7110808859335383e-05,
+      "loss": 0.2571,
+      "step": 8603
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7110473024029018e-05,
+      "loss": 0.2754,
+      "step": 8604
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.711013718872265e-05,
+      "loss": 0.3326,
+      "step": 8605
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7109801353416287e-05,
+      "loss": 0.2943,
+      "step": 8606
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.710946551810992e-05,
+      "loss": 0.3616,
+      "step": 8607
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7109129682803553e-05,
+      "loss": 0.4137,
+      "step": 8608
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7108793847497188e-05,
+      "loss": 0.2918,
+      "step": 8609
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7108458012190822e-05,
+      "loss": 0.4364,
+      "step": 8610
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7108122176884457e-05,
+      "loss": 0.3389,
+      "step": 8611
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7107786341578092e-05,
+      "loss": 0.2169,
+      "step": 8612
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7107450506271726e-05,
+      "loss": 0.4746,
+      "step": 8613
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7107114670965358e-05,
+      "loss": 0.5544,
+      "step": 8614
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7106778835658996e-05,
+      "loss": 0.2492,
+      "step": 8615
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7106443000352627e-05,
+      "loss": 0.2868,
+      "step": 8616
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.710610716504626e-05,
+      "loss": 0.268,
+      "step": 8617
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7105771329739896e-05,
+      "loss": 0.2009,
+      "step": 8618
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.710543549443353e-05,
+      "loss": 0.2847,
+      "step": 8619
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7105099659127166e-05,
+      "loss": 0.371,
+      "step": 8620
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.71047638238208e-05,
+      "loss": 0.222,
+      "step": 8621
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7104427988514435e-05,
+      "loss": 0.4179,
+      "step": 8622
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7104092153208066e-05,
+      "loss": 0.3412,
+      "step": 8623
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7103756317901704e-05,
+      "loss": 0.4345,
+      "step": 8624
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7103420482595335e-05,
+      "loss": 0.3724,
+      "step": 8625
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.710308464728897e-05,
+      "loss": 0.2975,
+      "step": 8626
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7102748811982605e-05,
+      "loss": 0.2899,
+      "step": 8627
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.710241297667624e-05,
+      "loss": 0.3276,
+      "step": 8628
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7102077141369874e-05,
+      "loss": 0.4003,
+      "step": 8629
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.710174130606351e-05,
+      "loss": 0.4473,
+      "step": 8630
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7101405470757143e-05,
+      "loss": 0.2089,
+      "step": 8631
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7101069635450775e-05,
+      "loss": 0.2834,
+      "step": 8632
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7100733800144413e-05,
+      "loss": 0.4095,
+      "step": 8633
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.7100397964838044e-05,
+      "loss": 0.2882,
+      "step": 8634
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.710006212953168e-05,
+      "loss": 0.4244,
+      "step": 8635
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7099726294225313e-05,
+      "loss": 0.407,
+      "step": 8636
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7099390458918948e-05,
+      "loss": 0.2859,
+      "step": 8637
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7099054623612582e-05,
+      "loss": 0.223,
+      "step": 8638
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7098718788306217e-05,
+      "loss": 0.3392,
+      "step": 8639
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7098382952999852e-05,
+      "loss": 0.3256,
+      "step": 8640
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7098047117693483e-05,
+      "loss": 0.2221,
+      "step": 8641
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.709771128238712e-05,
+      "loss": 0.2982,
+      "step": 8642
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7097375447080752e-05,
+      "loss": 0.4556,
+      "step": 8643
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7097039611774387e-05,
+      "loss": 0.2558,
+      "step": 8644
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.709670377646802e-05,
+      "loss": 0.3288,
+      "step": 8645
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7096367941161656e-05,
+      "loss": 0.4008,
+      "step": 8646
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7096032105855287e-05,
+      "loss": 0.3971,
+      "step": 8647
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7095696270548926e-05,
+      "loss": 0.2785,
+      "step": 8648
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7095360435242557e-05,
+      "loss": 0.3384,
+      "step": 8649
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.709502459993619e-05,
+      "loss": 0.3367,
+      "step": 8650
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7094688764629826e-05,
+      "loss": 0.3438,
+      "step": 8651
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.709435292932346e-05,
+      "loss": 0.3123,
+      "step": 8652
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7094017094017095e-05,
+      "loss": 0.3401,
+      "step": 8653
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.709368125871073e-05,
+      "loss": 0.4095,
+      "step": 8654
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7093345423404365e-05,
+      "loss": 0.4327,
+      "step": 8655
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7093009588097996e-05,
+      "loss": 0.2953,
+      "step": 8656
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7092673752791634e-05,
+      "loss": 0.2535,
+      "step": 8657
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7092337917485265e-05,
+      "loss": 0.283,
+      "step": 8658
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.70920020821789e-05,
+      "loss": 0.4124,
+      "step": 8659
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7091666246872535e-05,
+      "loss": 0.3954,
+      "step": 8660
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.709133041156617e-05,
+      "loss": 0.3154,
+      "step": 8661
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7090994576259804e-05,
+      "loss": 0.2834,
+      "step": 8662
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.709065874095344e-05,
+      "loss": 0.4188,
+      "step": 8663
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7090322905647073e-05,
+      "loss": 0.444,
+      "step": 8664
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7089987070340704e-05,
+      "loss": 0.5405,
+      "step": 8665
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7089651235034342e-05,
+      "loss": 0.3599,
+      "step": 8666
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7089315399727974e-05,
+      "loss": 0.4439,
+      "step": 8667
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7088979564421608e-05,
+      "loss": 0.4097,
+      "step": 8668
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7088643729115243e-05,
+      "loss": 0.1662,
+      "step": 8669
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7088307893808878e-05,
+      "loss": 0.2434,
+      "step": 8670
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7087972058502512e-05,
+      "loss": 0.2837,
+      "step": 8671
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7087636223196147e-05,
+      "loss": 0.2637,
+      "step": 8672
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.708730038788978e-05,
+      "loss": 0.2759,
+      "step": 8673
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7086964552583413e-05,
+      "loss": 0.34,
+      "step": 8674
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.708662871727705e-05,
+      "loss": 0.2917,
+      "step": 8675
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7086292881970682e-05,
+      "loss": 0.3274,
+      "step": 8676
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7085957046664317e-05,
+      "loss": 0.2385,
+      "step": 8677
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.708562121135795e-05,
+      "loss": 0.3329,
+      "step": 8678
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7085285376051586e-05,
+      "loss": 0.1783,
+      "step": 8679
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.708494954074522e-05,
+      "loss": 0.2986,
+      "step": 8680
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7084613705438855e-05,
+      "loss": 0.254,
+      "step": 8681
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.708427787013249e-05,
+      "loss": 0.2189,
+      "step": 8682
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.708394203482612e-05,
+      "loss": 0.2541,
+      "step": 8683
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.708360619951976e-05,
+      "loss": 0.5753,
+      "step": 8684
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.708327036421339e-05,
+      "loss": 0.3738,
+      "step": 8685
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7082934528907025e-05,
+      "loss": 0.4969,
+      "step": 8686
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.708259869360066e-05,
+      "loss": 0.4952,
+      "step": 8687
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7082262858294294e-05,
+      "loss": 0.1991,
+      "step": 8688
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.708192702298793e-05,
+      "loss": 0.3727,
+      "step": 8689
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7081591187681564e-05,
+      "loss": 0.3826,
+      "step": 8690
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.70812553523752e-05,
+      "loss": 0.353,
+      "step": 8691
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.708091951706883e-05,
+      "loss": 0.3226,
+      "step": 8692
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7080583681762464e-05,
+      "loss": 0.3652,
+      "step": 8693
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.70802478464561e-05,
+      "loss": 0.3457,
+      "step": 8694
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7079912011149734e-05,
+      "loss": 0.4652,
+      "step": 8695
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7079576175843368e-05,
+      "loss": 0.3729,
+      "step": 8696
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7079240340537003e-05,
+      "loss": 0.2656,
+      "step": 8697
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7078904505230634e-05,
+      "loss": 0.3498,
+      "step": 8698
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7078568669924272e-05,
+      "loss": 0.4195,
+      "step": 8699
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7078232834617903e-05,
+      "loss": 0.3406,
+      "step": 8700
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7077896999311538e-05,
+      "loss": 0.3516,
+      "step": 8701
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7077561164005173e-05,
+      "loss": 0.2461,
+      "step": 8702
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7077225328698807e-05,
+      "loss": 0.4288,
+      "step": 8703
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7076889493392442e-05,
+      "loss": 0.3314,
+      "step": 8704
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7076553658086077e-05,
+      "loss": 0.404,
+      "step": 8705
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.707621782277971e-05,
+      "loss": 0.365,
+      "step": 8706
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7075881987473343e-05,
+      "loss": 0.3909,
+      "step": 8707
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.707554615216698e-05,
+      "loss": 0.2943,
+      "step": 8708
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7075210316860612e-05,
+      "loss": 0.3707,
+      "step": 8709
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7074874481554247e-05,
+      "loss": 0.3615,
+      "step": 8710
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.707453864624788e-05,
+      "loss": 0.2463,
+      "step": 8711
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7074202810941516e-05,
+      "loss": 0.5239,
+      "step": 8712
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.707386697563515e-05,
+      "loss": 0.5151,
+      "step": 8713
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7073531140328785e-05,
+      "loss": 0.322,
+      "step": 8714
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.707319530502242e-05,
+      "loss": 0.3445,
+      "step": 8715
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.707285946971605e-05,
+      "loss": 0.1934,
+      "step": 8716
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.707252363440969e-05,
+      "loss": 0.1654,
+      "step": 8717
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.707218779910332e-05,
+      "loss": 0.4023,
+      "step": 8718
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7071851963796955e-05,
+      "loss": 0.2935,
+      "step": 8719
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.707151612849059e-05,
+      "loss": 0.2526,
+      "step": 8720
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7071180293184224e-05,
+      "loss": 0.232,
+      "step": 8721
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.707084445787786e-05,
+      "loss": 0.312,
+      "step": 8722
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7070508622571494e-05,
+      "loss": 0.265,
+      "step": 8723
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7070172787265128e-05,
+      "loss": 0.2024,
+      "step": 8724
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.706983695195876e-05,
+      "loss": 0.4848,
+      "step": 8725
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7069501116652398e-05,
+      "loss": 0.2616,
+      "step": 8726
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.706916528134603e-05,
+      "loss": 0.2709,
+      "step": 8727
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7068829446039663e-05,
+      "loss": 0.442,
+      "step": 8728
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7068493610733298e-05,
+      "loss": 0.4994,
+      "step": 8729
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7068157775426933e-05,
+      "loss": 0.4553,
+      "step": 8730
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7067821940120567e-05,
+      "loss": 0.3108,
+      "step": 8731
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7067486104814202e-05,
+      "loss": 0.4441,
+      "step": 8732
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7067150269507837e-05,
+      "loss": 0.401,
+      "step": 8733
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7066814434201468e-05,
+      "loss": 0.4292,
+      "step": 8734
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7066478598895106e-05,
+      "loss": 0.3304,
+      "step": 8735
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7066142763588737e-05,
+      "loss": 0.4382,
+      "step": 8736
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7065806928282372e-05,
+      "loss": 0.3761,
+      "step": 8737
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7065471092976007e-05,
+      "loss": 0.3506,
+      "step": 8738
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.706513525766964e-05,
+      "loss": 0.3386,
+      "step": 8739
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7064799422363272e-05,
+      "loss": 0.3221,
+      "step": 8740
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.706446358705691e-05,
+      "loss": 0.2565,
+      "step": 8741
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7064127751750542e-05,
+      "loss": 0.3444,
+      "step": 8742
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7063791916444176e-05,
+      "loss": 0.364,
+      "step": 8743
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.706345608113781e-05,
+      "loss": 0.388,
+      "step": 8744
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7063120245831446e-05,
+      "loss": 0.3783,
+      "step": 8745
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.706278441052508e-05,
+      "loss": 0.2617,
+      "step": 8746
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7062448575218715e-05,
+      "loss": 0.3191,
+      "step": 8747
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.706211273991235e-05,
+      "loss": 0.3256,
+      "step": 8748
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.706177690460598e-05,
+      "loss": 0.3257,
+      "step": 8749
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.706144106929962e-05,
+      "loss": 0.2942,
+      "step": 8750
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.706110523399325e-05,
+      "loss": 0.3038,
+      "step": 8751
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7060769398686885e-05,
+      "loss": 0.4151,
+      "step": 8752
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.706043356338052e-05,
+      "loss": 0.3742,
+      "step": 8753
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7060097728074154e-05,
+      "loss": 0.2896,
+      "step": 8754
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.705976189276779e-05,
+      "loss": 0.3103,
+      "step": 8755
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.705942605746142e-05,
+      "loss": 0.2605,
+      "step": 8756
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7059090222155058e-05,
+      "loss": 0.4445,
+      "step": 8757
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.705875438684869e-05,
+      "loss": 0.4136,
+      "step": 8758
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7058418551542324e-05,
+      "loss": 0.179,
+      "step": 8759
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.705808271623596e-05,
+      "loss": 0.337,
+      "step": 8760
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7057746880929593e-05,
+      "loss": 0.347,
+      "step": 8761
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7057411045623228e-05,
+      "loss": 0.4206,
+      "step": 8762
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7057075210316863e-05,
+      "loss": 0.2449,
+      "step": 8763
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7056739375010497e-05,
+      "loss": 0.2405,
+      "step": 8764
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.705640353970413e-05,
+      "loss": 0.5787,
+      "step": 8765
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7056067704397766e-05,
+      "loss": 0.2935,
+      "step": 8766
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7055731869091398e-05,
+      "loss": 0.4106,
+      "step": 8767
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7055396033785032e-05,
+      "loss": 0.3198,
+      "step": 8768
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7055060198478667e-05,
+      "loss": 0.3414,
+      "step": 8769
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.70547243631723e-05,
+      "loss": 0.3646,
+      "step": 8770
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7054388527865936e-05,
+      "loss": 0.2488,
+      "step": 8771
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.705405269255957e-05,
+      "loss": 0.3212,
+      "step": 8772
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7053716857253206e-05,
+      "loss": 0.3336,
+      "step": 8773
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7053381021946837e-05,
+      "loss": 0.4634,
+      "step": 8774
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7053045186640475e-05,
+      "loss": 0.4237,
+      "step": 8775
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7052709351334106e-05,
+      "loss": 0.2646,
+      "step": 8776
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.705237351602774e-05,
+      "loss": 0.2232,
+      "step": 8777
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7052037680721375e-05,
+      "loss": 0.3685,
+      "step": 8778
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.705170184541501e-05,
+      "loss": 0.3761,
+      "step": 8779
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7051366010108645e-05,
+      "loss": 0.3371,
+      "step": 8780
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.705103017480228e-05,
+      "loss": 0.3219,
+      "step": 8781
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.705069433949591e-05,
+      "loss": 0.3576,
+      "step": 8782
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7050358504189545e-05,
+      "loss": 0.2996,
+      "step": 8783
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.705002266888318e-05,
+      "loss": 0.4529,
+      "step": 8784
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7049686833576815e-05,
+      "loss": 0.4636,
+      "step": 8785
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.704935099827045e-05,
+      "loss": 0.2604,
+      "step": 8786
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7049015162964084e-05,
+      "loss": 0.3065,
+      "step": 8787
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.704867932765772e-05,
+      "loss": 0.2367,
+      "step": 8788
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.704834349235135e-05,
+      "loss": 0.2244,
+      "step": 8789
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7048007657044988e-05,
+      "loss": 0.1903,
+      "step": 8790
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.704767182173862e-05,
+      "loss": 0.2745,
+      "step": 8791
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7047335986432254e-05,
+      "loss": 0.4822,
+      "step": 8792
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.704700015112589e-05,
+      "loss": 0.2958,
+      "step": 8793
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7046664315819523e-05,
+      "loss": 0.366,
+      "step": 8794
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7046328480513158e-05,
+      "loss": 0.3089,
+      "step": 8795
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7045992645206792e-05,
+      "loss": 0.2717,
+      "step": 8796
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7045656809900427e-05,
+      "loss": 0.3898,
+      "step": 8797
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7045320974594058e-05,
+      "loss": 0.4294,
+      "step": 8798
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7044985139287696e-05,
+      "loss": 0.3062,
+      "step": 8799
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7044649303981328e-05,
+      "loss": 0.3153,
+      "step": 8800
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7044313468674962e-05,
+      "loss": 0.4513,
+      "step": 8801
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7043977633368597e-05,
+      "loss": 0.3939,
+      "step": 8802
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.704364179806223e-05,
+      "loss": 0.2294,
+      "step": 8803
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7043305962755866e-05,
+      "loss": 0.3182,
+      "step": 8804
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.70429701274495e-05,
+      "loss": 0.3353,
+      "step": 8805
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7042634292143135e-05,
+      "loss": 0.1714,
+      "step": 8806
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7042298456836767e-05,
+      "loss": 0.4142,
+      "step": 8807
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7041962621530405e-05,
+      "loss": 0.2678,
+      "step": 8808
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7041626786224036e-05,
+      "loss": 0.1541,
+      "step": 8809
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.704129095091767e-05,
+      "loss": 0.3536,
+      "step": 8810
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7040955115611305e-05,
+      "loss": 0.4577,
+      "step": 8811
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.704061928030494e-05,
+      "loss": 0.2714,
+      "step": 8812
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7040283444998575e-05,
+      "loss": 0.2115,
+      "step": 8813
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.703994760969221e-05,
+      "loss": 0.3169,
+      "step": 8814
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7039611774385844e-05,
+      "loss": 0.4132,
+      "step": 8815
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7039275939079475e-05,
+      "loss": 0.336,
+      "step": 8816
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7038940103773113e-05,
+      "loss": 0.1808,
+      "step": 8817
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7038604268466744e-05,
+      "loss": 0.1538,
+      "step": 8818
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.703826843316038e-05,
+      "loss": 0.3377,
+      "step": 8819
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7037932597854014e-05,
+      "loss": 0.4088,
+      "step": 8820
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.703759676254765e-05,
+      "loss": 0.313,
+      "step": 8821
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7037260927241283e-05,
+      "loss": 0.2995,
+      "step": 8822
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7036925091934918e-05,
+      "loss": 0.2424,
+      "step": 8823
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7036589256628552e-05,
+      "loss": 0.4258,
+      "step": 8824
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7036253421322184e-05,
+      "loss": 0.3274,
+      "step": 8825
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.703591758601582e-05,
+      "loss": 0.2286,
+      "step": 8826
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7035581750709453e-05,
+      "loss": 0.2991,
+      "step": 8827
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7035245915403087e-05,
+      "loss": 0.2671,
+      "step": 8828
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7034910080096722e-05,
+      "loss": 0.3837,
+      "step": 8829
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7034574244790357e-05,
+      "loss": 0.2187,
+      "step": 8830
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7034238409483988e-05,
+      "loss": 0.4068,
+      "step": 8831
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7033902574177626e-05,
+      "loss": 0.4968,
+      "step": 8832
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.7033566738871257e-05,
+      "loss": 0.2159,
+      "step": 8833
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7033230903564892e-05,
+      "loss": 0.3246,
+      "step": 8834
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7032895068258527e-05,
+      "loss": 0.3365,
+      "step": 8835
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.703255923295216e-05,
+      "loss": 0.314,
+      "step": 8836
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7032223397645796e-05,
+      "loss": 0.3972,
+      "step": 8837
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.703188756233943e-05,
+      "loss": 0.3379,
+      "step": 8838
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7031551727033065e-05,
+      "loss": 0.3228,
+      "step": 8839
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7031215891726696e-05,
+      "loss": 0.3658,
+      "step": 8840
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7030880056420335e-05,
+      "loss": 0.3874,
+      "step": 8841
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7030544221113966e-05,
+      "loss": 0.2891,
+      "step": 8842
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.70302083858076e-05,
+      "loss": 0.3761,
+      "step": 8843
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7029872550501235e-05,
+      "loss": 0.3495,
+      "step": 8844
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.702953671519487e-05,
+      "loss": 0.5107,
+      "step": 8845
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7029200879888504e-05,
+      "loss": 0.316,
+      "step": 8846
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.702886504458214e-05,
+      "loss": 0.4211,
+      "step": 8847
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7028529209275774e-05,
+      "loss": 0.32,
+      "step": 8848
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7028193373969405e-05,
+      "loss": 0.2919,
+      "step": 8849
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7027857538663043e-05,
+      "loss": 0.3174,
+      "step": 8850
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7027521703356674e-05,
+      "loss": 0.2777,
+      "step": 8851
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.702718586805031e-05,
+      "loss": 0.3264,
+      "step": 8852
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7026850032743944e-05,
+      "loss": 0.3131,
+      "step": 8853
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7026514197437578e-05,
+      "loss": 0.3334,
+      "step": 8854
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7026178362131213e-05,
+      "loss": 0.2629,
+      "step": 8855
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7025842526824847e-05,
+      "loss": 0.2368,
+      "step": 8856
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7025506691518482e-05,
+      "loss": 0.1723,
+      "step": 8857
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7025170856212113e-05,
+      "loss": 0.3738,
+      "step": 8858
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.702483502090575e-05,
+      "loss": 0.2173,
+      "step": 8859
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7024499185599383e-05,
+      "loss": 0.3951,
+      "step": 8860
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7024163350293017e-05,
+      "loss": 0.2301,
+      "step": 8861
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7023827514986652e-05,
+      "loss": 0.2816,
+      "step": 8862
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7023491679680287e-05,
+      "loss": 0.3128,
+      "step": 8863
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.702315584437392e-05,
+      "loss": 0.3656,
+      "step": 8864
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7022820009067556e-05,
+      "loss": 0.4493,
+      "step": 8865
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.702248417376119e-05,
+      "loss": 0.2648,
+      "step": 8866
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7022148338454822e-05,
+      "loss": 0.4954,
+      "step": 8867
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.702181250314846e-05,
+      "loss": 0.2652,
+      "step": 8868
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.702147666784209e-05,
+      "loss": 0.2483,
+      "step": 8869
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7021140832535726e-05,
+      "loss": 0.4169,
+      "step": 8870
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.702080499722936e-05,
+      "loss": 0.4391,
+      "step": 8871
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7020469161922995e-05,
+      "loss": 0.4193,
+      "step": 8872
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7020133326616626e-05,
+      "loss": 0.493,
+      "step": 8873
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7019797491310264e-05,
+      "loss": 0.3633,
+      "step": 8874
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7019461656003896e-05,
+      "loss": 0.387,
+      "step": 8875
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.701912582069753e-05,
+      "loss": 0.4327,
+      "step": 8876
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7018789985391165e-05,
+      "loss": 0.382,
+      "step": 8877
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.70184541500848e-05,
+      "loss": 0.3205,
+      "step": 8878
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7018118314778434e-05,
+      "loss": 0.2569,
+      "step": 8879
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.701778247947207e-05,
+      "loss": 0.3141,
+      "step": 8880
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7017446644165703e-05,
+      "loss": 0.3583,
+      "step": 8881
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7017110808859335e-05,
+      "loss": 0.291,
+      "step": 8882
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7016774973552973e-05,
+      "loss": 0.2324,
+      "step": 8883
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7016439138246604e-05,
+      "loss": 0.3329,
+      "step": 8884
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.701610330294024e-05,
+      "loss": 0.2844,
+      "step": 8885
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7015767467633873e-05,
+      "loss": 0.2819,
+      "step": 8886
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7015431632327508e-05,
+      "loss": 0.4339,
+      "step": 8887
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7015095797021143e-05,
+      "loss": 0.4359,
+      "step": 8888
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7014759961714777e-05,
+      "loss": 0.3475,
+      "step": 8889
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7014424126408412e-05,
+      "loss": 0.6225,
+      "step": 8890
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7014088291102043e-05,
+      "loss": 0.2344,
+      "step": 8891
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.701375245579568e-05,
+      "loss": 0.5113,
+      "step": 8892
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7013416620489312e-05,
+      "loss": 0.2355,
+      "step": 8893
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7013080785182947e-05,
+      "loss": 0.2612,
+      "step": 8894
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7012744949876582e-05,
+      "loss": 0.3762,
+      "step": 8895
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7012409114570216e-05,
+      "loss": 0.3651,
+      "step": 8896
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.701207327926385e-05,
+      "loss": 0.2184,
+      "step": 8897
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7011737443957486e-05,
+      "loss": 0.3041,
+      "step": 8898
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.701140160865112e-05,
+      "loss": 0.3634,
+      "step": 8899
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.701106577334475e-05,
+      "loss": 0.2494,
+      "step": 8900
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.701072993803839e-05,
+      "loss": 0.3281,
+      "step": 8901
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.701039410273202e-05,
+      "loss": 0.3031,
+      "step": 8902
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7010058267425656e-05,
+      "loss": 0.5747,
+      "step": 8903
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.700972243211929e-05,
+      "loss": 0.4609,
+      "step": 8904
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7009386596812925e-05,
+      "loss": 0.4227,
+      "step": 8905
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.700905076150656e-05,
+      "loss": 0.3263,
+      "step": 8906
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7008714926200194e-05,
+      "loss": 0.5458,
+      "step": 8907
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.700837909089383e-05,
+      "loss": 0.278,
+      "step": 8908
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.700804325558746e-05,
+      "loss": 0.3689,
+      "step": 8909
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7007707420281098e-05,
+      "loss": 0.3463,
+      "step": 8910
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.700737158497473e-05,
+      "loss": 0.4772,
+      "step": 8911
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7007035749668364e-05,
+      "loss": 0.3133,
+      "step": 8912
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7006699914362e-05,
+      "loss": 0.4039,
+      "step": 8913
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7006364079055633e-05,
+      "loss": 0.244,
+      "step": 8914
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7006028243749268e-05,
+      "loss": 0.3769,
+      "step": 8915
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7005692408442903e-05,
+      "loss": 0.2385,
+      "step": 8916
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7005356573136534e-05,
+      "loss": 0.2738,
+      "step": 8917
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.700502073783017e-05,
+      "loss": 0.4176,
+      "step": 8918
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7004684902523803e-05,
+      "loss": 0.207,
+      "step": 8919
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7004349067217438e-05,
+      "loss": 0.2854,
+      "step": 8920
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7004013231911072e-05,
+      "loss": 0.318,
+      "step": 8921
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7003677396604707e-05,
+      "loss": 0.3306,
+      "step": 8922
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7003341561298342e-05,
+      "loss": 0.4661,
+      "step": 8923
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7003005725991973e-05,
+      "loss": 0.5737,
+      "step": 8924
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.700266989068561e-05,
+      "loss": 0.2274,
+      "step": 8925
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7002334055379242e-05,
+      "loss": 0.2754,
+      "step": 8926
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7001998220072877e-05,
+      "loss": 0.4318,
+      "step": 8927
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.700166238476651e-05,
+      "loss": 0.2278,
+      "step": 8928
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7001326549460146e-05,
+      "loss": 0.1767,
+      "step": 8929
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.700099071415378e-05,
+      "loss": 0.2228,
+      "step": 8930
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.7000654878847416e-05,
+      "loss": 0.3934,
+      "step": 8931
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.700031904354105e-05,
+      "loss": 0.3981,
+      "step": 8932
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.699998320823468e-05,
+      "loss": 0.2317,
+      "step": 8933
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6999647372928316e-05,
+      "loss": 0.3954,
+      "step": 8934
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.699931153762195e-05,
+      "loss": 0.2923,
+      "step": 8935
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6998975702315585e-05,
+      "loss": 0.2936,
+      "step": 8936
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.699863986700922e-05,
+      "loss": 0.3005,
+      "step": 8937
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6998304031702855e-05,
+      "loss": 0.2478,
+      "step": 8938
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.699796819639649e-05,
+      "loss": 0.4829,
+      "step": 8939
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.699763236109012e-05,
+      "loss": 0.2524,
+      "step": 8940
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.699729652578376e-05,
+      "loss": 0.2055,
+      "step": 8941
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.699696069047739e-05,
+      "loss": 0.3274,
+      "step": 8942
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6996624855171025e-05,
+      "loss": 0.2268,
+      "step": 8943
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.699628901986466e-05,
+      "loss": 0.299,
+      "step": 8944
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6995953184558294e-05,
+      "loss": 0.299,
+      "step": 8945
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.699561734925193e-05,
+      "loss": 0.466,
+      "step": 8946
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6995281513945563e-05,
+      "loss": 0.279,
+      "step": 8947
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6994945678639198e-05,
+      "loss": 0.3464,
+      "step": 8948
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.699460984333283e-05,
+      "loss": 0.3575,
+      "step": 8949
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6994274008026467e-05,
+      "loss": 0.3464,
+      "step": 8950
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6993938172720098e-05,
+      "loss": 0.3388,
+      "step": 8951
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6993602337413733e-05,
+      "loss": 0.2473,
+      "step": 8952
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6993266502107368e-05,
+      "loss": 0.2748,
+      "step": 8953
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6992930666801002e-05,
+      "loss": 0.3817,
+      "step": 8954
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6992594831494637e-05,
+      "loss": 0.2608,
+      "step": 8955
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.699225899618827e-05,
+      "loss": 0.2556,
+      "step": 8956
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6991923160881906e-05,
+      "loss": 0.2454,
+      "step": 8957
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6991587325575537e-05,
+      "loss": 0.2592,
+      "step": 8958
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6991251490269175e-05,
+      "loss": 0.3798,
+      "step": 8959
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6990915654962807e-05,
+      "loss": 0.3422,
+      "step": 8960
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.699057981965644e-05,
+      "loss": 0.2603,
+      "step": 8961
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6990243984350076e-05,
+      "loss": 0.3107,
+      "step": 8962
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.698990814904371e-05,
+      "loss": 0.2653,
+      "step": 8963
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6989572313737342e-05,
+      "loss": 0.2941,
+      "step": 8964
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.698923647843098e-05,
+      "loss": 0.3161,
+      "step": 8965
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.698890064312461e-05,
+      "loss": 0.263,
+      "step": 8966
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6988564807818246e-05,
+      "loss": 0.3774,
+      "step": 8967
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.698822897251188e-05,
+      "loss": 0.4445,
+      "step": 8968
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6987893137205515e-05,
+      "loss": 0.213,
+      "step": 8969
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.698755730189915e-05,
+      "loss": 0.4475,
+      "step": 8970
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6987221466592784e-05,
+      "loss": 0.3604,
+      "step": 8971
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.698688563128642e-05,
+      "loss": 0.4056,
+      "step": 8972
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.698654979598005e-05,
+      "loss": 0.3752,
+      "step": 8973
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.698621396067369e-05,
+      "loss": 0.2749,
+      "step": 8974
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.698587812536732e-05,
+      "loss": 0.2828,
+      "step": 8975
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6985542290060954e-05,
+      "loss": 0.2818,
+      "step": 8976
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.698520645475459e-05,
+      "loss": 0.2346,
+      "step": 8977
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6984870619448224e-05,
+      "loss": 0.3624,
+      "step": 8978
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6984534784141858e-05,
+      "loss": 0.2721,
+      "step": 8979
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6984198948835493e-05,
+      "loss": 0.2928,
+      "step": 8980
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6983863113529128e-05,
+      "loss": 0.2824,
+      "step": 8981
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.698352727822276e-05,
+      "loss": 0.2297,
+      "step": 8982
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6983191442916397e-05,
+      "loss": 0.3453,
+      "step": 8983
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6982855607610028e-05,
+      "loss": 0.3646,
+      "step": 8984
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6982519772303663e-05,
+      "loss": 0.3441,
+      "step": 8985
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6982183936997297e-05,
+      "loss": 0.45,
+      "step": 8986
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6981848101690932e-05,
+      "loss": 0.3515,
+      "step": 8987
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6981512266384567e-05,
+      "loss": 0.2343,
+      "step": 8988
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.69811764310782e-05,
+      "loss": 0.4812,
+      "step": 8989
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6980840595771836e-05,
+      "loss": 0.2721,
+      "step": 8990
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6980504760465467e-05,
+      "loss": 0.3547,
+      "step": 8991
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6980168925159105e-05,
+      "loss": 0.247,
+      "step": 8992
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6979833089852737e-05,
+      "loss": 0.2417,
+      "step": 8993
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.697949725454637e-05,
+      "loss": 0.3986,
+      "step": 8994
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6979161419240006e-05,
+      "loss": 0.2807,
+      "step": 8995
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.697882558393364e-05,
+      "loss": 0.126,
+      "step": 8996
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6978489748627275e-05,
+      "loss": 0.2478,
+      "step": 8997
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.697815391332091e-05,
+      "loss": 0.2677,
+      "step": 8998
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6977818078014544e-05,
+      "loss": 0.3784,
+      "step": 8999
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6977482242708176e-05,
+      "loss": 0.2649,
+      "step": 9000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6977146407401814e-05,
+      "loss": 0.3468,
+      "step": 9001
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6976810572095445e-05,
+      "loss": 0.3015,
+      "step": 9002
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.697647473678908e-05,
+      "loss": 0.2055,
+      "step": 9003
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6976138901482714e-05,
+      "loss": 0.3647,
+      "step": 9004
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.697580306617635e-05,
+      "loss": 0.203,
+      "step": 9005
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6975467230869984e-05,
+      "loss": 0.3333,
+      "step": 9006
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6975131395563618e-05,
+      "loss": 0.2562,
+      "step": 9007
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.697479556025725e-05,
+      "loss": 0.2447,
+      "step": 9008
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6974459724950884e-05,
+      "loss": 0.287,
+      "step": 9009
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.697412388964452e-05,
+      "loss": 0.3356,
+      "step": 9010
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6973788054338153e-05,
+      "loss": 0.2619,
+      "step": 9011
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6973452219031788e-05,
+      "loss": 0.358,
+      "step": 9012
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6973116383725423e-05,
+      "loss": 0.2437,
+      "step": 9013
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6972780548419057e-05,
+      "loss": 0.2799,
+      "step": 9014
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.697244471311269e-05,
+      "loss": 0.6111,
+      "step": 9015
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6972108877806327e-05,
+      "loss": 0.5095,
+      "step": 9016
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6971773042499958e-05,
+      "loss": 0.3265,
+      "step": 9017
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6971437207193593e-05,
+      "loss": 0.2373,
+      "step": 9018
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6971101371887227e-05,
+      "loss": 0.2901,
+      "step": 9019
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6970765536580862e-05,
+      "loss": 0.3881,
+      "step": 9020
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6970429701274497e-05,
+      "loss": 0.2847,
+      "step": 9021
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.697009386596813e-05,
+      "loss": 0.3431,
+      "step": 9022
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6969758030661766e-05,
+      "loss": 0.3383,
+      "step": 9023
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6969422195355397e-05,
+      "loss": 0.3382,
+      "step": 9024
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6969086360049035e-05,
+      "loss": 0.4028,
+      "step": 9025
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6968750524742666e-05,
+      "loss": 0.2423,
+      "step": 9026
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.69684146894363e-05,
+      "loss": 0.4618,
+      "step": 9027
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6968078854129936e-05,
+      "loss": 0.357,
+      "step": 9028
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.696774301882357e-05,
+      "loss": 0.3212,
+      "step": 9029
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6967407183517205e-05,
+      "loss": 0.2947,
+      "step": 9030
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.696707134821084e-05,
+      "loss": 0.4456,
+      "step": 9031
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6966735512904474e-05,
+      "loss": 0.5266,
+      "step": 9032
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6966399677598106e-05,
+      "loss": 0.2084,
+      "step": 9033
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6966063842291744e-05,
+      "loss": 0.2195,
+      "step": 9034
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6965728006985375e-05,
+      "loss": 0.3807,
+      "step": 9035
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.696539217167901e-05,
+      "loss": 0.3793,
+      "step": 9036
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6965056336372644e-05,
+      "loss": 0.3754,
+      "step": 9037
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.696472050106628e-05,
+      "loss": 0.4629,
+      "step": 9038
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6964384665759913e-05,
+      "loss": 0.4108,
+      "step": 9039
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6964048830453548e-05,
+      "loss": 0.4447,
+      "step": 9040
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6963712995147183e-05,
+      "loss": 0.2248,
+      "step": 9041
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6963377159840814e-05,
+      "loss": 0.304,
+      "step": 9042
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6963041324534452e-05,
+      "loss": 0.3081,
+      "step": 9043
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6962705489228083e-05,
+      "loss": 0.331,
+      "step": 9044
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6962369653921718e-05,
+      "loss": 0.3903,
+      "step": 9045
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6962033818615353e-05,
+      "loss": 0.3571,
+      "step": 9046
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6961697983308987e-05,
+      "loss": 0.2953,
+      "step": 9047
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6961362148002622e-05,
+      "loss": 0.552,
+      "step": 9048
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6961026312696256e-05,
+      "loss": 0.2003,
+      "step": 9049
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.696069047738989e-05,
+      "loss": 0.2972,
+      "step": 9050
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6960354642083522e-05,
+      "loss": 0.3653,
+      "step": 9051
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.696001880677716e-05,
+      "loss": 0.297,
+      "step": 9052
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.695968297147079e-05,
+      "loss": 0.4176,
+      "step": 9053
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6959347136164426e-05,
+      "loss": 0.3725,
+      "step": 9054
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.695901130085806e-05,
+      "loss": 0.3401,
+      "step": 9055
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6958675465551696e-05,
+      "loss": 0.3467,
+      "step": 9056
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6958339630245327e-05,
+      "loss": 0.2416,
+      "step": 9057
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6958003794938965e-05,
+      "loss": 0.4962,
+      "step": 9058
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6957667959632596e-05,
+      "loss": 0.4061,
+      "step": 9059
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.695733212432623e-05,
+      "loss": 0.3617,
+      "step": 9060
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6956996289019865e-05,
+      "loss": 0.3757,
+      "step": 9061
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.69566604537135e-05,
+      "loss": 0.2695,
+      "step": 9062
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6956324618407135e-05,
+      "loss": 0.2943,
+      "step": 9063
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.695598878310077e-05,
+      "loss": 0.2346,
+      "step": 9064
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6955652947794404e-05,
+      "loss": 0.3822,
+      "step": 9065
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6955317112488035e-05,
+      "loss": 0.5116,
+      "step": 9066
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6954981277181673e-05,
+      "loss": 0.4748,
+      "step": 9067
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6954645441875305e-05,
+      "loss": 0.2598,
+      "step": 9068
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.695430960656894e-05,
+      "loss": 0.3298,
+      "step": 9069
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6953973771262574e-05,
+      "loss": 0.2409,
+      "step": 9070
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.695363793595621e-05,
+      "loss": 0.3194,
+      "step": 9071
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6953302100649843e-05,
+      "loss": 0.3045,
+      "step": 9072
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6952966265343478e-05,
+      "loss": 0.4242,
+      "step": 9073
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6952630430037113e-05,
+      "loss": 0.4038,
+      "step": 9074
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6952294594730744e-05,
+      "loss": 0.2859,
+      "step": 9075
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6951958759424382e-05,
+      "loss": 0.204,
+      "step": 9076
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6951622924118013e-05,
+      "loss": 0.3631,
+      "step": 9077
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6951287088811648e-05,
+      "loss": 0.1598,
+      "step": 9078
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6950951253505282e-05,
+      "loss": 0.4177,
+      "step": 9079
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6950615418198917e-05,
+      "loss": 0.3475,
+      "step": 9080
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.695027958289255e-05,
+      "loss": 0.2902,
+      "step": 9081
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6949943747586186e-05,
+      "loss": 0.2898,
+      "step": 9082
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.694960791227982e-05,
+      "loss": 0.4056,
+      "step": 9083
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6949272076973452e-05,
+      "loss": 0.1828,
+      "step": 9084
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.694893624166709e-05,
+      "loss": 0.2026,
+      "step": 9085
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.694860040636072e-05,
+      "loss": 0.5006,
+      "step": 9086
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6948264571054356e-05,
+      "loss": 0.359,
+      "step": 9087
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.694792873574799e-05,
+      "loss": 0.431,
+      "step": 9088
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6947592900441625e-05,
+      "loss": 0.3259,
+      "step": 9089
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.694725706513526e-05,
+      "loss": 0.2202,
+      "step": 9090
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6946921229828895e-05,
+      "loss": 0.2537,
+      "step": 9091
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.694658539452253e-05,
+      "loss": 0.4675,
+      "step": 9092
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.694624955921616e-05,
+      "loss": 0.3585,
+      "step": 9093
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.69459137239098e-05,
+      "loss": 0.4557,
+      "step": 9094
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.694557788860343e-05,
+      "loss": 0.2879,
+      "step": 9095
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6945242053297065e-05,
+      "loss": 0.5401,
+      "step": 9096
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.69449062179907e-05,
+      "loss": 0.2018,
+      "step": 9097
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6944570382684334e-05,
+      "loss": 0.373,
+      "step": 9098
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6944234547377965e-05,
+      "loss": 0.4232,
+      "step": 9099
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6943898712071603e-05,
+      "loss": 0.4552,
+      "step": 9100
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6943562876765234e-05,
+      "loss": 0.3444,
+      "step": 9101
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.694322704145887e-05,
+      "loss": 0.4901,
+      "step": 9102
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6942891206152504e-05,
+      "loss": 0.4639,
+      "step": 9103
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.694255537084614e-05,
+      "loss": 0.3611,
+      "step": 9104
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6942219535539773e-05,
+      "loss": 0.2566,
+      "step": 9105
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6941883700233408e-05,
+      "loss": 0.3182,
+      "step": 9106
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6941547864927042e-05,
+      "loss": 0.2219,
+      "step": 9107
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6941212029620674e-05,
+      "loss": 0.1602,
+      "step": 9108
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.694087619431431e-05,
+      "loss": 0.4589,
+      "step": 9109
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6940540359007943e-05,
+      "loss": 0.2602,
+      "step": 9110
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6940204523701578e-05,
+      "loss": 0.3659,
+      "step": 9111
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6939868688395212e-05,
+      "loss": 0.4879,
+      "step": 9112
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6939532853088847e-05,
+      "loss": 0.5022,
+      "step": 9113
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.693919701778248e-05,
+      "loss": 0.3597,
+      "step": 9114
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6938861182476113e-05,
+      "loss": 0.4337,
+      "step": 9115
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.693852534716975e-05,
+      "loss": 0.4495,
+      "step": 9116
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6938189511863382e-05,
+      "loss": 0.4003,
+      "step": 9117
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6937853676557017e-05,
+      "loss": 0.3328,
+      "step": 9118
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.693751784125065e-05,
+      "loss": 0.2978,
+      "step": 9119
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6937182005944286e-05,
+      "loss": 0.3294,
+      "step": 9120
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.693684617063792e-05,
+      "loss": 0.3077,
+      "step": 9121
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6936510335331555e-05,
+      "loss": 0.3778,
+      "step": 9122
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.693617450002519e-05,
+      "loss": 0.4466,
+      "step": 9123
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.693583866471882e-05,
+      "loss": 0.1631,
+      "step": 9124
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.693550282941246e-05,
+      "loss": 0.4134,
+      "step": 9125
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.693516699410609e-05,
+      "loss": 0.1859,
+      "step": 9126
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6934831158799725e-05,
+      "loss": 0.3254,
+      "step": 9127
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.693449532349336e-05,
+      "loss": 0.2246,
+      "step": 9128
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6934159488186994e-05,
+      "loss": 0.4663,
+      "step": 9129
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.693382365288063e-05,
+      "loss": 0.3015,
+      "step": 9130
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6933487817574264e-05,
+      "loss": 0.2828,
+      "step": 9131
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.69331519822679e-05,
+      "loss": 0.2531,
+      "step": 9132
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.693281614696153e-05,
+      "loss": 0.22,
+      "step": 9133
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6932480311655168e-05,
+      "loss": 0.3914,
+      "step": 9134
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.69321444763488e-05,
+      "loss": 0.2406,
+      "step": 9135
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6931808641042434e-05,
+      "loss": 0.2049,
+      "step": 9136
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6931472805736068e-05,
+      "loss": 0.3786,
+      "step": 9137
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6931136970429703e-05,
+      "loss": 0.2872,
+      "step": 9138
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6930801135123337e-05,
+      "loss": 0.3862,
+      "step": 9139
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6930465299816972e-05,
+      "loss": 0.2747,
+      "step": 9140
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6930129464510607e-05,
+      "loss": 0.2844,
+      "step": 9141
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6929793629204238e-05,
+      "loss": 0.4166,
+      "step": 9142
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6929457793897873e-05,
+      "loss": 0.3548,
+      "step": 9143
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6929121958591507e-05,
+      "loss": 0.2876,
+      "step": 9144
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6928786123285142e-05,
+      "loss": 0.2587,
+      "step": 9145
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6928450287978777e-05,
+      "loss": 0.3881,
+      "step": 9146
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.692811445267241e-05,
+      "loss": 0.3108,
+      "step": 9147
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6927778617366043e-05,
+      "loss": 0.5098,
+      "step": 9148
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.692744278205968e-05,
+      "loss": 0.3036,
+      "step": 9149
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6927106946753312e-05,
+      "loss": 0.3069,
+      "step": 9150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6926771111446946e-05,
+      "loss": 0.2435,
+      "step": 9151
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.692643527614058e-05,
+      "loss": 0.2068,
+      "step": 9152
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6926099440834216e-05,
+      "loss": 0.3382,
+      "step": 9153
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.692576360552785e-05,
+      "loss": 0.2304,
+      "step": 9154
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6925427770221485e-05,
+      "loss": 0.4399,
+      "step": 9155
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.692509193491512e-05,
+      "loss": 0.3445,
+      "step": 9156
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.692475609960875e-05,
+      "loss": 0.3187,
+      "step": 9157
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.692442026430239e-05,
+      "loss": 0.1407,
+      "step": 9158
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.692408442899602e-05,
+      "loss": 0.3467,
+      "step": 9159
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6923748593689655e-05,
+      "loss": 0.3592,
+      "step": 9160
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.692341275838329e-05,
+      "loss": 0.3999,
+      "step": 9161
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6923076923076924e-05,
+      "loss": 0.5301,
+      "step": 9162
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.692274108777056e-05,
+      "loss": 0.2713,
+      "step": 9163
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6922405252464193e-05,
+      "loss": 0.49,
+      "step": 9164
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6922069417157828e-05,
+      "loss": 0.3277,
+      "step": 9165
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.692173358185146e-05,
+      "loss": 0.4266,
+      "step": 9166
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6921397746545097e-05,
+      "loss": 0.4468,
+      "step": 9167
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.692106191123873e-05,
+      "loss": 0.21,
+      "step": 9168
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6920726075932363e-05,
+      "loss": 0.2806,
+      "step": 9169
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6920390240625998e-05,
+      "loss": 0.2801,
+      "step": 9170
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6920054405319633e-05,
+      "loss": 0.6341,
+      "step": 9171
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6919718570013267e-05,
+      "loss": 0.4275,
+      "step": 9172
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6919382734706902e-05,
+      "loss": 0.3149,
+      "step": 9173
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6919046899400537e-05,
+      "loss": 0.3678,
+      "step": 9174
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6918711064094168e-05,
+      "loss": 0.3077,
+      "step": 9175
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6918375228787806e-05,
+      "loss": 0.1895,
+      "step": 9176
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6918039393481437e-05,
+      "loss": 0.4311,
+      "step": 9177
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6917703558175072e-05,
+      "loss": 0.4837,
+      "step": 9178
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6917367722868706e-05,
+      "loss": 0.1783,
+      "step": 9179
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.691703188756234e-05,
+      "loss": 0.3552,
+      "step": 9180
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6916696052255976e-05,
+      "loss": 0.4466,
+      "step": 9181
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.691636021694961e-05,
+      "loss": 0.2521,
+      "step": 9182
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6916024381643245e-05,
+      "loss": 0.2246,
+      "step": 9183
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6915688546336876e-05,
+      "loss": 0.3434,
+      "step": 9184
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6915352711030514e-05,
+      "loss": 0.198,
+      "step": 9185
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6915016875724146e-05,
+      "loss": 0.2449,
+      "step": 9186
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.691468104041778e-05,
+      "loss": 0.4361,
+      "step": 9187
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6914345205111415e-05,
+      "loss": 0.4615,
+      "step": 9188
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.691400936980505e-05,
+      "loss": 0.2301,
+      "step": 9189
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.691367353449868e-05,
+      "loss": 0.2962,
+      "step": 9190
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.691333769919232e-05,
+      "loss": 0.2403,
+      "step": 9191
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.691300186388595e-05,
+      "loss": 0.4324,
+      "step": 9192
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6912666028579585e-05,
+      "loss": 0.2678,
+      "step": 9193
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.691233019327322e-05,
+      "loss": 0.2214,
+      "step": 9194
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6911994357966854e-05,
+      "loss": 0.3516,
+      "step": 9195
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.691165852266049e-05,
+      "loss": 0.41,
+      "step": 9196
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6911322687354123e-05,
+      "loss": 0.5062,
+      "step": 9197
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6910986852047758e-05,
+      "loss": 0.1653,
+      "step": 9198
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.691065101674139e-05,
+      "loss": 0.3621,
+      "step": 9199
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6910315181435027e-05,
+      "loss": 0.2334,
+      "step": 9200
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.690997934612866e-05,
+      "loss": 0.2951,
+      "step": 9201
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6909643510822293e-05,
+      "loss": 0.2752,
+      "step": 9202
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6909307675515928e-05,
+      "loss": 0.3895,
+      "step": 9203
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6908971840209562e-05,
+      "loss": 0.3985,
+      "step": 9204
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6908636004903197e-05,
+      "loss": 0.4159,
+      "step": 9205
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6908300169596832e-05,
+      "loss": 0.4046,
+      "step": 9206
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6907964334290466e-05,
+      "loss": 0.2996,
+      "step": 9207
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6907628498984098e-05,
+      "loss": 0.2132,
+      "step": 9208
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6907292663677736e-05,
+      "loss": 0.4442,
+      "step": 9209
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6906956828371367e-05,
+      "loss": 0.2691,
+      "step": 9210
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6906620993065e-05,
+      "loss": 0.4621,
+      "step": 9211
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6906285157758636e-05,
+      "loss": 0.325,
+      "step": 9212
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.690594932245227e-05,
+      "loss": 0.3409,
+      "step": 9213
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6905613487145906e-05,
+      "loss": 0.4359,
+      "step": 9214
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.690527765183954e-05,
+      "loss": 0.4395,
+      "step": 9215
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6904941816533175e-05,
+      "loss": 0.2964,
+      "step": 9216
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6904605981226806e-05,
+      "loss": 0.4046,
+      "step": 9217
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6904270145920444e-05,
+      "loss": 0.2861,
+      "step": 9218
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6903934310614075e-05,
+      "loss": 0.3521,
+      "step": 9219
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.690359847530771e-05,
+      "loss": 0.3233,
+      "step": 9220
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6903262640001345e-05,
+      "loss": 0.29,
+      "step": 9221
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.690292680469498e-05,
+      "loss": 0.4433,
+      "step": 9222
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6902590969388614e-05,
+      "loss": 0.3323,
+      "step": 9223
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.690225513408225e-05,
+      "loss": 0.2853,
+      "step": 9224
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6901919298775883e-05,
+      "loss": 0.264,
+      "step": 9225
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6901583463469515e-05,
+      "loss": 0.3708,
+      "step": 9226
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6901247628163153e-05,
+      "loss": 0.4061,
+      "step": 9227
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6900911792856784e-05,
+      "loss": 0.3423,
+      "step": 9228
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.690057595755042e-05,
+      "loss": 0.2444,
+      "step": 9229
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.6900240122244053e-05,
+      "loss": 0.447,
+      "step": 9230
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6899904286937688e-05,
+      "loss": 0.4258,
+      "step": 9231
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6899568451631322e-05,
+      "loss": 0.3911,
+      "step": 9232
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6899232616324957e-05,
+      "loss": 0.264,
+      "step": 9233
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.689889678101859e-05,
+      "loss": 0.2801,
+      "step": 9234
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6898560945712223e-05,
+      "loss": 0.3076,
+      "step": 9235
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6898225110405858e-05,
+      "loss": 0.4808,
+      "step": 9236
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6897889275099492e-05,
+      "loss": 0.3795,
+      "step": 9237
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6897553439793127e-05,
+      "loss": 0.3437,
+      "step": 9238
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.689721760448676e-05,
+      "loss": 0.3344,
+      "step": 9239
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6896881769180396e-05,
+      "loss": 0.3672,
+      "step": 9240
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6896545933874027e-05,
+      "loss": 0.5303,
+      "step": 9241
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6896210098567665e-05,
+      "loss": 0.4062,
+      "step": 9242
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6895874263261297e-05,
+      "loss": 0.3119,
+      "step": 9243
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.689553842795493e-05,
+      "loss": 0.436,
+      "step": 9244
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6895202592648566e-05,
+      "loss": 0.2794,
+      "step": 9245
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.68948667573422e-05,
+      "loss": 0.3168,
+      "step": 9246
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6894530922035835e-05,
+      "loss": 0.3393,
+      "step": 9247
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.689419508672947e-05,
+      "loss": 0.3921,
+      "step": 9248
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6893859251423105e-05,
+      "loss": 0.4097,
+      "step": 9249
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6893523416116736e-05,
+      "loss": 0.3257,
+      "step": 9250
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6893187580810374e-05,
+      "loss": 0.4214,
+      "step": 9251
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6892851745504005e-05,
+      "loss": 0.2364,
+      "step": 9252
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.689251591019764e-05,
+      "loss": 0.4239,
+      "step": 9253
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6892180074891274e-05,
+      "loss": 0.4902,
+      "step": 9254
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.689184423958491e-05,
+      "loss": 0.2827,
+      "step": 9255
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6891508404278544e-05,
+      "loss": 0.2909,
+      "step": 9256
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.689117256897218e-05,
+      "loss": 0.2598,
+      "step": 9257
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6890836733665813e-05,
+      "loss": 0.3713,
+      "step": 9258
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6890500898359444e-05,
+      "loss": 0.1834,
+      "step": 9259
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6890165063053082e-05,
+      "loss": 0.3,
+      "step": 9260
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6889829227746714e-05,
+      "loss": 0.2985,
+      "step": 9261
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6889493392440348e-05,
+      "loss": 0.3212,
+      "step": 9262
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6889157557133983e-05,
+      "loss": 0.4072,
+      "step": 9263
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6888821721827618e-05,
+      "loss": 0.2156,
+      "step": 9264
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6888485886521252e-05,
+      "loss": 0.3255,
+      "step": 9265
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6888150051214887e-05,
+      "loss": 0.3045,
+      "step": 9266
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.688781421590852e-05,
+      "loss": 0.341,
+      "step": 9267
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6887478380602153e-05,
+      "loss": 0.3328,
+      "step": 9268
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.688714254529579e-05,
+      "loss": 0.4449,
+      "step": 9269
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6886806709989422e-05,
+      "loss": 0.2771,
+      "step": 9270
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6886470874683057e-05,
+      "loss": 0.1816,
+      "step": 9271
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.688613503937669e-05,
+      "loss": 0.381,
+      "step": 9272
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6885799204070326e-05,
+      "loss": 0.4443,
+      "step": 9273
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.688546336876396e-05,
+      "loss": 0.2964,
+      "step": 9274
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6885127533457595e-05,
+      "loss": 0.238,
+      "step": 9275
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.688479169815123e-05,
+      "loss": 0.2283,
+      "step": 9276
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.688445586284486e-05,
+      "loss": 0.3179,
+      "step": 9277
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6884120027538496e-05,
+      "loss": 0.3537,
+      "step": 9278
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.688378419223213e-05,
+      "loss": 0.4388,
+      "step": 9279
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6883448356925765e-05,
+      "loss": 0.4981,
+      "step": 9280
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.68831125216194e-05,
+      "loss": 0.4145,
+      "step": 9281
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6882776686313034e-05,
+      "loss": 0.2521,
+      "step": 9282
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6882440851006666e-05,
+      "loss": 0.2661,
+      "step": 9283
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6882105015700304e-05,
+      "loss": 0.287,
+      "step": 9284
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6881769180393935e-05,
+      "loss": 0.3838,
+      "step": 9285
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.688143334508757e-05,
+      "loss": 0.3995,
+      "step": 9286
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6881097509781204e-05,
+      "loss": 0.2727,
+      "step": 9287
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.688076167447484e-05,
+      "loss": 0.3605,
+      "step": 9288
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6880425839168474e-05,
+      "loss": 0.3276,
+      "step": 9289
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6880090003862108e-05,
+      "loss": 0.3581,
+      "step": 9290
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6879754168555743e-05,
+      "loss": 0.288,
+      "step": 9291
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6879418333249374e-05,
+      "loss": 0.5008,
+      "step": 9292
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6879082497943012e-05,
+      "loss": 0.314,
+      "step": 9293
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6878746662636643e-05,
+      "loss": 0.4605,
+      "step": 9294
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6878410827330278e-05,
+      "loss": 0.4181,
+      "step": 9295
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6878074992023913e-05,
+      "loss": 0.2419,
+      "step": 9296
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6877739156717547e-05,
+      "loss": 0.3598,
+      "step": 9297
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6877403321411182e-05,
+      "loss": 0.2288,
+      "step": 9298
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6877067486104813e-05,
+      "loss": 0.2526,
+      "step": 9299
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.687673165079845e-05,
+      "loss": 0.3596,
+      "step": 9300
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6876395815492083e-05,
+      "loss": 0.5117,
+      "step": 9301
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6876059980185717e-05,
+      "loss": 0.2837,
+      "step": 9302
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6875724144879352e-05,
+      "loss": 0.3266,
+      "step": 9303
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6875388309572987e-05,
+      "loss": 0.4274,
+      "step": 9304
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.687505247426662e-05,
+      "loss": 0.3431,
+      "step": 9305
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6874716638960256e-05,
+      "loss": 0.282,
+      "step": 9306
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.687438080365389e-05,
+      "loss": 0.2432,
+      "step": 9307
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6874044968347522e-05,
+      "loss": 0.2475,
+      "step": 9308
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.687370913304116e-05,
+      "loss": 0.4508,
+      "step": 9309
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.687337329773479e-05,
+      "loss": 0.3604,
+      "step": 9310
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6873037462428426e-05,
+      "loss": 0.3369,
+      "step": 9311
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.687270162712206e-05,
+      "loss": 0.365,
+      "step": 9312
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6872365791815695e-05,
+      "loss": 0.4541,
+      "step": 9313
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.687202995650933e-05,
+      "loss": 0.2216,
+      "step": 9314
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6871694121202964e-05,
+      "loss": 0.2894,
+      "step": 9315
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.68713582858966e-05,
+      "loss": 0.3014,
+      "step": 9316
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.687102245059023e-05,
+      "loss": 0.315,
+      "step": 9317
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6870686615283868e-05,
+      "loss": 0.2514,
+      "step": 9318
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.68703507799775e-05,
+      "loss": 0.1995,
+      "step": 9319
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6870014944671134e-05,
+      "loss": 0.3748,
+      "step": 9320
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.686967910936477e-05,
+      "loss": 0.3187,
+      "step": 9321
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6869343274058403e-05,
+      "loss": 0.2016,
+      "step": 9322
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6869007438752038e-05,
+      "loss": 0.2381,
+      "step": 9323
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6868671603445673e-05,
+      "loss": 0.2731,
+      "step": 9324
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6868335768139304e-05,
+      "loss": 0.3205,
+      "step": 9325
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.686799993283294e-05,
+      "loss": 0.2122,
+      "step": 9326
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6867664097526573e-05,
+      "loss": 0.2658,
+      "step": 9327
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6867328262220208e-05,
+      "loss": 0.2383,
+      "step": 9328
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6866992426913843e-05,
+      "loss": 0.3788,
+      "step": 9329
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6866656591607477e-05,
+      "loss": 0.2707,
+      "step": 9330
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6866320756301112e-05,
+      "loss": 0.3186,
+      "step": 9331
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6865984920994743e-05,
+      "loss": 0.3429,
+      "step": 9332
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.686564908568838e-05,
+      "loss": 0.1489,
+      "step": 9333
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6865313250382012e-05,
+      "loss": 0.3104,
+      "step": 9334
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6864977415075647e-05,
+      "loss": 0.1746,
+      "step": 9335
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6864641579769282e-05,
+      "loss": 0.2929,
+      "step": 9336
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6864305744462916e-05,
+      "loss": 0.3136,
+      "step": 9337
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.686396990915655e-05,
+      "loss": 0.4238,
+      "step": 9338
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6863634073850186e-05,
+      "loss": 0.3217,
+      "step": 9339
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.686329823854382e-05,
+      "loss": 0.1771,
+      "step": 9340
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.686296240323745e-05,
+      "loss": 0.3157,
+      "step": 9341
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.686262656793109e-05,
+      "loss": 0.3594,
+      "step": 9342
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.686229073262472e-05,
+      "loss": 0.436,
+      "step": 9343
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6861954897318355e-05,
+      "loss": 0.3097,
+      "step": 9344
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.686161906201199e-05,
+      "loss": 0.1887,
+      "step": 9345
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6861283226705625e-05,
+      "loss": 0.4237,
+      "step": 9346
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.686094739139926e-05,
+      "loss": 0.3847,
+      "step": 9347
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6860611556092894e-05,
+      "loss": 0.1363,
+      "step": 9348
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.686027572078653e-05,
+      "loss": 0.2989,
+      "step": 9349
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.685993988548016e-05,
+      "loss": 0.2055,
+      "step": 9350
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6859604050173798e-05,
+      "loss": 0.1816,
+      "step": 9351
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.685926821486743e-05,
+      "loss": 0.2969,
+      "step": 9352
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6858932379561064e-05,
+      "loss": 0.2706,
+      "step": 9353
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.68585965442547e-05,
+      "loss": 0.2998,
+      "step": 9354
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6858260708948333e-05,
+      "loss": 0.3199,
+      "step": 9355
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6857924873641968e-05,
+      "loss": 0.3986,
+      "step": 9356
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6857589038335603e-05,
+      "loss": 0.2709,
+      "step": 9357
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6857253203029237e-05,
+      "loss": 0.602,
+      "step": 9358
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.685691736772287e-05,
+      "loss": 0.5094,
+      "step": 9359
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6856581532416506e-05,
+      "loss": 0.3553,
+      "step": 9360
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6856245697110138e-05,
+      "loss": 0.3797,
+      "step": 9361
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6855909861803772e-05,
+      "loss": 0.2157,
+      "step": 9362
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6855574026497407e-05,
+      "loss": 0.1754,
+      "step": 9363
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.685523819119104e-05,
+      "loss": 0.2827,
+      "step": 9364
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6854902355884676e-05,
+      "loss": 0.2908,
+      "step": 9365
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.685456652057831e-05,
+      "loss": 0.3526,
+      "step": 9366
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6854230685271946e-05,
+      "loss": 0.3455,
+      "step": 9367
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6853894849965577e-05,
+      "loss": 0.35,
+      "step": 9368
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.685355901465921e-05,
+      "loss": 0.4498,
+      "step": 9369
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6853223179352846e-05,
+      "loss": 0.3098,
+      "step": 9370
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.685288734404648e-05,
+      "loss": 0.4362,
+      "step": 9371
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6852551508740115e-05,
+      "loss": 0.4948,
+      "step": 9372
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.685221567343375e-05,
+      "loss": 0.3933,
+      "step": 9373
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.685187983812738e-05,
+      "loss": 0.3546,
+      "step": 9374
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.685154400282102e-05,
+      "loss": 0.4339,
+      "step": 9375
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.685120816751465e-05,
+      "loss": 0.361,
+      "step": 9376
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6850872332208285e-05,
+      "loss": 0.3517,
+      "step": 9377
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.685053649690192e-05,
+      "loss": 0.2109,
+      "step": 9378
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6850200661595555e-05,
+      "loss": 0.1976,
+      "step": 9379
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.684986482628919e-05,
+      "loss": 0.2365,
+      "step": 9380
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6849528990982824e-05,
+      "loss": 0.2918,
+      "step": 9381
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.684919315567646e-05,
+      "loss": 0.2857,
+      "step": 9382
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.684885732037009e-05,
+      "loss": 0.4472,
+      "step": 9383
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6848521485063728e-05,
+      "loss": 0.3169,
+      "step": 9384
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.684818564975736e-05,
+      "loss": 0.2982,
+      "step": 9385
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6847849814450994e-05,
+      "loss": 0.3739,
+      "step": 9386
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.684751397914463e-05,
+      "loss": 0.2546,
+      "step": 9387
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6847178143838263e-05,
+      "loss": 0.3105,
+      "step": 9388
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6846842308531898e-05,
+      "loss": 0.3411,
+      "step": 9389
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6846506473225532e-05,
+      "loss": 0.3939,
+      "step": 9390
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6846170637919167e-05,
+      "loss": 0.2812,
+      "step": 9391
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6845834802612798e-05,
+      "loss": 0.2509,
+      "step": 9392
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6845498967306436e-05,
+      "loss": 0.2387,
+      "step": 9393
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6845163132000068e-05,
+      "loss": 0.2255,
+      "step": 9394
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6844827296693702e-05,
+      "loss": 0.3406,
+      "step": 9395
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6844491461387337e-05,
+      "loss": 0.5826,
+      "step": 9396
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.684415562608097e-05,
+      "loss": 0.3291,
+      "step": 9397
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6843819790774606e-05,
+      "loss": 0.3975,
+      "step": 9398
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.684348395546824e-05,
+      "loss": 0.2743,
+      "step": 9399
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6843148120161875e-05,
+      "loss": 0.3366,
+      "step": 9400
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6842812284855507e-05,
+      "loss": 0.453,
+      "step": 9401
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6842476449549145e-05,
+      "loss": 0.1668,
+      "step": 9402
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6842140614242776e-05,
+      "loss": 0.2729,
+      "step": 9403
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.684180477893641e-05,
+      "loss": 0.3624,
+      "step": 9404
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6841468943630045e-05,
+      "loss": 0.3195,
+      "step": 9405
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.684113310832368e-05,
+      "loss": 0.371,
+      "step": 9406
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6840797273017315e-05,
+      "loss": 0.3313,
+      "step": 9407
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.684046143771095e-05,
+      "loss": 0.3285,
+      "step": 9408
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6840125602404584e-05,
+      "loss": 0.4508,
+      "step": 9409
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6839789767098215e-05,
+      "loss": 0.3597,
+      "step": 9410
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6839453931791853e-05,
+      "loss": 0.3407,
+      "step": 9411
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6839118096485484e-05,
+      "loss": 0.3254,
+      "step": 9412
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.683878226117912e-05,
+      "loss": 0.3858,
+      "step": 9413
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6838446425872754e-05,
+      "loss": 0.3427,
+      "step": 9414
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.683811059056639e-05,
+      "loss": 0.27,
+      "step": 9415
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.683777475526002e-05,
+      "loss": 0.2111,
+      "step": 9416
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6837438919953658e-05,
+      "loss": 0.3487,
+      "step": 9417
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.683710308464729e-05,
+      "loss": 0.3058,
+      "step": 9418
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6836767249340924e-05,
+      "loss": 0.5064,
+      "step": 9419
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6836431414034558e-05,
+      "loss": 0.2666,
+      "step": 9420
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6836095578728193e-05,
+      "loss": 0.3118,
+      "step": 9421
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6835759743421827e-05,
+      "loss": 0.1816,
+      "step": 9422
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6835423908115462e-05,
+      "loss": 0.4443,
+      "step": 9423
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6835088072809097e-05,
+      "loss": 0.2719,
+      "step": 9424
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6834752237502728e-05,
+      "loss": 0.2361,
+      "step": 9425
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6834416402196366e-05,
+      "loss": 0.2582,
+      "step": 9426
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6834080566889997e-05,
+      "loss": 0.2289,
+      "step": 9427
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6833744731583632e-05,
+      "loss": 0.4008,
+      "step": 9428
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.6833408896277267e-05,
+      "loss": 0.3654,
+      "step": 9429
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.68330730609709e-05,
+      "loss": 0.3163,
+      "step": 9430
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6832737225664536e-05,
+      "loss": 0.316,
+      "step": 9431
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.683240139035817e-05,
+      "loss": 0.3847,
+      "step": 9432
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6832065555051805e-05,
+      "loss": 0.2844,
+      "step": 9433
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6831729719745436e-05,
+      "loss": 0.2334,
+      "step": 9434
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6831393884439075e-05,
+      "loss": 0.4286,
+      "step": 9435
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6831058049132706e-05,
+      "loss": 0.3302,
+      "step": 9436
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.683072221382634e-05,
+      "loss": 0.3932,
+      "step": 9437
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6830386378519975e-05,
+      "loss": 0.3534,
+      "step": 9438
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.683005054321361e-05,
+      "loss": 0.3789,
+      "step": 9439
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6829714707907244e-05,
+      "loss": 0.2832,
+      "step": 9440
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.682937887260088e-05,
+      "loss": 0.361,
+      "step": 9441
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6829043037294514e-05,
+      "loss": 0.21,
+      "step": 9442
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6828707201988145e-05,
+      "loss": 0.3388,
+      "step": 9443
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6828371366681783e-05,
+      "loss": 0.2576,
+      "step": 9444
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6828035531375414e-05,
+      "loss": 0.2502,
+      "step": 9445
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.682769969606905e-05,
+      "loss": 0.5282,
+      "step": 9446
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6827363860762684e-05,
+      "loss": 0.4149,
+      "step": 9447
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6827028025456318e-05,
+      "loss": 0.3679,
+      "step": 9448
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6826692190149953e-05,
+      "loss": 0.3885,
+      "step": 9449
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6826356354843587e-05,
+      "loss": 0.3398,
+      "step": 9450
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6826020519537222e-05,
+      "loss": 0.3468,
+      "step": 9451
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6825684684230853e-05,
+      "loss": 0.3084,
+      "step": 9452
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.682534884892449e-05,
+      "loss": 0.4184,
+      "step": 9453
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6825013013618123e-05,
+      "loss": 0.3178,
+      "step": 9454
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6824677178311757e-05,
+      "loss": 0.2717,
+      "step": 9455
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6824341343005392e-05,
+      "loss": 0.318,
+      "step": 9456
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6824005507699027e-05,
+      "loss": 0.3301,
+      "step": 9457
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.682366967239266e-05,
+      "loss": 0.3573,
+      "step": 9458
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6823333837086296e-05,
+      "loss": 0.2005,
+      "step": 9459
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6822998001779927e-05,
+      "loss": 0.3357,
+      "step": 9460
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6822662166473562e-05,
+      "loss": 0.2921,
+      "step": 9461
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6822326331167196e-05,
+      "loss": 0.4286,
+      "step": 9462
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.682199049586083e-05,
+      "loss": 0.3286,
+      "step": 9463
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6821654660554466e-05,
+      "loss": 0.3879,
+      "step": 9464
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.68213188252481e-05,
+      "loss": 0.4674,
+      "step": 9465
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6820982989941735e-05,
+      "loss": 0.2922,
+      "step": 9466
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6820647154635366e-05,
+      "loss": 0.3478,
+      "step": 9467
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6820311319329004e-05,
+      "loss": 0.29,
+      "step": 9468
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6819975484022636e-05,
+      "loss": 0.3752,
+      "step": 9469
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.681963964871627e-05,
+      "loss": 0.3432,
+      "step": 9470
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6819303813409905e-05,
+      "loss": 0.5312,
+      "step": 9471
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.681896797810354e-05,
+      "loss": 0.3215,
+      "step": 9472
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6818632142797174e-05,
+      "loss": 0.3744,
+      "step": 9473
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.681829630749081e-05,
+      "loss": 0.4247,
+      "step": 9474
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6817960472184443e-05,
+      "loss": 0.2587,
+      "step": 9475
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6817624636878075e-05,
+      "loss": 0.3999,
+      "step": 9476
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.681728880157171e-05,
+      "loss": 0.2678,
+      "step": 9477
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6816952966265344e-05,
+      "loss": 0.5602,
+      "step": 9478
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.681661713095898e-05,
+      "loss": 0.2998,
+      "step": 9479
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6816281295652613e-05,
+      "loss": 0.2463,
+      "step": 9480
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6815945460346248e-05,
+      "loss": 0.313,
+      "step": 9481
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6815609625039883e-05,
+      "loss": 0.3134,
+      "step": 9482
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6815273789733514e-05,
+      "loss": 0.3854,
+      "step": 9483
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6814937954427152e-05,
+      "loss": 0.2669,
+      "step": 9484
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6814602119120783e-05,
+      "loss": 0.4475,
+      "step": 9485
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6814266283814418e-05,
+      "loss": 0.313,
+      "step": 9486
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6813930448508052e-05,
+      "loss": 0.2077,
+      "step": 9487
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6813594613201687e-05,
+      "loss": 0.2186,
+      "step": 9488
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6813258777895322e-05,
+      "loss": 0.1753,
+      "step": 9489
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6812922942588956e-05,
+      "loss": 0.3473,
+      "step": 9490
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.681258710728259e-05,
+      "loss": 0.4426,
+      "step": 9491
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6812251271976222e-05,
+      "loss": 0.42,
+      "step": 9492
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.681191543666986e-05,
+      "loss": 0.3116,
+      "step": 9493
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.681157960136349e-05,
+      "loss": 0.275,
+      "step": 9494
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6811243766057126e-05,
+      "loss": 0.3878,
+      "step": 9495
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.681090793075076e-05,
+      "loss": 0.4472,
+      "step": 9496
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6810572095444396e-05,
+      "loss": 0.2711,
+      "step": 9497
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.681023626013803e-05,
+      "loss": 0.3857,
+      "step": 9498
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6809900424831665e-05,
+      "loss": 0.2858,
+      "step": 9499
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.68095645895253e-05,
+      "loss": 0.3319,
+      "step": 9500
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.680922875421893e-05,
+      "loss": 0.1883,
+      "step": 9501
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.680889291891257e-05,
+      "loss": 0.3857,
+      "step": 9502
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.68085570836062e-05,
+      "loss": 0.2899,
+      "step": 9503
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6808221248299835e-05,
+      "loss": 0.3459,
+      "step": 9504
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.680788541299347e-05,
+      "loss": 0.2947,
+      "step": 9505
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6807549577687104e-05,
+      "loss": 0.405,
+      "step": 9506
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6807213742380735e-05,
+      "loss": 0.4405,
+      "step": 9507
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6806877907074373e-05,
+      "loss": 0.3241,
+      "step": 9508
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6806542071768005e-05,
+      "loss": 0.3625,
+      "step": 9509
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.680620623646164e-05,
+      "loss": 0.3565,
+      "step": 9510
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6805870401155274e-05,
+      "loss": 0.2733,
+      "step": 9511
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.680553456584891e-05,
+      "loss": 0.305,
+      "step": 9512
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6805198730542543e-05,
+      "loss": 0.4443,
+      "step": 9513
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6804862895236178e-05,
+      "loss": 0.2601,
+      "step": 9514
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6804527059929812e-05,
+      "loss": 0.3295,
+      "step": 9515
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6804191224623444e-05,
+      "loss": 0.3361,
+      "step": 9516
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6803855389317082e-05,
+      "loss": 0.3245,
+      "step": 9517
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6803519554010713e-05,
+      "loss": 0.3463,
+      "step": 9518
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6803183718704348e-05,
+      "loss": 0.427,
+      "step": 9519
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6802847883397982e-05,
+      "loss": 0.37,
+      "step": 9520
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6802512048091617e-05,
+      "loss": 0.2504,
+      "step": 9521
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.680217621278525e-05,
+      "loss": 0.2134,
+      "step": 9522
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6801840377478886e-05,
+      "loss": 0.5929,
+      "step": 9523
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.680150454217252e-05,
+      "loss": 0.3561,
+      "step": 9524
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6801168706866152e-05,
+      "loss": 0.2252,
+      "step": 9525
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.680083287155979e-05,
+      "loss": 0.3988,
+      "step": 9526
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.680049703625342e-05,
+      "loss": 0.1707,
+      "step": 9527
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6800161200947056e-05,
+      "loss": 0.2725,
+      "step": 9528
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.679982536564069e-05,
+      "loss": 0.3838,
+      "step": 9529
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6799489530334325e-05,
+      "loss": 0.4788,
+      "step": 9530
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.679915369502796e-05,
+      "loss": 0.3278,
+      "step": 9531
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6798817859721595e-05,
+      "loss": 0.4427,
+      "step": 9532
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.679848202441523e-05,
+      "loss": 0.2963,
+      "step": 9533
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.679814618910886e-05,
+      "loss": 0.2057,
+      "step": 9534
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.67978103538025e-05,
+      "loss": 0.2299,
+      "step": 9535
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.679747451849613e-05,
+      "loss": 0.3551,
+      "step": 9536
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6797138683189765e-05,
+      "loss": 0.2657,
+      "step": 9537
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.67968028478834e-05,
+      "loss": 0.2787,
+      "step": 9538
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6796467012577034e-05,
+      "loss": 0.4251,
+      "step": 9539
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.679613117727067e-05,
+      "loss": 0.4638,
+      "step": 9540
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6795795341964303e-05,
+      "loss": 0.3697,
+      "step": 9541
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6795459506657938e-05,
+      "loss": 0.2428,
+      "step": 9542
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.679512367135157e-05,
+      "loss": 0.3707,
+      "step": 9543
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6794787836045207e-05,
+      "loss": 0.3193,
+      "step": 9544
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6794452000738838e-05,
+      "loss": 0.2302,
+      "step": 9545
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6794116165432473e-05,
+      "loss": 0.2377,
+      "step": 9546
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6793780330126108e-05,
+      "loss": 0.2673,
+      "step": 9547
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6793444494819742e-05,
+      "loss": 0.2876,
+      "step": 9548
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6793108659513377e-05,
+      "loss": 0.3851,
+      "step": 9549
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.679277282420701e-05,
+      "loss": 0.3312,
+      "step": 9550
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6792436988900643e-05,
+      "loss": 0.2989,
+      "step": 9551
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6792101153594277e-05,
+      "loss": 0.2398,
+      "step": 9552
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6791765318287912e-05,
+      "loss": 0.3211,
+      "step": 9553
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6791429482981547e-05,
+      "loss": 0.4288,
+      "step": 9554
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.679109364767518e-05,
+      "loss": 0.3445,
+      "step": 9555
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6790757812368816e-05,
+      "loss": 0.4038,
+      "step": 9556
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.679042197706245e-05,
+      "loss": 0.2925,
+      "step": 9557
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6790086141756082e-05,
+      "loss": 0.3554,
+      "step": 9558
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.678975030644972e-05,
+      "loss": 0.4786,
+      "step": 9559
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.678941447114335e-05,
+      "loss": 0.3712,
+      "step": 9560
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6789078635836986e-05,
+      "loss": 0.2225,
+      "step": 9561
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.678874280053062e-05,
+      "loss": 0.1551,
+      "step": 9562
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6788406965224255e-05,
+      "loss": 0.3944,
+      "step": 9563
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.678807112991789e-05,
+      "loss": 0.2672,
+      "step": 9564
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6787735294611524e-05,
+      "loss": 0.2861,
+      "step": 9565
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.678739945930516e-05,
+      "loss": 0.2751,
+      "step": 9566
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.678706362399879e-05,
+      "loss": 0.3482,
+      "step": 9567
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.678672778869243e-05,
+      "loss": 0.4906,
+      "step": 9568
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.678639195338606e-05,
+      "loss": 0.3128,
+      "step": 9569
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6786056118079694e-05,
+      "loss": 0.2714,
+      "step": 9570
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.678572028277333e-05,
+      "loss": 0.2886,
+      "step": 9571
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6785384447466964e-05,
+      "loss": 0.2543,
+      "step": 9572
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6785048612160598e-05,
+      "loss": 0.2857,
+      "step": 9573
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6784712776854233e-05,
+      "loss": 0.3817,
+      "step": 9574
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6784376941547868e-05,
+      "loss": 0.4935,
+      "step": 9575
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.67840411062415e-05,
+      "loss": 0.4552,
+      "step": 9576
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6783705270935137e-05,
+      "loss": 0.4204,
+      "step": 9577
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6783369435628768e-05,
+      "loss": 0.4783,
+      "step": 9578
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6783033600322403e-05,
+      "loss": 0.4954,
+      "step": 9579
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6782697765016037e-05,
+      "loss": 0.3253,
+      "step": 9580
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6782361929709672e-05,
+      "loss": 0.322,
+      "step": 9581
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6782026094403307e-05,
+      "loss": 0.247,
+      "step": 9582
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.678169025909694e-05,
+      "loss": 0.3627,
+      "step": 9583
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6781354423790576e-05,
+      "loss": 0.4249,
+      "step": 9584
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6781018588484207e-05,
+      "loss": 0.3786,
+      "step": 9585
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6780682753177845e-05,
+      "loss": 0.4674,
+      "step": 9586
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6780346917871477e-05,
+      "loss": 0.3289,
+      "step": 9587
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.678001108256511e-05,
+      "loss": 0.2917,
+      "step": 9588
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6779675247258746e-05,
+      "loss": 0.337,
+      "step": 9589
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.677933941195238e-05,
+      "loss": 0.1924,
+      "step": 9590
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6779003576646015e-05,
+      "loss": 0.3934,
+      "step": 9591
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.677866774133965e-05,
+      "loss": 0.3069,
+      "step": 9592
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6778331906033284e-05,
+      "loss": 0.3357,
+      "step": 9593
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6777996070726916e-05,
+      "loss": 0.1837,
+      "step": 9594
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.677766023542055e-05,
+      "loss": 0.3225,
+      "step": 9595
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6777324400114185e-05,
+      "loss": 0.3271,
+      "step": 9596
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.677698856480782e-05,
+      "loss": 0.2442,
+      "step": 9597
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6776652729501454e-05,
+      "loss": 0.2965,
+      "step": 9598
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.677631689419509e-05,
+      "loss": 0.301,
+      "step": 9599
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.677598105888872e-05,
+      "loss": 0.3182,
+      "step": 9600
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6775645223582358e-05,
+      "loss": 0.3132,
+      "step": 9601
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.677530938827599e-05,
+      "loss": 0.3438,
+      "step": 9602
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6774973552969624e-05,
+      "loss": 0.3268,
+      "step": 9603
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.677463771766326e-05,
+      "loss": 0.4384,
+      "step": 9604
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6774301882356893e-05,
+      "loss": 0.3118,
+      "step": 9605
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6773966047050528e-05,
+      "loss": 0.3306,
+      "step": 9606
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6773630211744163e-05,
+      "loss": 0.4086,
+      "step": 9607
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6773294376437797e-05,
+      "loss": 0.4172,
+      "step": 9608
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.677295854113143e-05,
+      "loss": 0.2971,
+      "step": 9609
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6772622705825067e-05,
+      "loss": 0.3414,
+      "step": 9610
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6772286870518698e-05,
+      "loss": 0.1924,
+      "step": 9611
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6771951035212333e-05,
+      "loss": 0.2623,
+      "step": 9612
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6771615199905967e-05,
+      "loss": 0.2392,
+      "step": 9613
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6771279364599602e-05,
+      "loss": 0.3238,
+      "step": 9614
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6770943529293237e-05,
+      "loss": 0.3368,
+      "step": 9615
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.677060769398687e-05,
+      "loss": 0.4167,
+      "step": 9616
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6770271858680506e-05,
+      "loss": 0.3756,
+      "step": 9617
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6769936023374137e-05,
+      "loss": 0.315,
+      "step": 9618
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6769600188067775e-05,
+      "loss": 0.3134,
+      "step": 9619
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6769264352761406e-05,
+      "loss": 0.2502,
+      "step": 9620
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.676892851745504e-05,
+      "loss": 0.223,
+      "step": 9621
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6768592682148676e-05,
+      "loss": 0.3818,
+      "step": 9622
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.676825684684231e-05,
+      "loss": 0.3597,
+      "step": 9623
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6767921011535945e-05,
+      "loss": 0.2159,
+      "step": 9624
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.676758517622958e-05,
+      "loss": 0.3412,
+      "step": 9625
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6767249340923214e-05,
+      "loss": 0.3092,
+      "step": 9626
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.6766913505616845e-05,
+      "loss": 0.3395,
+      "step": 9627
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6766577670310484e-05,
+      "loss": 0.467,
+      "step": 9628
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6766241835004115e-05,
+      "loss": 0.3254,
+      "step": 9629
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.676590599969775e-05,
+      "loss": 0.3526,
+      "step": 9630
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6765570164391384e-05,
+      "loss": 0.181,
+      "step": 9631
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.676523432908502e-05,
+      "loss": 0.2877,
+      "step": 9632
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6764898493778653e-05,
+      "loss": 0.3087,
+      "step": 9633
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6764562658472288e-05,
+      "loss": 0.3618,
+      "step": 9634
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6764226823165923e-05,
+      "loss": 0.3662,
+      "step": 9635
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6763890987859554e-05,
+      "loss": 0.4321,
+      "step": 9636
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6763555152553192e-05,
+      "loss": 0.2076,
+      "step": 9637
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6763219317246823e-05,
+      "loss": 0.4316,
+      "step": 9638
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6762883481940458e-05,
+      "loss": 0.3196,
+      "step": 9639
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6762547646634093e-05,
+      "loss": 0.3606,
+      "step": 9640
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6762211811327727e-05,
+      "loss": 0.2095,
+      "step": 9641
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.676187597602136e-05,
+      "loss": 0.2372,
+      "step": 9642
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6761540140714996e-05,
+      "loss": 0.429,
+      "step": 9643
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6761204305408628e-05,
+      "loss": 0.3908,
+      "step": 9644
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6760868470102262e-05,
+      "loss": 0.4075,
+      "step": 9645
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6760532634795897e-05,
+      "loss": 0.3208,
+      "step": 9646
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.676019679948953e-05,
+      "loss": 0.39,
+      "step": 9647
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6759860964183166e-05,
+      "loss": 0.3106,
+      "step": 9648
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.67595251288768e-05,
+      "loss": 0.3049,
+      "step": 9649
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6759189293570436e-05,
+      "loss": 0.3819,
+      "step": 9650
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6758853458264067e-05,
+      "loss": 0.29,
+      "step": 9651
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6758517622957705e-05,
+      "loss": 0.4094,
+      "step": 9652
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6758181787651336e-05,
+      "loss": 0.4455,
+      "step": 9653
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.675784595234497e-05,
+      "loss": 0.3847,
+      "step": 9654
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6757510117038605e-05,
+      "loss": 0.2595,
+      "step": 9655
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.675717428173224e-05,
+      "loss": 0.3246,
+      "step": 9656
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6756838446425875e-05,
+      "loss": 0.2254,
+      "step": 9657
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6756502611119506e-05,
+      "loss": 0.2863,
+      "step": 9658
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6756166775813144e-05,
+      "loss": 0.5127,
+      "step": 9659
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6755830940506775e-05,
+      "loss": 0.4293,
+      "step": 9660
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.675549510520041e-05,
+      "loss": 0.2979,
+      "step": 9661
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6755159269894045e-05,
+      "loss": 0.2788,
+      "step": 9662
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.675482343458768e-05,
+      "loss": 0.2716,
+      "step": 9663
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6754487599281314e-05,
+      "loss": 0.3695,
+      "step": 9664
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.675415176397495e-05,
+      "loss": 0.3998,
+      "step": 9665
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6753815928668583e-05,
+      "loss": 0.3016,
+      "step": 9666
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6753480093362214e-05,
+      "loss": 0.3059,
+      "step": 9667
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6753144258055852e-05,
+      "loss": 0.3433,
+      "step": 9668
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6752808422749484e-05,
+      "loss": 0.2239,
+      "step": 9669
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.675247258744312e-05,
+      "loss": 0.31,
+      "step": 9670
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6752136752136753e-05,
+      "loss": 0.2373,
+      "step": 9671
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6751800916830388e-05,
+      "loss": 0.2996,
+      "step": 9672
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6751465081524022e-05,
+      "loss": 0.2445,
+      "step": 9673
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6751129246217657e-05,
+      "loss": 0.5605,
+      "step": 9674
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.675079341091129e-05,
+      "loss": 0.3913,
+      "step": 9675
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6750457575604923e-05,
+      "loss": 0.2214,
+      "step": 9676
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.675012174029856e-05,
+      "loss": 0.2173,
+      "step": 9677
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6749785904992192e-05,
+      "loss": 0.5159,
+      "step": 9678
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6749450069685827e-05,
+      "loss": 0.3541,
+      "step": 9679
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.674911423437946e-05,
+      "loss": 0.3632,
+      "step": 9680
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6748778399073096e-05,
+      "loss": 0.3514,
+      "step": 9681
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.674844256376673e-05,
+      "loss": 0.3471,
+      "step": 9682
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6748106728460365e-05,
+      "loss": 0.268,
+      "step": 9683
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6747770893154e-05,
+      "loss": 0.1653,
+      "step": 9684
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.674743505784763e-05,
+      "loss": 0.3462,
+      "step": 9685
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6747099222541266e-05,
+      "loss": 0.3668,
+      "step": 9686
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.67467633872349e-05,
+      "loss": 0.1929,
+      "step": 9687
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6746427551928535e-05,
+      "loss": 0.2858,
+      "step": 9688
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.674609171662217e-05,
+      "loss": 0.3352,
+      "step": 9689
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6745755881315805e-05,
+      "loss": 0.4005,
+      "step": 9690
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6745420046009436e-05,
+      "loss": 0.4546,
+      "step": 9691
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6745084210703074e-05,
+      "loss": 0.5572,
+      "step": 9692
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6744748375396705e-05,
+      "loss": 0.2105,
+      "step": 9693
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.674441254009034e-05,
+      "loss": 0.2889,
+      "step": 9694
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6744076704783974e-05,
+      "loss": 0.3754,
+      "step": 9695
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.674374086947761e-05,
+      "loss": 0.4072,
+      "step": 9696
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6743405034171244e-05,
+      "loss": 0.3364,
+      "step": 9697
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.674306919886488e-05,
+      "loss": 0.3237,
+      "step": 9698
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6742733363558513e-05,
+      "loss": 0.5164,
+      "step": 9699
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6742397528252144e-05,
+      "loss": 0.5186,
+      "step": 9700
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6742061692945782e-05,
+      "loss": 0.4449,
+      "step": 9701
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6741725857639414e-05,
+      "loss": 0.2605,
+      "step": 9702
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6741390022333048e-05,
+      "loss": 0.4901,
+      "step": 9703
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6741054187026683e-05,
+      "loss": 0.3748,
+      "step": 9704
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6740718351720318e-05,
+      "loss": 0.4585,
+      "step": 9705
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6740382516413952e-05,
+      "loss": 0.3296,
+      "step": 9706
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6740046681107587e-05,
+      "loss": 0.4097,
+      "step": 9707
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.673971084580122e-05,
+      "loss": 0.4026,
+      "step": 9708
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6739375010494853e-05,
+      "loss": 0.4106,
+      "step": 9709
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.673903917518849e-05,
+      "loss": 0.2068,
+      "step": 9710
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6738703339882122e-05,
+      "loss": 0.3161,
+      "step": 9711
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6738367504575757e-05,
+      "loss": 0.4077,
+      "step": 9712
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.673803166926939e-05,
+      "loss": 0.3219,
+      "step": 9713
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6737695833963026e-05,
+      "loss": 0.3039,
+      "step": 9714
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.673735999865666e-05,
+      "loss": 0.2863,
+      "step": 9715
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6737024163350295e-05,
+      "loss": 0.2597,
+      "step": 9716
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.673668832804393e-05,
+      "loss": 0.3135,
+      "step": 9717
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.673635249273756e-05,
+      "loss": 0.1783,
+      "step": 9718
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.67360166574312e-05,
+      "loss": 0.2682,
+      "step": 9719
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.673568082212483e-05,
+      "loss": 0.5214,
+      "step": 9720
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6735344986818465e-05,
+      "loss": 0.3218,
+      "step": 9721
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.67350091515121e-05,
+      "loss": 0.3271,
+      "step": 9722
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6734673316205734e-05,
+      "loss": 0.4584,
+      "step": 9723
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.673433748089937e-05,
+      "loss": 0.5138,
+      "step": 9724
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6734001645593004e-05,
+      "loss": 0.3631,
+      "step": 9725
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.673366581028664e-05,
+      "loss": 0.3409,
+      "step": 9726
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.673332997498027e-05,
+      "loss": 0.2563,
+      "step": 9727
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6732994139673908e-05,
+      "loss": 0.4106,
+      "step": 9728
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.673265830436754e-05,
+      "loss": 0.3577,
+      "step": 9729
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6732322469061174e-05,
+      "loss": 0.3221,
+      "step": 9730
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6731986633754808e-05,
+      "loss": 0.3195,
+      "step": 9731
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6731650798448443e-05,
+      "loss": 0.3854,
+      "step": 9732
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6731314963142074e-05,
+      "loss": 0.4091,
+      "step": 9733
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6730979127835712e-05,
+      "loss": 0.572,
+      "step": 9734
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6730643292529343e-05,
+      "loss": 0.2992,
+      "step": 9735
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6730307457222978e-05,
+      "loss": 0.4836,
+      "step": 9736
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6729971621916613e-05,
+      "loss": 0.2335,
+      "step": 9737
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6729635786610247e-05,
+      "loss": 0.3515,
+      "step": 9738
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6729299951303882e-05,
+      "loss": 0.2716,
+      "step": 9739
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6728964115997517e-05,
+      "loss": 0.2248,
+      "step": 9740
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.672862828069115e-05,
+      "loss": 0.4149,
+      "step": 9741
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6728292445384783e-05,
+      "loss": 0.3659,
+      "step": 9742
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.672795661007842e-05,
+      "loss": 0.3253,
+      "step": 9743
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6727620774772052e-05,
+      "loss": 0.4051,
+      "step": 9744
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6727284939465686e-05,
+      "loss": 0.4885,
+      "step": 9745
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.672694910415932e-05,
+      "loss": 0.3806,
+      "step": 9746
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6726613268852956e-05,
+      "loss": 0.2161,
+      "step": 9747
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.672627743354659e-05,
+      "loss": 0.2575,
+      "step": 9748
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6725941598240225e-05,
+      "loss": 0.2432,
+      "step": 9749
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.672560576293386e-05,
+      "loss": 0.4041,
+      "step": 9750
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.672526992762749e-05,
+      "loss": 0.4645,
+      "step": 9751
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.672493409232113e-05,
+      "loss": 0.3473,
+      "step": 9752
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.672459825701476e-05,
+      "loss": 0.4512,
+      "step": 9753
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6724262421708395e-05,
+      "loss": 0.3436,
+      "step": 9754
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.672392658640203e-05,
+      "loss": 0.2998,
+      "step": 9755
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6723590751095664e-05,
+      "loss": 0.5616,
+      "step": 9756
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.67232549157893e-05,
+      "loss": 0.2346,
+      "step": 9757
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6722919080482933e-05,
+      "loss": 0.2614,
+      "step": 9758
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6722583245176568e-05,
+      "loss": 0.3405,
+      "step": 9759
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.67222474098702e-05,
+      "loss": 0.3646,
+      "step": 9760
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6721911574563837e-05,
+      "loss": 0.2887,
+      "step": 9761
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.672157573925747e-05,
+      "loss": 0.4373,
+      "step": 9762
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6721239903951103e-05,
+      "loss": 0.2677,
+      "step": 9763
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6720904068644738e-05,
+      "loss": 0.2,
+      "step": 9764
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6720568233338373e-05,
+      "loss": 0.3566,
+      "step": 9765
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6720232398032007e-05,
+      "loss": 0.3766,
+      "step": 9766
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6719896562725642e-05,
+      "loss": 0.3995,
+      "step": 9767
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6719560727419277e-05,
+      "loss": 0.3485,
+      "step": 9768
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6719224892112908e-05,
+      "loss": 0.5112,
+      "step": 9769
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6718889056806546e-05,
+      "loss": 0.3926,
+      "step": 9770
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6718553221500177e-05,
+      "loss": 0.183,
+      "step": 9771
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6718217386193812e-05,
+      "loss": 0.3698,
+      "step": 9772
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6717881550887446e-05,
+      "loss": 0.2649,
+      "step": 9773
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.671754571558108e-05,
+      "loss": 0.4629,
+      "step": 9774
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6717209880274716e-05,
+      "loss": 0.5008,
+      "step": 9775
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.671687404496835e-05,
+      "loss": 0.478,
+      "step": 9776
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.671653820966198e-05,
+      "loss": 0.2158,
+      "step": 9777
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6716202374355616e-05,
+      "loss": 0.3301,
+      "step": 9778
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.671586653904925e-05,
+      "loss": 0.3617,
+      "step": 9779
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6715530703742886e-05,
+      "loss": 0.391,
+      "step": 9780
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.671519486843652e-05,
+      "loss": 0.3465,
+      "step": 9781
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6714859033130155e-05,
+      "loss": 0.3341,
+      "step": 9782
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.671452319782379e-05,
+      "loss": 0.4459,
+      "step": 9783
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.671418736251742e-05,
+      "loss": 0.1444,
+      "step": 9784
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.671385152721106e-05,
+      "loss": 0.3145,
+      "step": 9785
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.671351569190469e-05,
+      "loss": 0.3121,
+      "step": 9786
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6713179856598325e-05,
+      "loss": 0.3724,
+      "step": 9787
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.671284402129196e-05,
+      "loss": 0.3562,
+      "step": 9788
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6712508185985594e-05,
+      "loss": 0.1664,
+      "step": 9789
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.671217235067923e-05,
+      "loss": 0.2084,
+      "step": 9790
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6711836515372863e-05,
+      "loss": 0.2046,
+      "step": 9791
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6711500680066498e-05,
+      "loss": 0.3323,
+      "step": 9792
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.671116484476013e-05,
+      "loss": 0.3308,
+      "step": 9793
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6710829009453767e-05,
+      "loss": 0.4189,
+      "step": 9794
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.67104931741474e-05,
+      "loss": 0.48,
+      "step": 9795
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6710157338841033e-05,
+      "loss": 0.4629,
+      "step": 9796
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6709821503534668e-05,
+      "loss": 0.3888,
+      "step": 9797
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6709485668228302e-05,
+      "loss": 0.4725,
+      "step": 9798
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6709149832921937e-05,
+      "loss": 0.2708,
+      "step": 9799
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6708813997615572e-05,
+      "loss": 0.2434,
+      "step": 9800
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6708478162309206e-05,
+      "loss": 0.2485,
+      "step": 9801
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6708142327002838e-05,
+      "loss": 0.2496,
+      "step": 9802
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6707806491696476e-05,
+      "loss": 0.5796,
+      "step": 9803
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6707470656390107e-05,
+      "loss": 0.5268,
+      "step": 9804
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.670713482108374e-05,
+      "loss": 0.2686,
+      "step": 9805
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6706798985777376e-05,
+      "loss": 0.298,
+      "step": 9806
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.670646315047101e-05,
+      "loss": 0.3235,
+      "step": 9807
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6706127315164646e-05,
+      "loss": 0.2187,
+      "step": 9808
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.670579147985828e-05,
+      "loss": 0.286,
+      "step": 9809
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6705455644551915e-05,
+      "loss": 0.309,
+      "step": 9810
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6705119809245546e-05,
+      "loss": 0.3852,
+      "step": 9811
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6704783973939184e-05,
+      "loss": 0.4522,
+      "step": 9812
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6704448138632815e-05,
+      "loss": 0.3165,
+      "step": 9813
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.670411230332645e-05,
+      "loss": 0.5028,
+      "step": 9814
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6703776468020085e-05,
+      "loss": 0.5094,
+      "step": 9815
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.670344063271372e-05,
+      "loss": 0.4168,
+      "step": 9816
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6703104797407354e-05,
+      "loss": 0.4515,
+      "step": 9817
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.670276896210099e-05,
+      "loss": 0.3568,
+      "step": 9818
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6702433126794623e-05,
+      "loss": 0.2733,
+      "step": 9819
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6702097291488255e-05,
+      "loss": 0.4405,
+      "step": 9820
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.670176145618189e-05,
+      "loss": 0.3432,
+      "step": 9821
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6701425620875524e-05,
+      "loss": 0.4381,
+      "step": 9822
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.670108978556916e-05,
+      "loss": 0.5124,
+      "step": 9823
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6700753950262793e-05,
+      "loss": 0.4914,
+      "step": 9824
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.6700418114956428e-05,
+      "loss": 0.3829,
+      "step": 9825
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.670008227965006e-05,
+      "loss": 0.3312,
+      "step": 9826
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6699746444343697e-05,
+      "loss": 0.2761,
+      "step": 9827
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.669941060903733e-05,
+      "loss": 0.2452,
+      "step": 9828
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6699074773730963e-05,
+      "loss": 0.2257,
+      "step": 9829
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6698738938424598e-05,
+      "loss": 0.3382,
+      "step": 9830
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6698403103118232e-05,
+      "loss": 0.3207,
+      "step": 9831
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6698067267811867e-05,
+      "loss": 0.3793,
+      "step": 9832
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.66977314325055e-05,
+      "loss": 0.31,
+      "step": 9833
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6697395597199136e-05,
+      "loss": 0.3457,
+      "step": 9834
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6697059761892767e-05,
+      "loss": 0.2435,
+      "step": 9835
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6696723926586402e-05,
+      "loss": 0.3786,
+      "step": 9836
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6696388091280037e-05,
+      "loss": 0.2393,
+      "step": 9837
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.669605225597367e-05,
+      "loss": 0.3468,
+      "step": 9838
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6695716420667306e-05,
+      "loss": 0.336,
+      "step": 9839
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.669538058536094e-05,
+      "loss": 0.3289,
+      "step": 9840
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6695044750054575e-05,
+      "loss": 0.3481,
+      "step": 9841
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6694708914748207e-05,
+      "loss": 0.3351,
+      "step": 9842
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6694373079441845e-05,
+      "loss": 0.3568,
+      "step": 9843
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6694037244135476e-05,
+      "loss": 0.3177,
+      "step": 9844
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.669370140882911e-05,
+      "loss": 0.3437,
+      "step": 9845
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6693365573522745e-05,
+      "loss": 0.3408,
+      "step": 9846
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.669302973821638e-05,
+      "loss": 0.2929,
+      "step": 9847
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6692693902910014e-05,
+      "loss": 0.3142,
+      "step": 9848
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.669235806760365e-05,
+      "loss": 0.4858,
+      "step": 9849
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6692022232297284e-05,
+      "loss": 0.2281,
+      "step": 9850
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6691686396990915e-05,
+      "loss": 0.3423,
+      "step": 9851
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6691350561684553e-05,
+      "loss": 0.4046,
+      "step": 9852
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6691014726378184e-05,
+      "loss": 0.2826,
+      "step": 9853
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.669067889107182e-05,
+      "loss": 0.3611,
+      "step": 9854
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6690343055765454e-05,
+      "loss": 0.3464,
+      "step": 9855
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6690007220459088e-05,
+      "loss": 0.3393,
+      "step": 9856
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6689671385152723e-05,
+      "loss": 0.4129,
+      "step": 9857
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6689335549846358e-05,
+      "loss": 0.3491,
+      "step": 9858
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6688999714539992e-05,
+      "loss": 0.4481,
+      "step": 9859
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6688663879233623e-05,
+      "loss": 0.317,
+      "step": 9860
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.668832804392726e-05,
+      "loss": 0.1965,
+      "step": 9861
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6687992208620893e-05,
+      "loss": 0.1772,
+      "step": 9862
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6687656373314527e-05,
+      "loss": 0.3409,
+      "step": 9863
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6687320538008162e-05,
+      "loss": 0.3029,
+      "step": 9864
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6686984702701797e-05,
+      "loss": 0.3057,
+      "step": 9865
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6686648867395428e-05,
+      "loss": 0.323,
+      "step": 9866
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6686313032089066e-05,
+      "loss": 0.3508,
+      "step": 9867
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6685977196782697e-05,
+      "loss": 0.35,
+      "step": 9868
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6685641361476332e-05,
+      "loss": 0.3177,
+      "step": 9869
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6685305526169967e-05,
+      "loss": 0.4066,
+      "step": 9870
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.66849696908636e-05,
+      "loss": 0.4481,
+      "step": 9871
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6684633855557236e-05,
+      "loss": 0.3414,
+      "step": 9872
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.668429802025087e-05,
+      "loss": 0.1603,
+      "step": 9873
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6683962184944505e-05,
+      "loss": 0.6339,
+      "step": 9874
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6683626349638136e-05,
+      "loss": 0.3123,
+      "step": 9875
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6683290514331774e-05,
+      "loss": 0.352,
+      "step": 9876
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6682954679025406e-05,
+      "loss": 0.273,
+      "step": 9877
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.668261884371904e-05,
+      "loss": 0.3748,
+      "step": 9878
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6682283008412675e-05,
+      "loss": 0.3631,
+      "step": 9879
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.668194717310631e-05,
+      "loss": 0.2748,
+      "step": 9880
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6681611337799944e-05,
+      "loss": 0.2602,
+      "step": 9881
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.668127550249358e-05,
+      "loss": 0.4054,
+      "step": 9882
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6680939667187214e-05,
+      "loss": 0.3257,
+      "step": 9883
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6680603831880845e-05,
+      "loss": 0.2792,
+      "step": 9884
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6680267996574483e-05,
+      "loss": 0.3177,
+      "step": 9885
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6679932161268114e-05,
+      "loss": 0.4092,
+      "step": 9886
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.667959632596175e-05,
+      "loss": 0.4032,
+      "step": 9887
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6679260490655383e-05,
+      "loss": 0.4739,
+      "step": 9888
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6678924655349018e-05,
+      "loss": 0.237,
+      "step": 9889
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6678588820042653e-05,
+      "loss": 0.2342,
+      "step": 9890
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6678252984736287e-05,
+      "loss": 0.2888,
+      "step": 9891
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6677917149429922e-05,
+      "loss": 0.3992,
+      "step": 9892
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6677581314123553e-05,
+      "loss": 0.1982,
+      "step": 9893
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.667724547881719e-05,
+      "loss": 0.2321,
+      "step": 9894
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6676909643510823e-05,
+      "loss": 0.2053,
+      "step": 9895
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6676573808204457e-05,
+      "loss": 0.4793,
+      "step": 9896
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6676237972898092e-05,
+      "loss": 0.1824,
+      "step": 9897
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6675902137591727e-05,
+      "loss": 0.2733,
+      "step": 9898
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.667556630228536e-05,
+      "loss": 0.3684,
+      "step": 9899
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6675230466978996e-05,
+      "loss": 0.6904,
+      "step": 9900
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.667489463167263e-05,
+      "loss": 0.5431,
+      "step": 9901
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6674558796366262e-05,
+      "loss": 0.5765,
+      "step": 9902
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.66742229610599e-05,
+      "loss": 0.2675,
+      "step": 9903
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.667388712575353e-05,
+      "loss": 0.3567,
+      "step": 9904
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6673551290447166e-05,
+      "loss": 0.3794,
+      "step": 9905
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.66732154551408e-05,
+      "loss": 0.3092,
+      "step": 9906
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6672879619834435e-05,
+      "loss": 0.271,
+      "step": 9907
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.667254378452807e-05,
+      "loss": 0.2908,
+      "step": 9908
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6672207949221704e-05,
+      "loss": 0.3578,
+      "step": 9909
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.667187211391534e-05,
+      "loss": 0.4197,
+      "step": 9910
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.667153627860897e-05,
+      "loss": 0.2845,
+      "step": 9911
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6671200443302605e-05,
+      "loss": 0.4271,
+      "step": 9912
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.667086460799624e-05,
+      "loss": 0.4798,
+      "step": 9913
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6670528772689874e-05,
+      "loss": 0.1999,
+      "step": 9914
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.667019293738351e-05,
+      "loss": 0.3808,
+      "step": 9915
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6669857102077143e-05,
+      "loss": 0.314,
+      "step": 9916
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6669521266770775e-05,
+      "loss": 0.4063,
+      "step": 9917
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6669185431464413e-05,
+      "loss": 0.302,
+      "step": 9918
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6668849596158044e-05,
+      "loss": 0.2724,
+      "step": 9919
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.666851376085168e-05,
+      "loss": 0.3486,
+      "step": 9920
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6668177925545313e-05,
+      "loss": 0.3303,
+      "step": 9921
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6667842090238948e-05,
+      "loss": 0.4202,
+      "step": 9922
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6667506254932583e-05,
+      "loss": 0.6058,
+      "step": 9923
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6667170419626217e-05,
+      "loss": 0.3091,
+      "step": 9924
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6666834584319852e-05,
+      "loss": 0.2529,
+      "step": 9925
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6666498749013483e-05,
+      "loss": 0.2331,
+      "step": 9926
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.666616291370712e-05,
+      "loss": 0.4162,
+      "step": 9927
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6665827078400752e-05,
+      "loss": 0.3907,
+      "step": 9928
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6665491243094387e-05,
+      "loss": 0.1739,
+      "step": 9929
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.666515540778802e-05,
+      "loss": 0.3701,
+      "step": 9930
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6664819572481656e-05,
+      "loss": 0.2575,
+      "step": 9931
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.666448373717529e-05,
+      "loss": 0.3763,
+      "step": 9932
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6664147901868926e-05,
+      "loss": 0.4975,
+      "step": 9933
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.666381206656256e-05,
+      "loss": 0.3997,
+      "step": 9934
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.666347623125619e-05,
+      "loss": 0.3841,
+      "step": 9935
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.666314039594983e-05,
+      "loss": 0.3865,
+      "step": 9936
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.666280456064346e-05,
+      "loss": 0.3802,
+      "step": 9937
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6662468725337095e-05,
+      "loss": 0.3456,
+      "step": 9938
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.666213289003073e-05,
+      "loss": 0.156,
+      "step": 9939
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6661797054724365e-05,
+      "loss": 0.3144,
+      "step": 9940
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6661461219418e-05,
+      "loss": 0.6417,
+      "step": 9941
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6661125384111634e-05,
+      "loss": 0.4836,
+      "step": 9942
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.666078954880527e-05,
+      "loss": 0.3297,
+      "step": 9943
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.66604537134989e-05,
+      "loss": 0.3639,
+      "step": 9944
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6660117878192538e-05,
+      "loss": 0.4597,
+      "step": 9945
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.665978204288617e-05,
+      "loss": 0.3221,
+      "step": 9946
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6659446207579804e-05,
+      "loss": 0.2899,
+      "step": 9947
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.665911037227344e-05,
+      "loss": 0.3626,
+      "step": 9948
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6658774536967073e-05,
+      "loss": 0.2846,
+      "step": 9949
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6658438701660708e-05,
+      "loss": 0.2822,
+      "step": 9950
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6658102866354343e-05,
+      "loss": 0.4409,
+      "step": 9951
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6657767031047977e-05,
+      "loss": 0.3916,
+      "step": 9952
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.665743119574161e-05,
+      "loss": 0.3012,
+      "step": 9953
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6657095360435246e-05,
+      "loss": 0.2452,
+      "step": 9954
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6656759525128878e-05,
+      "loss": 0.2045,
+      "step": 9955
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6656423689822512e-05,
+      "loss": 0.4335,
+      "step": 9956
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6656087854516147e-05,
+      "loss": 0.3394,
+      "step": 9957
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.665575201920978e-05,
+      "loss": 0.2969,
+      "step": 9958
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6655416183903413e-05,
+      "loss": 0.2688,
+      "step": 9959
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.665508034859705e-05,
+      "loss": 0.4919,
+      "step": 9960
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6654744513290682e-05,
+      "loss": 0.5024,
+      "step": 9961
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6654408677984317e-05,
+      "loss": 0.2108,
+      "step": 9962
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.665407284267795e-05,
+      "loss": 0.5837,
+      "step": 9963
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6653737007371586e-05,
+      "loss": 0.3461,
+      "step": 9964
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.665340117206522e-05,
+      "loss": 0.1941,
+      "step": 9965
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6653065336758855e-05,
+      "loss": 0.3413,
+      "step": 9966
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.665272950145249e-05,
+      "loss": 0.3604,
+      "step": 9967
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.665239366614612e-05,
+      "loss": 0.4254,
+      "step": 9968
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.665205783083976e-05,
+      "loss": 0.1921,
+      "step": 9969
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.665172199553339e-05,
+      "loss": 0.3564,
+      "step": 9970
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6651386160227025e-05,
+      "loss": 0.2575,
+      "step": 9971
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.665105032492066e-05,
+      "loss": 0.4729,
+      "step": 9972
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6650714489614295e-05,
+      "loss": 0.4846,
+      "step": 9973
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.665037865430793e-05,
+      "loss": 0.4123,
+      "step": 9974
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6650042819001564e-05,
+      "loss": 0.1604,
+      "step": 9975
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.66497069836952e-05,
+      "loss": 0.3201,
+      "step": 9976
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.664937114838883e-05,
+      "loss": 0.1586,
+      "step": 9977
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6649035313082468e-05,
+      "loss": 0.2595,
+      "step": 9978
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.66486994777761e-05,
+      "loss": 0.3679,
+      "step": 9979
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6648363642469734e-05,
+      "loss": 0.4908,
+      "step": 9980
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.664802780716337e-05,
+      "loss": 0.4006,
+      "step": 9981
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6647691971857003e-05,
+      "loss": 0.2788,
+      "step": 9982
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6647356136550638e-05,
+      "loss": 0.6035,
+      "step": 9983
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6647020301244272e-05,
+      "loss": 0.4259,
+      "step": 9984
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6646684465937907e-05,
+      "loss": 0.2039,
+      "step": 9985
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6646348630631538e-05,
+      "loss": 0.5213,
+      "step": 9986
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6646012795325176e-05,
+      "loss": 0.3049,
+      "step": 9987
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6645676960018808e-05,
+      "loss": 0.372,
+      "step": 9988
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6645341124712442e-05,
+      "loss": 0.3343,
+      "step": 9989
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6645005289406077e-05,
+      "loss": 0.3613,
+      "step": 9990
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.664466945409971e-05,
+      "loss": 0.2676,
+      "step": 9991
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6644333618793346e-05,
+      "loss": 0.2451,
+      "step": 9992
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.664399778348698e-05,
+      "loss": 0.4242,
+      "step": 9993
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6643661948180615e-05,
+      "loss": 0.3066,
+      "step": 9994
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6643326112874247e-05,
+      "loss": 0.2865,
+      "step": 9995
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6642990277567885e-05,
+      "loss": 0.4098,
+      "step": 9996
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6642654442261516e-05,
+      "loss": 0.15,
+      "step": 9997
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.664231860695515e-05,
+      "loss": 0.182,
+      "step": 9998
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6641982771648785e-05,
+      "loss": 0.4975,
+      "step": 9999
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.664164693634242e-05,
+      "loss": 0.1989,
+      "step": 10000
+    },
+    {
+      "epoch": 0.5,
+      "eval_accuracy": 0.8759871876813671,
+      "eval_f1_microaverage": 0.8759871876813671,
+      "eval_loss": 0.33414140343666077,
+      "eval_precision_macroaverage": 0.7432116183627921,
+      "eval_precision_microaverage": 0.8759871876813671,
+      "eval_recall_macroaverage": 0.6253252741754043,
+      "eval_recall_microaverage": 0.8759871876813671,
+      "eval_runtime": 1177.381,
+      "eval_samples_per_second": 231.223,
+      "eval_steps_per_second": 3.613,
+      "step": 10000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6641311101036055e-05,
+      "loss": 0.3473,
+      "step": 10001
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.664097526572969e-05,
+      "loss": 0.4707,
+      "step": 10002
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.664063943042332e-05,
+      "loss": 0.25,
+      "step": 10003
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6640303595116955e-05,
+      "loss": 0.3441,
+      "step": 10004
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.663996775981059e-05,
+      "loss": 0.2054,
+      "step": 10005
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6639631924504224e-05,
+      "loss": 0.3836,
+      "step": 10006
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.663929608919786e-05,
+      "loss": 0.4647,
+      "step": 10007
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6638960253891494e-05,
+      "loss": 0.4231,
+      "step": 10008
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.663862441858513e-05,
+      "loss": 0.2778,
+      "step": 10009
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.663828858327876e-05,
+      "loss": 0.3618,
+      "step": 10010
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6637952747972398e-05,
+      "loss": 0.3272,
+      "step": 10011
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.663761691266603e-05,
+      "loss": 0.3361,
+      "step": 10012
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6637281077359664e-05,
+      "loss": 0.2672,
+      "step": 10013
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6636945242053298e-05,
+      "loss": 0.2385,
+      "step": 10014
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6636609406746933e-05,
+      "loss": 0.4725,
+      "step": 10015
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6636273571440567e-05,
+      "loss": 0.3877,
+      "step": 10016
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.66359377361342e-05,
+      "loss": 0.3288,
+      "step": 10017
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6635601900827837e-05,
+      "loss": 0.3302,
+      "step": 10018
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6635266065521468e-05,
+      "loss": 0.3018,
+      "step": 10019
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6634930230215103e-05,
+      "loss": 0.309,
+      "step": 10020
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6634594394908737e-05,
+      "loss": 0.3949,
+      "step": 10021
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6634258559602372e-05,
+      "loss": 0.4454,
+      "step": 10022
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.6633922724296007e-05,
+      "loss": 0.4627,
+      "step": 10023
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.663358688898964e-05,
+      "loss": 0.3625,
+      "step": 10024
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6633251053683276e-05,
+      "loss": 0.4733,
+      "step": 10025
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6632915218376907e-05,
+      "loss": 0.193,
+      "step": 10026
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6632579383070545e-05,
+      "loss": 0.2321,
+      "step": 10027
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6632243547764176e-05,
+      "loss": 0.2178,
+      "step": 10028
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.663190771245781e-05,
+      "loss": 0.4154,
+      "step": 10029
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6631571877151446e-05,
+      "loss": 0.3497,
+      "step": 10030
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.663123604184508e-05,
+      "loss": 0.4278,
+      "step": 10031
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6630900206538715e-05,
+      "loss": 0.2932,
+      "step": 10032
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.663056437123235e-05,
+      "loss": 0.1894,
+      "step": 10033
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6630228535925984e-05,
+      "loss": 0.4584,
+      "step": 10034
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6629892700619616e-05,
+      "loss": 0.413,
+      "step": 10035
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6629556865313254e-05,
+      "loss": 0.2372,
+      "step": 10036
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6629221030006885e-05,
+      "loss": 0.2485,
+      "step": 10037
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.662888519470052e-05,
+      "loss": 0.4734,
+      "step": 10038
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6628549359394154e-05,
+      "loss": 0.2185,
+      "step": 10039
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.662821352408779e-05,
+      "loss": 0.3128,
+      "step": 10040
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6627877688781423e-05,
+      "loss": 0.3231,
+      "step": 10041
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6627541853475058e-05,
+      "loss": 0.3906,
+      "step": 10042
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6627206018168693e-05,
+      "loss": 0.2777,
+      "step": 10043
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6626870182862324e-05,
+      "loss": 0.2665,
+      "step": 10044
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6626534347555962e-05,
+      "loss": 0.3222,
+      "step": 10045
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6626198512249593e-05,
+      "loss": 0.2065,
+      "step": 10046
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6625862676943228e-05,
+      "loss": 0.2072,
+      "step": 10047
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6625526841636863e-05,
+      "loss": 0.3706,
+      "step": 10048
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6625191006330497e-05,
+      "loss": 0.3869,
+      "step": 10049
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.662485517102413e-05,
+      "loss": 0.3828,
+      "step": 10050
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6624519335717767e-05,
+      "loss": 0.3625,
+      "step": 10051
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6624183500411398e-05,
+      "loss": 0.5685,
+      "step": 10052
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6623847665105032e-05,
+      "loss": 0.2815,
+      "step": 10053
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6623511829798667e-05,
+      "loss": 0.3683,
+      "step": 10054
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6623175994492302e-05,
+      "loss": 0.2549,
+      "step": 10055
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6622840159185936e-05,
+      "loss": 0.3225,
+      "step": 10056
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.662250432387957e-05,
+      "loss": 0.2565,
+      "step": 10057
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6622168488573206e-05,
+      "loss": 0.4173,
+      "step": 10058
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6621832653266837e-05,
+      "loss": 0.3216,
+      "step": 10059
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6621496817960475e-05,
+      "loss": 0.2284,
+      "step": 10060
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6621160982654106e-05,
+      "loss": 0.3495,
+      "step": 10061
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.662082514734774e-05,
+      "loss": 0.2816,
+      "step": 10062
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6620489312041376e-05,
+      "loss": 0.2962,
+      "step": 10063
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.662015347673501e-05,
+      "loss": 0.2714,
+      "step": 10064
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6619817641428645e-05,
+      "loss": 0.3421,
+      "step": 10065
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.661948180612228e-05,
+      "loss": 0.4232,
+      "step": 10066
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6619145970815914e-05,
+      "loss": 0.5454,
+      "step": 10067
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6618810135509545e-05,
+      "loss": 0.3384,
+      "step": 10068
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6618474300203183e-05,
+      "loss": 0.2908,
+      "step": 10069
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6618138464896815e-05,
+      "loss": 0.434,
+      "step": 10070
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.661780262959045e-05,
+      "loss": 0.4527,
+      "step": 10071
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6617466794284084e-05,
+      "loss": 0.3712,
+      "step": 10072
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.661713095897772e-05,
+      "loss": 0.2637,
+      "step": 10073
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6616795123671353e-05,
+      "loss": 0.3796,
+      "step": 10074
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6616459288364988e-05,
+      "loss": 0.4214,
+      "step": 10075
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6616123453058623e-05,
+      "loss": 0.3623,
+      "step": 10076
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6615787617752254e-05,
+      "loss": 0.1548,
+      "step": 10077
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6615451782445892e-05,
+      "loss": 0.3425,
+      "step": 10078
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6615115947139523e-05,
+      "loss": 0.3091,
+      "step": 10079
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6614780111833158e-05,
+      "loss": 0.2021,
+      "step": 10080
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6614444276526792e-05,
+      "loss": 0.3132,
+      "step": 10081
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6614108441220427e-05,
+      "loss": 0.2677,
+      "step": 10082
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6613772605914062e-05,
+      "loss": 0.4061,
+      "step": 10083
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6613436770607696e-05,
+      "loss": 0.2899,
+      "step": 10084
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.661310093530133e-05,
+      "loss": 0.2491,
+      "step": 10085
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6612765099994962e-05,
+      "loss": 0.2278,
+      "step": 10086
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.66124292646886e-05,
+      "loss": 0.6773,
+      "step": 10087
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.661209342938223e-05,
+      "loss": 0.3482,
+      "step": 10088
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6611757594075866e-05,
+      "loss": 0.3791,
+      "step": 10089
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.66114217587695e-05,
+      "loss": 0.3105,
+      "step": 10090
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6611085923463136e-05,
+      "loss": 0.3721,
+      "step": 10091
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6610750088156767e-05,
+      "loss": 0.2379,
+      "step": 10092
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6610414252850405e-05,
+      "loss": 0.3869,
+      "step": 10093
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6610078417544036e-05,
+      "loss": 0.3265,
+      "step": 10094
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.660974258223767e-05,
+      "loss": 0.303,
+      "step": 10095
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6609406746931305e-05,
+      "loss": 0.269,
+      "step": 10096
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.660907091162494e-05,
+      "loss": 0.3712,
+      "step": 10097
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6608735076318575e-05,
+      "loss": 0.485,
+      "step": 10098
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.660839924101221e-05,
+      "loss": 0.2592,
+      "step": 10099
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6608063405705844e-05,
+      "loss": 0.2975,
+      "step": 10100
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6607727570399475e-05,
+      "loss": 0.305,
+      "step": 10101
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6607391735093113e-05,
+      "loss": 0.2984,
+      "step": 10102
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6607055899786745e-05,
+      "loss": 0.335,
+      "step": 10103
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.660672006448038e-05,
+      "loss": 0.3154,
+      "step": 10104
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6606384229174014e-05,
+      "loss": 0.2754,
+      "step": 10105
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.660604839386765e-05,
+      "loss": 0.2938,
+      "step": 10106
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6605712558561283e-05,
+      "loss": 0.289,
+      "step": 10107
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6605376723254918e-05,
+      "loss": 0.3016,
+      "step": 10108
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6605040887948552e-05,
+      "loss": 0.3951,
+      "step": 10109
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6604705052642184e-05,
+      "loss": 0.2304,
+      "step": 10110
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6604369217335822e-05,
+      "loss": 0.3004,
+      "step": 10111
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6604033382029453e-05,
+      "loss": 0.3153,
+      "step": 10112
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6603697546723088e-05,
+      "loss": 0.2888,
+      "step": 10113
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6603361711416722e-05,
+      "loss": 0.4718,
+      "step": 10114
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6603025876110357e-05,
+      "loss": 0.3228,
+      "step": 10115
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.660269004080399e-05,
+      "loss": 0.3692,
+      "step": 10116
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6602354205497626e-05,
+      "loss": 0.2862,
+      "step": 10117
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.660201837019126e-05,
+      "loss": 0.2519,
+      "step": 10118
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6601682534884892e-05,
+      "loss": 0.3768,
+      "step": 10119
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.660134669957853e-05,
+      "loss": 0.3245,
+      "step": 10120
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.660101086427216e-05,
+      "loss": 0.4527,
+      "step": 10121
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6600675028965796e-05,
+      "loss": 0.2391,
+      "step": 10122
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.660033919365943e-05,
+      "loss": 0.4196,
+      "step": 10123
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6600003358353065e-05,
+      "loss": 0.3392,
+      "step": 10124
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.65996675230467e-05,
+      "loss": 0.3597,
+      "step": 10125
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6599331687740335e-05,
+      "loss": 0.1505,
+      "step": 10126
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.659899585243397e-05,
+      "loss": 0.2882,
+      "step": 10127
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.65986600171276e-05,
+      "loss": 0.2321,
+      "step": 10128
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.659832418182124e-05,
+      "loss": 0.2287,
+      "step": 10129
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.659798834651487e-05,
+      "loss": 0.3957,
+      "step": 10130
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6597652511208504e-05,
+      "loss": 0.1907,
+      "step": 10131
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.659731667590214e-05,
+      "loss": 0.1951,
+      "step": 10132
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6596980840595774e-05,
+      "loss": 0.2712,
+      "step": 10133
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.659664500528941e-05,
+      "loss": 0.4004,
+      "step": 10134
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6596309169983043e-05,
+      "loss": 0.3958,
+      "step": 10135
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6595973334676678e-05,
+      "loss": 0.3455,
+      "step": 10136
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.659563749937031e-05,
+      "loss": 0.2201,
+      "step": 10137
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6595301664063944e-05,
+      "loss": 0.3068,
+      "step": 10138
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6594965828757578e-05,
+      "loss": 0.2565,
+      "step": 10139
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6594629993451213e-05,
+      "loss": 0.374,
+      "step": 10140
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6594294158144848e-05,
+      "loss": 0.3844,
+      "step": 10141
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6593958322838482e-05,
+      "loss": 0.4879,
+      "step": 10142
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6593622487532113e-05,
+      "loss": 0.1792,
+      "step": 10143
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.659328665222575e-05,
+      "loss": 0.5459,
+      "step": 10144
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6592950816919383e-05,
+      "loss": 0.2645,
+      "step": 10145
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6592614981613017e-05,
+      "loss": 0.2803,
+      "step": 10146
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6592279146306652e-05,
+      "loss": 0.4421,
+      "step": 10147
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6591943311000287e-05,
+      "loss": 0.3001,
+      "step": 10148
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.659160747569392e-05,
+      "loss": 0.1765,
+      "step": 10149
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6591271640387556e-05,
+      "loss": 0.3775,
+      "step": 10150
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.659093580508119e-05,
+      "loss": 0.4042,
+      "step": 10151
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6590599969774822e-05,
+      "loss": 0.4225,
+      "step": 10152
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.659026413446846e-05,
+      "loss": 0.3066,
+      "step": 10153
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.658992829916209e-05,
+      "loss": 0.3579,
+      "step": 10154
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6589592463855726e-05,
+      "loss": 0.4412,
+      "step": 10155
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.658925662854936e-05,
+      "loss": 0.4418,
+      "step": 10156
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6588920793242995e-05,
+      "loss": 0.3349,
+      "step": 10157
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.658858495793663e-05,
+      "loss": 0.4023,
+      "step": 10158
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6588249122630264e-05,
+      "loss": 0.3387,
+      "step": 10159
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.65879132873239e-05,
+      "loss": 0.3028,
+      "step": 10160
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.658757745201753e-05,
+      "loss": 0.282,
+      "step": 10161
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.658724161671117e-05,
+      "loss": 0.2629,
+      "step": 10162
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.65869057814048e-05,
+      "loss": 0.2522,
+      "step": 10163
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6586569946098434e-05,
+      "loss": 0.48,
+      "step": 10164
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.658623411079207e-05,
+      "loss": 0.3665,
+      "step": 10165
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6585898275485704e-05,
+      "loss": 0.4776,
+      "step": 10166
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6585562440179338e-05,
+      "loss": 0.3617,
+      "step": 10167
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6585226604872973e-05,
+      "loss": 0.1812,
+      "step": 10168
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6584890769566608e-05,
+      "loss": 0.4397,
+      "step": 10169
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.658455493426024e-05,
+      "loss": 0.2641,
+      "step": 10170
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6584219098953877e-05,
+      "loss": 0.1879,
+      "step": 10171
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6583883263647508e-05,
+      "loss": 0.2333,
+      "step": 10172
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6583547428341143e-05,
+      "loss": 0.2734,
+      "step": 10173
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6583211593034777e-05,
+      "loss": 0.3386,
+      "step": 10174
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6582875757728412e-05,
+      "loss": 0.3237,
+      "step": 10175
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6582539922422047e-05,
+      "loss": 0.2322,
+      "step": 10176
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.658220408711568e-05,
+      "loss": 0.2244,
+      "step": 10177
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6581868251809316e-05,
+      "loss": 0.3777,
+      "step": 10178
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6581532416502947e-05,
+      "loss": 0.3282,
+      "step": 10179
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6581196581196585e-05,
+      "loss": 0.329,
+      "step": 10180
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6580860745890217e-05,
+      "loss": 0.3543,
+      "step": 10181
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.658052491058385e-05,
+      "loss": 0.4077,
+      "step": 10182
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6580189075277486e-05,
+      "loss": 0.3999,
+      "step": 10183
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.657985323997112e-05,
+      "loss": 0.2352,
+      "step": 10184
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6579517404664752e-05,
+      "loss": 0.3055,
+      "step": 10185
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.657918156935839e-05,
+      "loss": 0.2712,
+      "step": 10186
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.657884573405202e-05,
+      "loss": 0.2572,
+      "step": 10187
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6578509898745656e-05,
+      "loss": 0.3519,
+      "step": 10188
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.657817406343929e-05,
+      "loss": 0.2639,
+      "step": 10189
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6577838228132925e-05,
+      "loss": 0.2602,
+      "step": 10190
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.657750239282656e-05,
+      "loss": 0.346,
+      "step": 10191
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6577166557520194e-05,
+      "loss": 0.345,
+      "step": 10192
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.657683072221383e-05,
+      "loss": 0.4714,
+      "step": 10193
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.657649488690746e-05,
+      "loss": 0.2124,
+      "step": 10194
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6576159051601098e-05,
+      "loss": 0.5478,
+      "step": 10195
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.657582321629473e-05,
+      "loss": 0.3221,
+      "step": 10196
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6575487380988364e-05,
+      "loss": 0.3413,
+      "step": 10197
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6575151545682e-05,
+      "loss": 0.3166,
+      "step": 10198
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6574815710375633e-05,
+      "loss": 0.3502,
+      "step": 10199
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6574479875069268e-05,
+      "loss": 0.2835,
+      "step": 10200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.65741440397629e-05,
+      "loss": 0.4215,
+      "step": 10201
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6573808204456537e-05,
+      "loss": 0.3416,
+      "step": 10202
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.657347236915017e-05,
+      "loss": 0.4825,
+      "step": 10203
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6573136533843803e-05,
+      "loss": 0.4276,
+      "step": 10204
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6572800698537438e-05,
+      "loss": 0.2661,
+      "step": 10205
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6572464863231073e-05,
+      "loss": 0.3463,
+      "step": 10206
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6572129027924707e-05,
+      "loss": 0.4142,
+      "step": 10207
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6571793192618342e-05,
+      "loss": 0.3885,
+      "step": 10208
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6571457357311976e-05,
+      "loss": 0.4229,
+      "step": 10209
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6571121522005608e-05,
+      "loss": 0.5077,
+      "step": 10210
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6570785686699246e-05,
+      "loss": 0.3188,
+      "step": 10211
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6570449851392877e-05,
+      "loss": 0.3346,
+      "step": 10212
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6570114016086512e-05,
+      "loss": 0.4523,
+      "step": 10213
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6569778180780146e-05,
+      "loss": 0.4247,
+      "step": 10214
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.656944234547378e-05,
+      "loss": 0.2546,
+      "step": 10215
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6569106510167416e-05,
+      "loss": 0.4226,
+      "step": 10216
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.656877067486105e-05,
+      "loss": 0.208,
+      "step": 10217
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6568434839554685e-05,
+      "loss": 0.296,
+      "step": 10218
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6568099004248316e-05,
+      "loss": 0.2267,
+      "step": 10219
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6567763168941954e-05,
+      "loss": 0.4101,
+      "step": 10220
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6567427333635585e-05,
+      "loss": 0.4377,
+      "step": 10221
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.656709149832922e-05,
+      "loss": 0.3121,
+      "step": 10222
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.6566755663022855e-05,
+      "loss": 0.2455,
+      "step": 10223
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.656641982771649e-05,
+      "loss": 0.3883,
+      "step": 10224
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6566083992410124e-05,
+      "loss": 0.4676,
+      "step": 10225
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.656574815710376e-05,
+      "loss": 0.2593,
+      "step": 10226
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.656541232179739e-05,
+      "loss": 0.3233,
+      "step": 10227
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6565076486491025e-05,
+      "loss": 0.409,
+      "step": 10228
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.656474065118466e-05,
+      "loss": 0.1967,
+      "step": 10229
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6564404815878294e-05,
+      "loss": 0.29,
+      "step": 10230
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.656406898057193e-05,
+      "loss": 0.4317,
+      "step": 10231
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6563733145265563e-05,
+      "loss": 0.2808,
+      "step": 10232
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6563397309959198e-05,
+      "loss": 0.4764,
+      "step": 10233
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.656306147465283e-05,
+      "loss": 0.4147,
+      "step": 10234
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6562725639346467e-05,
+      "loss": 0.5418,
+      "step": 10235
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.65623898040401e-05,
+      "loss": 0.3537,
+      "step": 10236
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6562053968733733e-05,
+      "loss": 0.54,
+      "step": 10237
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6561718133427368e-05,
+      "loss": 0.2302,
+      "step": 10238
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6561382298121002e-05,
+      "loss": 0.3016,
+      "step": 10239
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6561046462814637e-05,
+      "loss": 0.2168,
+      "step": 10240
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.656071062750827e-05,
+      "loss": 0.3732,
+      "step": 10241
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6560374792201906e-05,
+      "loss": 0.2574,
+      "step": 10242
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6560038956895538e-05,
+      "loss": 0.2614,
+      "step": 10243
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6559703121589176e-05,
+      "loss": 0.273,
+      "step": 10244
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6559367286282807e-05,
+      "loss": 0.3533,
+      "step": 10245
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.655903145097644e-05,
+      "loss": 0.3502,
+      "step": 10246
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6558695615670076e-05,
+      "loss": 0.3485,
+      "step": 10247
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.655835978036371e-05,
+      "loss": 0.3467,
+      "step": 10248
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6558023945057345e-05,
+      "loss": 0.2422,
+      "step": 10249
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.655768810975098e-05,
+      "loss": 0.3402,
+      "step": 10250
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6557352274444615e-05,
+      "loss": 0.4336,
+      "step": 10251
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6557016439138246e-05,
+      "loss": 0.3509,
+      "step": 10252
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6556680603831884e-05,
+      "loss": 0.3836,
+      "step": 10253
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6556344768525515e-05,
+      "loss": 0.4336,
+      "step": 10254
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.655600893321915e-05,
+      "loss": 0.2952,
+      "step": 10255
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6555673097912785e-05,
+      "loss": 0.3642,
+      "step": 10256
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.655533726260642e-05,
+      "loss": 0.309,
+      "step": 10257
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6555001427300054e-05,
+      "loss": 0.3929,
+      "step": 10258
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.655466559199369e-05,
+      "loss": 0.1948,
+      "step": 10259
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6554329756687323e-05,
+      "loss": 0.2866,
+      "step": 10260
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6553993921380954e-05,
+      "loss": 0.3003,
+      "step": 10261
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6553658086074592e-05,
+      "loss": 0.181,
+      "step": 10262
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6553322250768224e-05,
+      "loss": 0.4655,
+      "step": 10263
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.655298641546186e-05,
+      "loss": 0.3904,
+      "step": 10264
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6552650580155493e-05,
+      "loss": 0.3006,
+      "step": 10265
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6552314744849128e-05,
+      "loss": 0.2068,
+      "step": 10266
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6551978909542762e-05,
+      "loss": 0.2775,
+      "step": 10267
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6551643074236397e-05,
+      "loss": 0.2502,
+      "step": 10268
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.655130723893003e-05,
+      "loss": 0.3371,
+      "step": 10269
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6550971403623663e-05,
+      "loss": 0.2906,
+      "step": 10270
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.65506355683173e-05,
+      "loss": 0.3955,
+      "step": 10271
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6550299733010932e-05,
+      "loss": 0.3037,
+      "step": 10272
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6549963897704567e-05,
+      "loss": 0.3492,
+      "step": 10273
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.65496280623982e-05,
+      "loss": 0.2751,
+      "step": 10274
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6549292227091836e-05,
+      "loss": 0.3344,
+      "step": 10275
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6548956391785467e-05,
+      "loss": 0.1007,
+      "step": 10276
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6548620556479105e-05,
+      "loss": 0.3728,
+      "step": 10277
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6548284721172737e-05,
+      "loss": 0.5676,
+      "step": 10278
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.654794888586637e-05,
+      "loss": 0.2287,
+      "step": 10279
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6547613050560006e-05,
+      "loss": 0.3785,
+      "step": 10280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.654727721525364e-05,
+      "loss": 0.3613,
+      "step": 10281
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6546941379947275e-05,
+      "loss": 0.2279,
+      "step": 10282
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.654660554464091e-05,
+      "loss": 0.2824,
+      "step": 10283
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6546269709334545e-05,
+      "loss": 0.2499,
+      "step": 10284
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6545933874028176e-05,
+      "loss": 0.352,
+      "step": 10285
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6545598038721814e-05,
+      "loss": 0.2285,
+      "step": 10286
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6545262203415445e-05,
+      "loss": 0.3395,
+      "step": 10287
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.654492636810908e-05,
+      "loss": 0.2853,
+      "step": 10288
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6544590532802714e-05,
+      "loss": 0.3156,
+      "step": 10289
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.654425469749635e-05,
+      "loss": 0.3238,
+      "step": 10290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6543918862189984e-05,
+      "loss": 0.3399,
+      "step": 10291
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.654358302688362e-05,
+      "loss": 0.2573,
+      "step": 10292
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6543247191577253e-05,
+      "loss": 0.2774,
+      "step": 10293
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6542911356270884e-05,
+      "loss": 0.5296,
+      "step": 10294
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6542575520964522e-05,
+      "loss": 0.2697,
+      "step": 10295
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6542239685658154e-05,
+      "loss": 0.2515,
+      "step": 10296
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6541903850351788e-05,
+      "loss": 0.2156,
+      "step": 10297
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6541568015045423e-05,
+      "loss": 0.3655,
+      "step": 10298
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6541232179739057e-05,
+      "loss": 0.2642,
+      "step": 10299
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6540896344432692e-05,
+      "loss": 0.393,
+      "step": 10300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6540560509126327e-05,
+      "loss": 0.2046,
+      "step": 10301
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.654022467381996e-05,
+      "loss": 0.2595,
+      "step": 10302
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6539888838513593e-05,
+      "loss": 0.2111,
+      "step": 10303
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.653955300320723e-05,
+      "loss": 0.3424,
+      "step": 10304
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6539217167900862e-05,
+      "loss": 0.3645,
+      "step": 10305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6538881332594497e-05,
+      "loss": 0.3993,
+      "step": 10306
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.653854549728813e-05,
+      "loss": 0.2407,
+      "step": 10307
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6538209661981766e-05,
+      "loss": 0.3116,
+      "step": 10308
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.65378738266754e-05,
+      "loss": 0.3643,
+      "step": 10309
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6537537991369035e-05,
+      "loss": 0.3557,
+      "step": 10310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.653720215606267e-05,
+      "loss": 0.2443,
+      "step": 10311
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.65368663207563e-05,
+      "loss": 0.2509,
+      "step": 10312
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.653653048544994e-05,
+      "loss": 0.3611,
+      "step": 10313
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.653619465014357e-05,
+      "loss": 0.369,
+      "step": 10314
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6535858814837205e-05,
+      "loss": 0.433,
+      "step": 10315
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.653552297953084e-05,
+      "loss": 0.2689,
+      "step": 10316
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6535187144224474e-05,
+      "loss": 0.3811,
+      "step": 10317
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6534851308918106e-05,
+      "loss": 0.4147,
+      "step": 10318
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6534515473611744e-05,
+      "loss": 0.3804,
+      "step": 10319
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6534179638305375e-05,
+      "loss": 0.3014,
+      "step": 10320
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.653384380299901e-05,
+      "loss": 0.3543,
+      "step": 10321
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6533507967692644e-05,
+      "loss": 0.2664,
+      "step": 10322
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.653317213238628e-05,
+      "loss": 0.4606,
+      "step": 10323
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6532836297079914e-05,
+      "loss": 0.2363,
+      "step": 10324
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6532500461773548e-05,
+      "loss": 0.449,
+      "step": 10325
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6532164626467183e-05,
+      "loss": 0.3275,
+      "step": 10326
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6531828791160814e-05,
+      "loss": 0.4471,
+      "step": 10327
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6531492955854452e-05,
+      "loss": 0.2913,
+      "step": 10328
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6531157120548083e-05,
+      "loss": 0.2672,
+      "step": 10329
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6530821285241718e-05,
+      "loss": 0.2784,
+      "step": 10330
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6530485449935353e-05,
+      "loss": 0.2665,
+      "step": 10331
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6530149614628987e-05,
+      "loss": 0.2313,
+      "step": 10332
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6529813779322622e-05,
+      "loss": 0.3859,
+      "step": 10333
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6529477944016257e-05,
+      "loss": 0.329,
+      "step": 10334
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.652914210870989e-05,
+      "loss": 0.359,
+      "step": 10335
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6528806273403523e-05,
+      "loss": 0.3796,
+      "step": 10336
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.652847043809716e-05,
+      "loss": 0.2805,
+      "step": 10337
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6528134602790792e-05,
+      "loss": 0.4514,
+      "step": 10338
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6527798767484426e-05,
+      "loss": 0.435,
+      "step": 10339
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.652746293217806e-05,
+      "loss": 0.3287,
+      "step": 10340
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6527127096871696e-05,
+      "loss": 0.4171,
+      "step": 10341
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.652679126156533e-05,
+      "loss": 0.569,
+      "step": 10342
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6526455426258965e-05,
+      "loss": 0.4063,
+      "step": 10343
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.65261195909526e-05,
+      "loss": 0.3083,
+      "step": 10344
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.652578375564623e-05,
+      "loss": 0.2879,
+      "step": 10345
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.652544792033987e-05,
+      "loss": 0.2511,
+      "step": 10346
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.65251120850335e-05,
+      "loss": 0.2837,
+      "step": 10347
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6524776249727135e-05,
+      "loss": 0.1877,
+      "step": 10348
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.652444041442077e-05,
+      "loss": 0.2329,
+      "step": 10349
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6524104579114404e-05,
+      "loss": 0.3597,
+      "step": 10350
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.652376874380804e-05,
+      "loss": 0.338,
+      "step": 10351
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6523432908501673e-05,
+      "loss": 0.3712,
+      "step": 10352
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6523097073195308e-05,
+      "loss": 0.3392,
+      "step": 10353
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.652276123788894e-05,
+      "loss": 0.3747,
+      "step": 10354
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6522425402582577e-05,
+      "loss": 0.2133,
+      "step": 10355
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.652208956727621e-05,
+      "loss": 0.2187,
+      "step": 10356
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6521753731969843e-05,
+      "loss": 0.2567,
+      "step": 10357
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6521417896663478e-05,
+      "loss": 0.2977,
+      "step": 10358
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6521082061357113e-05,
+      "loss": 0.3972,
+      "step": 10359
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6520746226050747e-05,
+      "loss": 0.2192,
+      "step": 10360
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6520410390744382e-05,
+      "loss": 0.2385,
+      "step": 10361
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6520074555438013e-05,
+      "loss": 0.478,
+      "step": 10362
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6519738720131648e-05,
+      "loss": 0.2488,
+      "step": 10363
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6519402884825282e-05,
+      "loss": 0.3171,
+      "step": 10364
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6519067049518917e-05,
+      "loss": 0.2454,
+      "step": 10365
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6518731214212552e-05,
+      "loss": 0.4713,
+      "step": 10366
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6518395378906186e-05,
+      "loss": 0.358,
+      "step": 10367
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.651805954359982e-05,
+      "loss": 0.2433,
+      "step": 10368
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6517723708293452e-05,
+      "loss": 0.2469,
+      "step": 10369
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.651738787298709e-05,
+      "loss": 0.3061,
+      "step": 10370
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.651705203768072e-05,
+      "loss": 0.2674,
+      "step": 10371
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6516716202374356e-05,
+      "loss": 0.1893,
+      "step": 10372
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.651638036706799e-05,
+      "loss": 0.4747,
+      "step": 10373
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6516044531761626e-05,
+      "loss": 0.4909,
+      "step": 10374
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.651570869645526e-05,
+      "loss": 0.4187,
+      "step": 10375
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6515372861148895e-05,
+      "loss": 0.402,
+      "step": 10376
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.651503702584253e-05,
+      "loss": 0.2562,
+      "step": 10377
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.651470119053616e-05,
+      "loss": 0.3235,
+      "step": 10378
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6514365355229795e-05,
+      "loss": 0.3886,
+      "step": 10379
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.651402951992343e-05,
+      "loss": 0.2786,
+      "step": 10380
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6513693684617065e-05,
+      "loss": 0.2126,
+      "step": 10381
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.65133578493107e-05,
+      "loss": 0.3038,
+      "step": 10382
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6513022014004334e-05,
+      "loss": 0.271,
+      "step": 10383
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.651268617869797e-05,
+      "loss": 0.1951,
+      "step": 10384
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.65123503433916e-05,
+      "loss": 0.1791,
+      "step": 10385
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6512014508085238e-05,
+      "loss": 0.3226,
+      "step": 10386
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.651167867277887e-05,
+      "loss": 0.2706,
+      "step": 10387
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6511342837472504e-05,
+      "loss": 0.1773,
+      "step": 10388
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.651100700216614e-05,
+      "loss": 0.3659,
+      "step": 10389
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6510671166859773e-05,
+      "loss": 0.4304,
+      "step": 10390
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6510335331553408e-05,
+      "loss": 0.4133,
+      "step": 10391
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6509999496247042e-05,
+      "loss": 0.2072,
+      "step": 10392
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6509663660940677e-05,
+      "loss": 0.3696,
+      "step": 10393
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.650932782563431e-05,
+      "loss": 0.2069,
+      "step": 10394
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6508991990327946e-05,
+      "loss": 0.1567,
+      "step": 10395
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6508656155021578e-05,
+      "loss": 0.1705,
+      "step": 10396
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6508320319715212e-05,
+      "loss": 0.2143,
+      "step": 10397
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6507984484408847e-05,
+      "loss": 0.2906,
+      "step": 10398
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.650764864910248e-05,
+      "loss": 0.2374,
+      "step": 10399
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6507312813796116e-05,
+      "loss": 0.4557,
+      "step": 10400
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.650697697848975e-05,
+      "loss": 0.3826,
+      "step": 10401
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6506641143183386e-05,
+      "loss": 0.2725,
+      "step": 10402
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6506305307877017e-05,
+      "loss": 0.2466,
+      "step": 10403
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6505969472570655e-05,
+      "loss": 0.4656,
+      "step": 10404
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6505633637264286e-05,
+      "loss": 0.4444,
+      "step": 10405
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.650529780195792e-05,
+      "loss": 0.386,
+      "step": 10406
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6504961966651555e-05,
+      "loss": 0.3043,
+      "step": 10407
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.650462613134519e-05,
+      "loss": 0.482,
+      "step": 10408
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.650429029603882e-05,
+      "loss": 0.2691,
+      "step": 10409
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.650395446073246e-05,
+      "loss": 0.1906,
+      "step": 10410
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.650361862542609e-05,
+      "loss": 0.351,
+      "step": 10411
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6503282790119725e-05,
+      "loss": 0.2606,
+      "step": 10412
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.650294695481336e-05,
+      "loss": 0.1505,
+      "step": 10413
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6502611119506995e-05,
+      "loss": 0.3178,
+      "step": 10414
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.650227528420063e-05,
+      "loss": 0.2748,
+      "step": 10415
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6501939448894264e-05,
+      "loss": 0.3232,
+      "step": 10416
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.65016036135879e-05,
+      "loss": 0.4375,
+      "step": 10417
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.650126777828153e-05,
+      "loss": 0.5554,
+      "step": 10418
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6500931942975168e-05,
+      "loss": 0.4948,
+      "step": 10419
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.65005961076688e-05,
+      "loss": 0.3098,
+      "step": 10420
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.6500260272362434e-05,
+      "loss": 0.3533,
+      "step": 10421
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6499924437056068e-05,
+      "loss": 0.1979,
+      "step": 10422
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6499588601749703e-05,
+      "loss": 0.1764,
+      "step": 10423
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6499252766443338e-05,
+      "loss": 0.3089,
+      "step": 10424
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6498916931136972e-05,
+      "loss": 0.3527,
+      "step": 10425
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6498581095830607e-05,
+      "loss": 0.3211,
+      "step": 10426
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6498245260524238e-05,
+      "loss": 0.4085,
+      "step": 10427
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6497909425217876e-05,
+      "loss": 0.3902,
+      "step": 10428
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6497573589911507e-05,
+      "loss": 0.4782,
+      "step": 10429
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6497237754605142e-05,
+      "loss": 0.2135,
+      "step": 10430
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6496901919298777e-05,
+      "loss": 0.3759,
+      "step": 10431
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.649656608399241e-05,
+      "loss": 0.3246,
+      "step": 10432
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6496230248686046e-05,
+      "loss": 0.2698,
+      "step": 10433
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.649589441337968e-05,
+      "loss": 0.6142,
+      "step": 10434
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6495558578073315e-05,
+      "loss": 0.1986,
+      "step": 10435
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6495222742766947e-05,
+      "loss": 0.3307,
+      "step": 10436
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6494886907460585e-05,
+      "loss": 0.3762,
+      "step": 10437
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6494551072154216e-05,
+      "loss": 0.4382,
+      "step": 10438
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.649421523684785e-05,
+      "loss": 0.2586,
+      "step": 10439
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6493879401541485e-05,
+      "loss": 0.3249,
+      "step": 10440
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.649354356623512e-05,
+      "loss": 0.2695,
+      "step": 10441
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6493207730928754e-05,
+      "loss": 0.3333,
+      "step": 10442
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.649287189562239e-05,
+      "loss": 0.3522,
+      "step": 10443
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6492536060316024e-05,
+      "loss": 0.2231,
+      "step": 10444
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6492200225009655e-05,
+      "loss": 0.2924,
+      "step": 10445
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6491864389703293e-05,
+      "loss": 0.1942,
+      "step": 10446
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6491528554396924e-05,
+      "loss": 0.2071,
+      "step": 10447
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.649119271909056e-05,
+      "loss": 0.3099,
+      "step": 10448
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6490856883784194e-05,
+      "loss": 0.3979,
+      "step": 10449
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6490521048477828e-05,
+      "loss": 0.5141,
+      "step": 10450
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6490185213171463e-05,
+      "loss": 0.3243,
+      "step": 10451
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6489849377865098e-05,
+      "loss": 0.3002,
+      "step": 10452
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.648951354255873e-05,
+      "loss": 0.3893,
+      "step": 10453
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6489177707252363e-05,
+      "loss": 0.2974,
+      "step": 10454
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6488841871945998e-05,
+      "loss": 0.3509,
+      "step": 10455
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6488506036639633e-05,
+      "loss": 0.3151,
+      "step": 10456
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6488170201333267e-05,
+      "loss": 0.4984,
+      "step": 10457
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6487834366026902e-05,
+      "loss": 0.3398,
+      "step": 10458
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6487498530720537e-05,
+      "loss": 0.3068,
+      "step": 10459
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6487162695414168e-05,
+      "loss": 0.3977,
+      "step": 10460
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6486826860107806e-05,
+      "loss": 0.2537,
+      "step": 10461
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6486491024801437e-05,
+      "loss": 0.4208,
+      "step": 10462
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6486155189495072e-05,
+      "loss": 0.4536,
+      "step": 10463
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6485819354188707e-05,
+      "loss": 0.2244,
+      "step": 10464
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.648548351888234e-05,
+      "loss": 0.4574,
+      "step": 10465
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6485147683575976e-05,
+      "loss": 0.3771,
+      "step": 10466
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.648481184826961e-05,
+      "loss": 0.3137,
+      "step": 10467
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6484476012963245e-05,
+      "loss": 0.4392,
+      "step": 10468
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6484140177656876e-05,
+      "loss": 0.2362,
+      "step": 10469
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6483804342350514e-05,
+      "loss": 0.3392,
+      "step": 10470
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6483468507044146e-05,
+      "loss": 0.446,
+      "step": 10471
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.648313267173778e-05,
+      "loss": 0.226,
+      "step": 10472
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6482796836431415e-05,
+      "loss": 0.4455,
+      "step": 10473
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.648246100112505e-05,
+      "loss": 0.1845,
+      "step": 10474
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6482125165818684e-05,
+      "loss": 0.3117,
+      "step": 10475
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.648178933051232e-05,
+      "loss": 0.2942,
+      "step": 10476
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6481453495205954e-05,
+      "loss": 0.4232,
+      "step": 10477
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6481117659899585e-05,
+      "loss": 0.1293,
+      "step": 10478
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6480781824593223e-05,
+      "loss": 0.2622,
+      "step": 10479
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6480445989286854e-05,
+      "loss": 0.5207,
+      "step": 10480
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.648011015398049e-05,
+      "loss": 0.3487,
+      "step": 10481
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6479774318674123e-05,
+      "loss": 0.3825,
+      "step": 10482
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6479438483367758e-05,
+      "loss": 0.293,
+      "step": 10483
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6479102648061393e-05,
+      "loss": 0.2427,
+      "step": 10484
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6478766812755027e-05,
+      "loss": 0.2267,
+      "step": 10485
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6478430977448662e-05,
+      "loss": 0.2935,
+      "step": 10486
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6478095142142293e-05,
+      "loss": 0.2429,
+      "step": 10487
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.647775930683593e-05,
+      "loss": 0.4028,
+      "step": 10488
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6477423471529563e-05,
+      "loss": 0.2445,
+      "step": 10489
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6477087636223197e-05,
+      "loss": 0.403,
+      "step": 10490
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6476751800916832e-05,
+      "loss": 0.3478,
+      "step": 10491
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6476415965610467e-05,
+      "loss": 0.3151,
+      "step": 10492
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.64760801303041e-05,
+      "loss": 0.3428,
+      "step": 10493
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6475744294997736e-05,
+      "loss": 0.2789,
+      "step": 10494
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.647540845969137e-05,
+      "loss": 0.4047,
+      "step": 10495
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6475072624385002e-05,
+      "loss": 0.3978,
+      "step": 10496
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6474736789078636e-05,
+      "loss": 0.3325,
+      "step": 10497
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.647440095377227e-05,
+      "loss": 0.4279,
+      "step": 10498
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6474065118465906e-05,
+      "loss": 0.1944,
+      "step": 10499
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.647372928315954e-05,
+      "loss": 0.2772,
+      "step": 10500
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6473393447853175e-05,
+      "loss": 0.3544,
+      "step": 10501
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6473057612546806e-05,
+      "loss": 0.3644,
+      "step": 10502
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6472721777240444e-05,
+      "loss": 0.3107,
+      "step": 10503
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6472385941934076e-05,
+      "loss": 0.2904,
+      "step": 10504
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.647205010662771e-05,
+      "loss": 0.2847,
+      "step": 10505
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6471714271321345e-05,
+      "loss": 0.3974,
+      "step": 10506
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.647137843601498e-05,
+      "loss": 0.4427,
+      "step": 10507
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6471042600708614e-05,
+      "loss": 0.246,
+      "step": 10508
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.647070676540225e-05,
+      "loss": 0.2154,
+      "step": 10509
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6470370930095883e-05,
+      "loss": 0.3352,
+      "step": 10510
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6470035094789515e-05,
+      "loss": 0.4464,
+      "step": 10511
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6469699259483153e-05,
+      "loss": 0.2595,
+      "step": 10512
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6469363424176784e-05,
+      "loss": 0.3467,
+      "step": 10513
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.646902758887042e-05,
+      "loss": 0.2017,
+      "step": 10514
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6468691753564053e-05,
+      "loss": 0.2909,
+      "step": 10515
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6468355918257688e-05,
+      "loss": 0.2475,
+      "step": 10516
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6468020082951323e-05,
+      "loss": 0.2826,
+      "step": 10517
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6467684247644957e-05,
+      "loss": 0.2733,
+      "step": 10518
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6467348412338592e-05,
+      "loss": 0.3695,
+      "step": 10519
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6467012577032223e-05,
+      "loss": 0.181,
+      "step": 10520
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.646667674172586e-05,
+      "loss": 0.2766,
+      "step": 10521
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6466340906419492e-05,
+      "loss": 0.4111,
+      "step": 10522
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6466005071113127e-05,
+      "loss": 0.3053,
+      "step": 10523
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.646566923580676e-05,
+      "loss": 0.2271,
+      "step": 10524
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6465333400500396e-05,
+      "loss": 0.318,
+      "step": 10525
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.646499756519403e-05,
+      "loss": 0.1911,
+      "step": 10526
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6464661729887666e-05,
+      "loss": 0.4407,
+      "step": 10527
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.64643258945813e-05,
+      "loss": 0.351,
+      "step": 10528
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.646399005927493e-05,
+      "loss": 0.3973,
+      "step": 10529
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.646365422396857e-05,
+      "loss": 0.3159,
+      "step": 10530
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.64633183886622e-05,
+      "loss": 0.362,
+      "step": 10531
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6462982553355835e-05,
+      "loss": 0.3077,
+      "step": 10532
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.646264671804947e-05,
+      "loss": 0.3416,
+      "step": 10533
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6462310882743105e-05,
+      "loss": 0.3052,
+      "step": 10534
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.646197504743674e-05,
+      "loss": 0.3008,
+      "step": 10535
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6461639212130374e-05,
+      "loss": 0.3077,
+      "step": 10536
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.646130337682401e-05,
+      "loss": 0.3943,
+      "step": 10537
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.646096754151764e-05,
+      "loss": 0.3167,
+      "step": 10538
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6460631706211278e-05,
+      "loss": 0.1794,
+      "step": 10539
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.646029587090491e-05,
+      "loss": 0.4497,
+      "step": 10540
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6459960035598544e-05,
+      "loss": 0.2969,
+      "step": 10541
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.645962420029218e-05,
+      "loss": 0.441,
+      "step": 10542
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6459288364985813e-05,
+      "loss": 0.3386,
+      "step": 10543
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6458952529679444e-05,
+      "loss": 0.3758,
+      "step": 10544
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6458616694373082e-05,
+      "loss": 0.2574,
+      "step": 10545
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6458280859066714e-05,
+      "loss": 0.5141,
+      "step": 10546
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.645794502376035e-05,
+      "loss": 0.3053,
+      "step": 10547
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6457609188453983e-05,
+      "loss": 0.477,
+      "step": 10548
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6457273353147618e-05,
+      "loss": 0.2737,
+      "step": 10549
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6456937517841252e-05,
+      "loss": 0.2412,
+      "step": 10550
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6456601682534887e-05,
+      "loss": 0.4105,
+      "step": 10551
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.645626584722852e-05,
+      "loss": 0.2578,
+      "step": 10552
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6455930011922153e-05,
+      "loss": 0.3844,
+      "step": 10553
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.645559417661579e-05,
+      "loss": 0.4578,
+      "step": 10554
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6455258341309422e-05,
+      "loss": 0.3311,
+      "step": 10555
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6454922506003057e-05,
+      "loss": 0.4468,
+      "step": 10556
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.645458667069669e-05,
+      "loss": 0.3516,
+      "step": 10557
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6454250835390326e-05,
+      "loss": 0.3499,
+      "step": 10558
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.645391500008396e-05,
+      "loss": 0.2234,
+      "step": 10559
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6453579164777592e-05,
+      "loss": 0.2716,
+      "step": 10560
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.645324332947123e-05,
+      "loss": 0.4174,
+      "step": 10561
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.645290749416486e-05,
+      "loss": 0.2928,
+      "step": 10562
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6452571658858496e-05,
+      "loss": 0.2677,
+      "step": 10563
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.645223582355213e-05,
+      "loss": 0.3499,
+      "step": 10564
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6451899988245765e-05,
+      "loss": 0.3394,
+      "step": 10565
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.64515641529394e-05,
+      "loss": 0.284,
+      "step": 10566
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6451228317633035e-05,
+      "loss": 0.4318,
+      "step": 10567
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.645089248232667e-05,
+      "loss": 0.2005,
+      "step": 10568
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.64505566470203e-05,
+      "loss": 0.3128,
+      "step": 10569
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.645022081171394e-05,
+      "loss": 0.2457,
+      "step": 10570
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.644988497640757e-05,
+      "loss": 0.4697,
+      "step": 10571
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6449549141101204e-05,
+      "loss": 0.4286,
+      "step": 10572
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.644921330579484e-05,
+      "loss": 0.4856,
+      "step": 10573
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6448877470488474e-05,
+      "loss": 0.2867,
+      "step": 10574
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.644854163518211e-05,
+      "loss": 0.3227,
+      "step": 10575
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6448205799875743e-05,
+      "loss": 0.2192,
+      "step": 10576
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6447869964569378e-05,
+      "loss": 0.3947,
+      "step": 10577
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.644753412926301e-05,
+      "loss": 0.3749,
+      "step": 10578
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6447198293956647e-05,
+      "loss": 0.2622,
+      "step": 10579
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6446862458650278e-05,
+      "loss": 0.3491,
+      "step": 10580
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6446526623343913e-05,
+      "loss": 0.3493,
+      "step": 10581
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6446190788037548e-05,
+      "loss": 0.2027,
+      "step": 10582
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6445854952731182e-05,
+      "loss": 0.3688,
+      "step": 10583
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6445519117424817e-05,
+      "loss": 0.2679,
+      "step": 10584
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.644518328211845e-05,
+      "loss": 0.2695,
+      "step": 10585
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6444847446812086e-05,
+      "loss": 0.4548,
+      "step": 10586
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6444511611505717e-05,
+      "loss": 0.211,
+      "step": 10587
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6444175776199352e-05,
+      "loss": 0.4351,
+      "step": 10588
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6443839940892987e-05,
+      "loss": 0.2233,
+      "step": 10589
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.644350410558662e-05,
+      "loss": 0.275,
+      "step": 10590
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6443168270280256e-05,
+      "loss": 0.3818,
+      "step": 10591
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.644283243497389e-05,
+      "loss": 0.3856,
+      "step": 10592
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6442496599667522e-05,
+      "loss": 0.3236,
+      "step": 10593
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.644216076436116e-05,
+      "loss": 0.4924,
+      "step": 10594
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.644182492905479e-05,
+      "loss": 0.4558,
+      "step": 10595
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6441489093748426e-05,
+      "loss": 0.3702,
+      "step": 10596
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.644115325844206e-05,
+      "loss": 0.2509,
+      "step": 10597
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6440817423135695e-05,
+      "loss": 0.3121,
+      "step": 10598
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.644048158782933e-05,
+      "loss": 0.3698,
+      "step": 10599
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6440145752522964e-05,
+      "loss": 0.4931,
+      "step": 10600
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.64398099172166e-05,
+      "loss": 0.4131,
+      "step": 10601
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.643947408191023e-05,
+      "loss": 0.2148,
+      "step": 10602
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.643913824660387e-05,
+      "loss": 0.3262,
+      "step": 10603
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.64388024112975e-05,
+      "loss": 0.2977,
+      "step": 10604
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6438466575991134e-05,
+      "loss": 0.4059,
+      "step": 10605
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.643813074068477e-05,
+      "loss": 0.2405,
+      "step": 10606
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6437794905378404e-05,
+      "loss": 0.3585,
+      "step": 10607
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6437459070072038e-05,
+      "loss": 0.3823,
+      "step": 10608
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6437123234765673e-05,
+      "loss": 0.2939,
+      "step": 10609
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6436787399459307e-05,
+      "loss": 0.3647,
+      "step": 10610
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.643645156415294e-05,
+      "loss": 0.2167,
+      "step": 10611
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6436115728846577e-05,
+      "loss": 0.339,
+      "step": 10612
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6435779893540208e-05,
+      "loss": 0.3361,
+      "step": 10613
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6435444058233843e-05,
+      "loss": 0.1968,
+      "step": 10614
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6435108222927477e-05,
+      "loss": 0.2815,
+      "step": 10615
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6434772387621112e-05,
+      "loss": 0.4593,
+      "step": 10616
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6434436552314747e-05,
+      "loss": 0.5204,
+      "step": 10617
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.643410071700838e-05,
+      "loss": 0.2775,
+      "step": 10618
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6433764881702016e-05,
+      "loss": 0.3092,
+      "step": 10619
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.6433429046395647e-05,
+      "loss": 0.3824,
+      "step": 10620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6433093211089285e-05,
+      "loss": 0.3632,
+      "step": 10621
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6432757375782916e-05,
+      "loss": 0.4299,
+      "step": 10622
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.643242154047655e-05,
+      "loss": 0.2901,
+      "step": 10623
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6432085705170186e-05,
+      "loss": 0.5072,
+      "step": 10624
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.643174986986382e-05,
+      "loss": 0.1774,
+      "step": 10625
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6431414034557455e-05,
+      "loss": 0.2633,
+      "step": 10626
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.643107819925109e-05,
+      "loss": 0.4035,
+      "step": 10627
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6430742363944724e-05,
+      "loss": 0.3531,
+      "step": 10628
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6430406528638356e-05,
+      "loss": 0.1799,
+      "step": 10629
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6430070693331994e-05,
+      "loss": 0.2161,
+      "step": 10630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6429734858025625e-05,
+      "loss": 0.3523,
+      "step": 10631
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.642939902271926e-05,
+      "loss": 0.2831,
+      "step": 10632
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6429063187412894e-05,
+      "loss": 0.5395,
+      "step": 10633
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.642872735210653e-05,
+      "loss": 0.2883,
+      "step": 10634
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.642839151680016e-05,
+      "loss": 0.2061,
+      "step": 10635
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6428055681493798e-05,
+      "loss": 0.4699,
+      "step": 10636
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.642771984618743e-05,
+      "loss": 0.2935,
+      "step": 10637
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6427384010881064e-05,
+      "loss": 0.3155,
+      "step": 10638
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.64270481755747e-05,
+      "loss": 0.52,
+      "step": 10639
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6426712340268333e-05,
+      "loss": 0.3752,
+      "step": 10640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6426376504961968e-05,
+      "loss": 0.4105,
+      "step": 10641
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6426040669655603e-05,
+      "loss": 0.2764,
+      "step": 10642
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6425704834349237e-05,
+      "loss": 0.3693,
+      "step": 10643
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.642536899904287e-05,
+      "loss": 0.3452,
+      "step": 10644
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6425033163736507e-05,
+      "loss": 0.2521,
+      "step": 10645
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6424697328430138e-05,
+      "loss": 0.4098,
+      "step": 10646
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6424361493123772e-05,
+      "loss": 0.4417,
+      "step": 10647
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6424025657817407e-05,
+      "loss": 0.46,
+      "step": 10648
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6423689822511042e-05,
+      "loss": 0.2751,
+      "step": 10649
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6423353987204676e-05,
+      "loss": 0.2015,
+      "step": 10650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.642301815189831e-05,
+      "loss": 0.3064,
+      "step": 10651
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6422682316591946e-05,
+      "loss": 0.4011,
+      "step": 10652
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6422346481285577e-05,
+      "loss": 0.4625,
+      "step": 10653
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6422010645979215e-05,
+      "loss": 0.228,
+      "step": 10654
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6421674810672846e-05,
+      "loss": 0.2661,
+      "step": 10655
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.642133897536648e-05,
+      "loss": 0.4519,
+      "step": 10656
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6421003140060116e-05,
+      "loss": 0.3387,
+      "step": 10657
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.642066730475375e-05,
+      "loss": 0.5095,
+      "step": 10658
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6420331469447385e-05,
+      "loss": 0.1581,
+      "step": 10659
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.641999563414102e-05,
+      "loss": 0.3165,
+      "step": 10660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6419659798834654e-05,
+      "loss": 0.2034,
+      "step": 10661
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6419323963528285e-05,
+      "loss": 0.2682,
+      "step": 10662
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6418988128221923e-05,
+      "loss": 0.3546,
+      "step": 10663
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6418652292915555e-05,
+      "loss": 0.2693,
+      "step": 10664
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.641831645760919e-05,
+      "loss": 0.1872,
+      "step": 10665
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6417980622302824e-05,
+      "loss": 0.3278,
+      "step": 10666
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.641764478699646e-05,
+      "loss": 0.5674,
+      "step": 10667
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6417308951690093e-05,
+      "loss": 0.2982,
+      "step": 10668
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6416973116383728e-05,
+      "loss": 0.233,
+      "step": 10669
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6416637281077363e-05,
+      "loss": 0.315,
+      "step": 10670
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6416301445770994e-05,
+      "loss": 0.1793,
+      "step": 10671
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6415965610464632e-05,
+      "loss": 0.4416,
+      "step": 10672
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6415629775158263e-05,
+      "loss": 0.3576,
+      "step": 10673
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6415293939851898e-05,
+      "loss": 0.4262,
+      "step": 10674
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6414958104545532e-05,
+      "loss": 0.2404,
+      "step": 10675
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6414622269239167e-05,
+      "loss": 0.4103,
+      "step": 10676
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6414286433932802e-05,
+      "loss": 0.2393,
+      "step": 10677
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6413950598626436e-05,
+      "loss": 0.2746,
+      "step": 10678
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6413614763320068e-05,
+      "loss": 0.293,
+      "step": 10679
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6413278928013702e-05,
+      "loss": 0.3707,
+      "step": 10680
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6412943092707337e-05,
+      "loss": 0.4318,
+      "step": 10681
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.641260725740097e-05,
+      "loss": 0.2887,
+      "step": 10682
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6412271422094606e-05,
+      "loss": 0.3365,
+      "step": 10683
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.641193558678824e-05,
+      "loss": 0.1087,
+      "step": 10684
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6411599751481876e-05,
+      "loss": 0.3565,
+      "step": 10685
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6411263916175507e-05,
+      "loss": 0.3451,
+      "step": 10686
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6410928080869145e-05,
+      "loss": 0.3148,
+      "step": 10687
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6410592245562776e-05,
+      "loss": 0.3217,
+      "step": 10688
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.641025641025641e-05,
+      "loss": 0.2977,
+      "step": 10689
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6409920574950045e-05,
+      "loss": 0.2442,
+      "step": 10690
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.640958473964368e-05,
+      "loss": 0.2736,
+      "step": 10691
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6409248904337315e-05,
+      "loss": 0.5107,
+      "step": 10692
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.640891306903095e-05,
+      "loss": 0.3759,
+      "step": 10693
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6408577233724584e-05,
+      "loss": 0.3028,
+      "step": 10694
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6408241398418215e-05,
+      "loss": 0.1922,
+      "step": 10695
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6407905563111853e-05,
+      "loss": 0.3175,
+      "step": 10696
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6407569727805485e-05,
+      "loss": 0.5374,
+      "step": 10697
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.640723389249912e-05,
+      "loss": 0.1751,
+      "step": 10698
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6406898057192754e-05,
+      "loss": 0.5394,
+      "step": 10699
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.640656222188639e-05,
+      "loss": 0.4111,
+      "step": 10700
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6406226386580023e-05,
+      "loss": 0.3812,
+      "step": 10701
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6405890551273658e-05,
+      "loss": 0.2793,
+      "step": 10702
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6405554715967292e-05,
+      "loss": 0.2138,
+      "step": 10703
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6405218880660924e-05,
+      "loss": 0.335,
+      "step": 10704
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6404883045354562e-05,
+      "loss": 0.4587,
+      "step": 10705
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6404547210048193e-05,
+      "loss": 0.3227,
+      "step": 10706
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6404211374741828e-05,
+      "loss": 0.3381,
+      "step": 10707
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6403875539435462e-05,
+      "loss": 0.2568,
+      "step": 10708
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6403539704129097e-05,
+      "loss": 0.3663,
+      "step": 10709
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.640320386882273e-05,
+      "loss": 0.2484,
+      "step": 10710
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6402868033516366e-05,
+      "loss": 0.449,
+      "step": 10711
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.640253219821e-05,
+      "loss": 0.313,
+      "step": 10712
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6402196362903632e-05,
+      "loss": 0.4177,
+      "step": 10713
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.640186052759727e-05,
+      "loss": 0.4083,
+      "step": 10714
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.64015246922909e-05,
+      "loss": 0.2923,
+      "step": 10715
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6401188856984536e-05,
+      "loss": 0.5169,
+      "step": 10716
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.640085302167817e-05,
+      "loss": 0.3433,
+      "step": 10717
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6400517186371805e-05,
+      "loss": 0.2695,
+      "step": 10718
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.640018135106544e-05,
+      "loss": 0.2786,
+      "step": 10719
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6399845515759075e-05,
+      "loss": 0.2363,
+      "step": 10720
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.639950968045271e-05,
+      "loss": 0.188,
+      "step": 10721
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.639917384514634e-05,
+      "loss": 0.2652,
+      "step": 10722
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6398838009839975e-05,
+      "loss": 0.4597,
+      "step": 10723
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.639850217453361e-05,
+      "loss": 0.2782,
+      "step": 10724
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6398166339227244e-05,
+      "loss": 0.3735,
+      "step": 10725
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.639783050392088e-05,
+      "loss": 0.2449,
+      "step": 10726
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6397494668614514e-05,
+      "loss": 0.1552,
+      "step": 10727
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6397158833308145e-05,
+      "loss": 0.2811,
+      "step": 10728
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6396822998001783e-05,
+      "loss": 0.2304,
+      "step": 10729
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6396487162695414e-05,
+      "loss": 0.409,
+      "step": 10730
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.639615132738905e-05,
+      "loss": 0.4131,
+      "step": 10731
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6395815492082684e-05,
+      "loss": 0.3179,
+      "step": 10732
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6395479656776318e-05,
+      "loss": 0.4212,
+      "step": 10733
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6395143821469953e-05,
+      "loss": 0.3132,
+      "step": 10734
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6394807986163588e-05,
+      "loss": 0.228,
+      "step": 10735
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6394472150857222e-05,
+      "loss": 0.318,
+      "step": 10736
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6394136315550853e-05,
+      "loss": 0.3141,
+      "step": 10737
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6393800480244488e-05,
+      "loss": 0.3579,
+      "step": 10738
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6393464644938123e-05,
+      "loss": 0.439,
+      "step": 10739
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6393128809631757e-05,
+      "loss": 0.3599,
+      "step": 10740
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6392792974325392e-05,
+      "loss": 0.4003,
+      "step": 10741
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6392457139019027e-05,
+      "loss": 0.3432,
+      "step": 10742
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.639212130371266e-05,
+      "loss": 0.383,
+      "step": 10743
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6391785468406293e-05,
+      "loss": 0.2365,
+      "step": 10744
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.639144963309993e-05,
+      "loss": 0.486,
+      "step": 10745
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6391113797793562e-05,
+      "loss": 0.3501,
+      "step": 10746
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6390777962487197e-05,
+      "loss": 0.3733,
+      "step": 10747
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.639044212718083e-05,
+      "loss": 0.3775,
+      "step": 10748
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6390106291874466e-05,
+      "loss": 0.3237,
+      "step": 10749
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.63897704565681e-05,
+      "loss": 0.2529,
+      "step": 10750
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6389434621261735e-05,
+      "loss": 0.242,
+      "step": 10751
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.638909878595537e-05,
+      "loss": 0.3205,
+      "step": 10752
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6388762950649e-05,
+      "loss": 0.2486,
+      "step": 10753
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.638842711534264e-05,
+      "loss": 0.2968,
+      "step": 10754
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.638809128003627e-05,
+      "loss": 0.3411,
+      "step": 10755
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6387755444729905e-05,
+      "loss": 0.4022,
+      "step": 10756
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.638741960942354e-05,
+      "loss": 0.3645,
+      "step": 10757
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6387083774117174e-05,
+      "loss": 0.3566,
+      "step": 10758
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.638674793881081e-05,
+      "loss": 0.3769,
+      "step": 10759
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6386412103504444e-05,
+      "loss": 0.3018,
+      "step": 10760
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6386076268198078e-05,
+      "loss": 0.3504,
+      "step": 10761
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.638574043289171e-05,
+      "loss": 0.3174,
+      "step": 10762
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6385404597585348e-05,
+      "loss": 0.2245,
+      "step": 10763
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.638506876227898e-05,
+      "loss": 0.4198,
+      "step": 10764
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6384732926972613e-05,
+      "loss": 0.3697,
+      "step": 10765
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6384397091666248e-05,
+      "loss": 0.2602,
+      "step": 10766
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6384061256359883e-05,
+      "loss": 0.3695,
+      "step": 10767
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6383725421053517e-05,
+      "loss": 0.2926,
+      "step": 10768
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6383389585747152e-05,
+      "loss": 0.5898,
+      "step": 10769
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6383053750440783e-05,
+      "loss": 0.3366,
+      "step": 10770
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6382717915134418e-05,
+      "loss": 0.5123,
+      "step": 10771
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6382382079828053e-05,
+      "loss": 0.3907,
+      "step": 10772
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6382046244521687e-05,
+      "loss": 0.4718,
+      "step": 10773
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6381710409215322e-05,
+      "loss": 0.2303,
+      "step": 10774
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6381374573908957e-05,
+      "loss": 0.3024,
+      "step": 10775
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.638103873860259e-05,
+      "loss": 0.3463,
+      "step": 10776
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6380702903296222e-05,
+      "loss": 0.2644,
+      "step": 10777
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.638036706798986e-05,
+      "loss": 0.3091,
+      "step": 10778
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6380031232683492e-05,
+      "loss": 0.69,
+      "step": 10779
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6379695397377126e-05,
+      "loss": 0.3927,
+      "step": 10780
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.637935956207076e-05,
+      "loss": 0.1758,
+      "step": 10781
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6379023726764396e-05,
+      "loss": 0.3805,
+      "step": 10782
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.637868789145803e-05,
+      "loss": 0.2705,
+      "step": 10783
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6378352056151665e-05,
+      "loss": 0.3167,
+      "step": 10784
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.63780162208453e-05,
+      "loss": 0.3696,
+      "step": 10785
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.637768038553893e-05,
+      "loss": 0.389,
+      "step": 10786
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.637734455023257e-05,
+      "loss": 0.3975,
+      "step": 10787
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.63770087149262e-05,
+      "loss": 0.2212,
+      "step": 10788
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6376672879619835e-05,
+      "loss": 0.2705,
+      "step": 10789
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.637633704431347e-05,
+      "loss": 0.2601,
+      "step": 10790
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6376001209007104e-05,
+      "loss": 0.2416,
+      "step": 10791
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.637566537370074e-05,
+      "loss": 0.3725,
+      "step": 10792
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6375329538394373e-05,
+      "loss": 0.1968,
+      "step": 10793
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6374993703088008e-05,
+      "loss": 0.3389,
+      "step": 10794
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.637465786778164e-05,
+      "loss": 0.4564,
+      "step": 10795
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6374322032475277e-05,
+      "loss": 0.2432,
+      "step": 10796
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.637398619716891e-05,
+      "loss": 0.4258,
+      "step": 10797
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6373650361862543e-05,
+      "loss": 0.5002,
+      "step": 10798
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6373314526556178e-05,
+      "loss": 0.1849,
+      "step": 10799
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6372978691249813e-05,
+      "loss": 0.3287,
+      "step": 10800
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6372642855943447e-05,
+      "loss": 0.4228,
+      "step": 10801
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6372307020637082e-05,
+      "loss": 0.1429,
+      "step": 10802
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6371971185330716e-05,
+      "loss": 0.252,
+      "step": 10803
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6371635350024348e-05,
+      "loss": 0.4076,
+      "step": 10804
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6371299514717986e-05,
+      "loss": 0.4567,
+      "step": 10805
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6370963679411617e-05,
+      "loss": 0.2622,
+      "step": 10806
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.637062784410525e-05,
+      "loss": 0.3412,
+      "step": 10807
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6370292008798886e-05,
+      "loss": 0.2355,
+      "step": 10808
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.636995617349252e-05,
+      "loss": 0.4719,
+      "step": 10809
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6369620338186156e-05,
+      "loss": 0.248,
+      "step": 10810
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.636928450287979e-05,
+      "loss": 0.2523,
+      "step": 10811
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6368948667573425e-05,
+      "loss": 0.4394,
+      "step": 10812
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6368612832267056e-05,
+      "loss": 0.3175,
+      "step": 10813
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.636827699696069e-05,
+      "loss": 0.3781,
+      "step": 10814
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6367941161654325e-05,
+      "loss": 0.3211,
+      "step": 10815
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.636760532634796e-05,
+      "loss": 0.1726,
+      "step": 10816
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6367269491041595e-05,
+      "loss": 0.2611,
+      "step": 10817
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.636693365573523e-05,
+      "loss": 0.2776,
+      "step": 10818
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.636659782042886e-05,
+      "loss": 0.5194,
+      "step": 10819
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.63662619851225e-05,
+      "loss": 0.4944,
+      "step": 10820
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.636592614981613e-05,
+      "loss": 0.2867,
+      "step": 10821
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6365590314509765e-05,
+      "loss": 0.1563,
+      "step": 10822
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.63652544792034e-05,
+      "loss": 0.2959,
+      "step": 10823
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6364918643897034e-05,
+      "loss": 0.2782,
+      "step": 10824
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.636458280859067e-05,
+      "loss": 0.4096,
+      "step": 10825
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6364246973284303e-05,
+      "loss": 0.4409,
+      "step": 10826
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6363911137977938e-05,
+      "loss": 0.5586,
+      "step": 10827
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.636357530267157e-05,
+      "loss": 0.3064,
+      "step": 10828
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6363239467365207e-05,
+      "loss": 0.3368,
+      "step": 10829
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.636290363205884e-05,
+      "loss": 0.2783,
+      "step": 10830
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6362567796752473e-05,
+      "loss": 0.3784,
+      "step": 10831
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6362231961446108e-05,
+      "loss": 0.4366,
+      "step": 10832
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6361896126139742e-05,
+      "loss": 0.4242,
+      "step": 10833
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6361560290833377e-05,
+      "loss": 0.3597,
+      "step": 10834
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.636122445552701e-05,
+      "loss": 0.3518,
+      "step": 10835
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6360888620220646e-05,
+      "loss": 0.287,
+      "step": 10836
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6360552784914278e-05,
+      "loss": 0.2414,
+      "step": 10837
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6360216949607916e-05,
+      "loss": 0.3631,
+      "step": 10838
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6359881114301547e-05,
+      "loss": 0.3407,
+      "step": 10839
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.635954527899518e-05,
+      "loss": 0.2794,
+      "step": 10840
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6359209443688816e-05,
+      "loss": 0.398,
+      "step": 10841
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.635887360838245e-05,
+      "loss": 0.4092,
+      "step": 10842
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6358537773076085e-05,
+      "loss": 0.3718,
+      "step": 10843
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.635820193776972e-05,
+      "loss": 0.2023,
+      "step": 10844
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6357866102463355e-05,
+      "loss": 0.2913,
+      "step": 10845
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6357530267156986e-05,
+      "loss": 0.453,
+      "step": 10846
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6357194431850624e-05,
+      "loss": 0.4671,
+      "step": 10847
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6356858596544255e-05,
+      "loss": 0.2354,
+      "step": 10848
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.635652276123789e-05,
+      "loss": 0.3714,
+      "step": 10849
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6356186925931525e-05,
+      "loss": 0.3821,
+      "step": 10850
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.635585109062516e-05,
+      "loss": 0.3584,
+      "step": 10851
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6355515255318794e-05,
+      "loss": 0.2569,
+      "step": 10852
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.635517942001243e-05,
+      "loss": 0.2837,
+      "step": 10853
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6354843584706063e-05,
+      "loss": 0.2797,
+      "step": 10854
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6354507749399694e-05,
+      "loss": 0.3292,
+      "step": 10855
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6354171914093332e-05,
+      "loss": 0.3564,
+      "step": 10856
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6353836078786964e-05,
+      "loss": 0.4177,
+      "step": 10857
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.63535002434806e-05,
+      "loss": 0.2912,
+      "step": 10858
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6353164408174233e-05,
+      "loss": 0.4629,
+      "step": 10859
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6352828572867868e-05,
+      "loss": 0.2987,
+      "step": 10860
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.63524927375615e-05,
+      "loss": 0.236,
+      "step": 10861
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6352156902255137e-05,
+      "loss": 0.2914,
+      "step": 10862
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6351821066948768e-05,
+      "loss": 0.4248,
+      "step": 10863
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6351485231642403e-05,
+      "loss": 0.4284,
+      "step": 10864
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6351149396336038e-05,
+      "loss": 0.2868,
+      "step": 10865
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6350813561029672e-05,
+      "loss": 0.1919,
+      "step": 10866
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6350477725723307e-05,
+      "loss": 0.3331,
+      "step": 10867
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.635014189041694e-05,
+      "loss": 0.2445,
+      "step": 10868
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6349806055110576e-05,
+      "loss": 0.3829,
+      "step": 10869
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6349470219804207e-05,
+      "loss": 0.3013,
+      "step": 10870
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6349134384497845e-05,
+      "loss": 0.3003,
+      "step": 10871
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6348798549191477e-05,
+      "loss": 0.2951,
+      "step": 10872
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.634846271388511e-05,
+      "loss": 0.5052,
+      "step": 10873
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6348126878578746e-05,
+      "loss": 0.3401,
+      "step": 10874
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.634779104327238e-05,
+      "loss": 0.3292,
+      "step": 10875
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6347455207966015e-05,
+      "loss": 0.2978,
+      "step": 10876
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.634711937265965e-05,
+      "loss": 0.2232,
+      "step": 10877
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6346783537353285e-05,
+      "loss": 0.2974,
+      "step": 10878
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6346447702046916e-05,
+      "loss": 0.3774,
+      "step": 10879
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6346111866740554e-05,
+      "loss": 0.335,
+      "step": 10880
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6345776031434185e-05,
+      "loss": 0.3217,
+      "step": 10881
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.634544019612782e-05,
+      "loss": 0.2228,
+      "step": 10882
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6345104360821454e-05,
+      "loss": 0.2227,
+      "step": 10883
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.634476852551509e-05,
+      "loss": 0.2978,
+      "step": 10884
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6344432690208724e-05,
+      "loss": 0.1914,
+      "step": 10885
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.634409685490236e-05,
+      "loss": 0.3329,
+      "step": 10886
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6343761019595993e-05,
+      "loss": 0.3757,
+      "step": 10887
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6343425184289624e-05,
+      "loss": 0.2668,
+      "step": 10888
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6343089348983262e-05,
+      "loss": 0.2256,
+      "step": 10889
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6342753513676894e-05,
+      "loss": 0.197,
+      "step": 10890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6342417678370528e-05,
+      "loss": 0.2907,
+      "step": 10891
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6342081843064163e-05,
+      "loss": 0.1753,
+      "step": 10892
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6341746007757797e-05,
+      "loss": 0.2701,
+      "step": 10893
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6341410172451432e-05,
+      "loss": 0.3467,
+      "step": 10894
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6341074337145067e-05,
+      "loss": 0.416,
+      "step": 10895
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.63407385018387e-05,
+      "loss": 0.2095,
+      "step": 10896
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6340402666532333e-05,
+      "loss": 0.5396,
+      "step": 10897
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.634006683122597e-05,
+      "loss": 0.2546,
+      "step": 10898
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6339730995919602e-05,
+      "loss": 0.2878,
+      "step": 10899
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6339395160613237e-05,
+      "loss": 0.295,
+      "step": 10900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.633905932530687e-05,
+      "loss": 0.2781,
+      "step": 10901
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6338723490000506e-05,
+      "loss": 0.2293,
+      "step": 10902
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.633838765469414e-05,
+      "loss": 0.4235,
+      "step": 10903
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6338051819387775e-05,
+      "loss": 0.1883,
+      "step": 10904
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6337715984081406e-05,
+      "loss": 0.2281,
+      "step": 10905
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.633738014877504e-05,
+      "loss": 0.3973,
+      "step": 10906
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6337044313468676e-05,
+      "loss": 0.4355,
+      "step": 10907
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.633670847816231e-05,
+      "loss": 0.418,
+      "step": 10908
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6336372642855945e-05,
+      "loss": 0.3248,
+      "step": 10909
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.633603680754958e-05,
+      "loss": 0.2871,
+      "step": 10910
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6335700972243214e-05,
+      "loss": 0.3939,
+      "step": 10911
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6335365136936846e-05,
+      "loss": 0.3729,
+      "step": 10912
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6335029301630484e-05,
+      "loss": 0.4119,
+      "step": 10913
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6334693466324115e-05,
+      "loss": 0.2527,
+      "step": 10914
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.633435763101775e-05,
+      "loss": 0.1791,
+      "step": 10915
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6334021795711384e-05,
+      "loss": 0.3798,
+      "step": 10916
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.633368596040502e-05,
+      "loss": 0.2654,
+      "step": 10917
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6333350125098654e-05,
+      "loss": 0.4367,
+      "step": 10918
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6333014289792285e-05,
+      "loss": 0.2531,
+      "step": 10919
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6332678454485923e-05,
+      "loss": 0.3685,
+      "step": 10920
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6332342619179554e-05,
+      "loss": 0.285,
+      "step": 10921
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.633200678387319e-05,
+      "loss": 0.2978,
+      "step": 10922
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6331670948566823e-05,
+      "loss": 0.4655,
+      "step": 10923
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6331335113260458e-05,
+      "loss": 0.3245,
+      "step": 10924
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6330999277954093e-05,
+      "loss": 0.3754,
+      "step": 10925
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6330663442647727e-05,
+      "loss": 0.2546,
+      "step": 10926
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6330327607341362e-05,
+      "loss": 0.3891,
+      "step": 10927
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6329991772034993e-05,
+      "loss": 0.2879,
+      "step": 10928
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.632965593672863e-05,
+      "loss": 0.4486,
+      "step": 10929
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6329320101422262e-05,
+      "loss": 0.2666,
+      "step": 10930
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6328984266115897e-05,
+      "loss": 0.2948,
+      "step": 10931
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6328648430809532e-05,
+      "loss": 0.318,
+      "step": 10932
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6328312595503166e-05,
+      "loss": 0.2579,
+      "step": 10933
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.63279767601968e-05,
+      "loss": 0.3765,
+      "step": 10934
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6327640924890436e-05,
+      "loss": 0.3723,
+      "step": 10935
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.632730508958407e-05,
+      "loss": 0.3508,
+      "step": 10936
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.63269692542777e-05,
+      "loss": 0.3686,
+      "step": 10937
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.632663341897134e-05,
+      "loss": 0.2459,
+      "step": 10938
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.632629758366497e-05,
+      "loss": 0.3928,
+      "step": 10939
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6325961748358606e-05,
+      "loss": 0.3254,
+      "step": 10940
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.632562591305224e-05,
+      "loss": 0.4632,
+      "step": 10941
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6325290077745875e-05,
+      "loss": 0.3511,
+      "step": 10942
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.632495424243951e-05,
+      "loss": 0.3977,
+      "step": 10943
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6324618407133144e-05,
+      "loss": 0.2986,
+      "step": 10944
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.632428257182678e-05,
+      "loss": 0.5138,
+      "step": 10945
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.632394673652041e-05,
+      "loss": 0.4194,
+      "step": 10946
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6323610901214048e-05,
+      "loss": 0.2349,
+      "step": 10947
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.632327506590768e-05,
+      "loss": 0.3366,
+      "step": 10948
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6322939230601314e-05,
+      "loss": 0.3264,
+      "step": 10949
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.632260339529495e-05,
+      "loss": 0.3051,
+      "step": 10950
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6322267559988583e-05,
+      "loss": 0.3945,
+      "step": 10951
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6321931724682215e-05,
+      "loss": 0.3667,
+      "step": 10952
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6321595889375853e-05,
+      "loss": 0.3293,
+      "step": 10953
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6321260054069484e-05,
+      "loss": 0.3024,
+      "step": 10954
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.632092421876312e-05,
+      "loss": 0.3428,
+      "step": 10955
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6320588383456753e-05,
+      "loss": 0.4533,
+      "step": 10956
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6320252548150388e-05,
+      "loss": 0.3169,
+      "step": 10957
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6319916712844022e-05,
+      "loss": 0.2269,
+      "step": 10958
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6319580877537657e-05,
+      "loss": 0.3863,
+      "step": 10959
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6319245042231292e-05,
+      "loss": 0.2451,
+      "step": 10960
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6318909206924923e-05,
+      "loss": 0.5505,
+      "step": 10961
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.631857337161856e-05,
+      "loss": 0.1851,
+      "step": 10962
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6318237536312192e-05,
+      "loss": 0.4027,
+      "step": 10963
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6317901701005827e-05,
+      "loss": 0.294,
+      "step": 10964
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.631756586569946e-05,
+      "loss": 0.2653,
+      "step": 10965
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6317230030393096e-05,
+      "loss": 0.3971,
+      "step": 10966
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.631689419508673e-05,
+      "loss": 0.343,
+      "step": 10967
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6316558359780366e-05,
+      "loss": 0.2262,
+      "step": 10968
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6316222524474e-05,
+      "loss": 0.3181,
+      "step": 10969
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.631588668916763e-05,
+      "loss": 0.1833,
+      "step": 10970
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.631555085386127e-05,
+      "loss": 0.6331,
+      "step": 10971
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.63152150185549e-05,
+      "loss": 0.3633,
+      "step": 10972
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6314879183248535e-05,
+      "loss": 0.3621,
+      "step": 10973
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.631454334794217e-05,
+      "loss": 0.3775,
+      "step": 10974
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6314207512635805e-05,
+      "loss": 0.4419,
+      "step": 10975
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.631387167732944e-05,
+      "loss": 0.3514,
+      "step": 10976
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6313535842023074e-05,
+      "loss": 0.1928,
+      "step": 10977
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.631320000671671e-05,
+      "loss": 0.2707,
+      "step": 10978
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.631286417141034e-05,
+      "loss": 0.2902,
+      "step": 10979
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6312528336103978e-05,
+      "loss": 0.312,
+      "step": 10980
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.631219250079761e-05,
+      "loss": 0.3786,
+      "step": 10981
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6311856665491244e-05,
+      "loss": 0.2585,
+      "step": 10982
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.631152083018488e-05,
+      "loss": 0.2241,
+      "step": 10983
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6311184994878513e-05,
+      "loss": 0.3699,
+      "step": 10984
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6310849159572148e-05,
+      "loss": 0.3339,
+      "step": 10985
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6310513324265782e-05,
+      "loss": 0.46,
+      "step": 10986
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6310177488959417e-05,
+      "loss": 0.2402,
+      "step": 10987
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.630984165365305e-05,
+      "loss": 0.3536,
+      "step": 10988
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6309505818346686e-05,
+      "loss": 0.3123,
+      "step": 10989
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6309169983040318e-05,
+      "loss": 0.6121,
+      "step": 10990
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6308834147733952e-05,
+      "loss": 0.2471,
+      "step": 10991
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6308498312427587e-05,
+      "loss": 0.2578,
+      "step": 10992
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.630816247712122e-05,
+      "loss": 0.2964,
+      "step": 10993
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6307826641814856e-05,
+      "loss": 0.5218,
+      "step": 10994
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.630749080650849e-05,
+      "loss": 0.2064,
+      "step": 10995
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6307154971202122e-05,
+      "loss": 0.4029,
+      "step": 10996
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6306819135895757e-05,
+      "loss": 0.3722,
+      "step": 10997
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.630648330058939e-05,
+      "loss": 0.2841,
+      "step": 10998
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6306147465283026e-05,
+      "loss": 0.3509,
+      "step": 10999
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.630581162997666e-05,
+      "loss": 0.2655,
+      "step": 11000
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6305475794670295e-05,
+      "loss": 0.3514,
+      "step": 11001
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.630513995936393e-05,
+      "loss": 0.2388,
+      "step": 11002
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.630480412405756e-05,
+      "loss": 0.2934,
+      "step": 11003
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.63044682887512e-05,
+      "loss": 0.2294,
+      "step": 11004
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.630413245344483e-05,
+      "loss": 0.2596,
+      "step": 11005
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6303796618138465e-05,
+      "loss": 0.302,
+      "step": 11006
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.63034607828321e-05,
+      "loss": 0.2294,
+      "step": 11007
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6303124947525734e-05,
+      "loss": 0.4817,
+      "step": 11008
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.630278911221937e-05,
+      "loss": 0.3161,
+      "step": 11009
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6302453276913004e-05,
+      "loss": 0.3441,
+      "step": 11010
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.630211744160664e-05,
+      "loss": 0.2998,
+      "step": 11011
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.630178160630027e-05,
+      "loss": 0.3461,
+      "step": 11012
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6301445770993908e-05,
+      "loss": 0.3572,
+      "step": 11013
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.630110993568754e-05,
+      "loss": 0.199,
+      "step": 11014
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6300774100381174e-05,
+      "loss": 0.2692,
+      "step": 11015
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6300438265074808e-05,
+      "loss": 0.5725,
+      "step": 11016
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.6300102429768443e-05,
+      "loss": 0.3589,
+      "step": 11017
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6299766594462078e-05,
+      "loss": 0.3206,
+      "step": 11018
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6299430759155712e-05,
+      "loss": 0.3047,
+      "step": 11019
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6299094923849347e-05,
+      "loss": 0.3444,
+      "step": 11020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6298759088542978e-05,
+      "loss": 0.1925,
+      "step": 11021
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6298423253236616e-05,
+      "loss": 0.2539,
+      "step": 11022
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6298087417930247e-05,
+      "loss": 0.1566,
+      "step": 11023
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6297751582623882e-05,
+      "loss": 0.2924,
+      "step": 11024
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6297415747317517e-05,
+      "loss": 0.315,
+      "step": 11025
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.629707991201115e-05,
+      "loss": 0.5534,
+      "step": 11026
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6296744076704786e-05,
+      "loss": 0.2778,
+      "step": 11027
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.629640824139842e-05,
+      "loss": 0.2354,
+      "step": 11028
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6296072406092055e-05,
+      "loss": 0.3291,
+      "step": 11029
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6295736570785687e-05,
+      "loss": 0.3613,
+      "step": 11030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6295400735479325e-05,
+      "loss": 0.2618,
+      "step": 11031
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6295064900172956e-05,
+      "loss": 0.2662,
+      "step": 11032
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.629472906486659e-05,
+      "loss": 0.2601,
+      "step": 11033
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6294393229560225e-05,
+      "loss": 0.4987,
+      "step": 11034
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.629405739425386e-05,
+      "loss": 0.2295,
+      "step": 11035
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6293721558947494e-05,
+      "loss": 0.3715,
+      "step": 11036
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.629338572364113e-05,
+      "loss": 0.2467,
+      "step": 11037
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6293049888334764e-05,
+      "loss": 0.3343,
+      "step": 11038
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6292714053028395e-05,
+      "loss": 0.4206,
+      "step": 11039
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.629237821772203e-05,
+      "loss": 0.3887,
+      "step": 11040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6292042382415664e-05,
+      "loss": 0.2688,
+      "step": 11041
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.62917065471093e-05,
+      "loss": 0.2862,
+      "step": 11042
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6291370711802934e-05,
+      "loss": 0.3179,
+      "step": 11043
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6291034876496568e-05,
+      "loss": 0.2986,
+      "step": 11044
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.62906990411902e-05,
+      "loss": 0.3,
+      "step": 11045
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6290363205883838e-05,
+      "loss": 0.3695,
+      "step": 11046
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.629002737057747e-05,
+      "loss": 0.2211,
+      "step": 11047
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6289691535271103e-05,
+      "loss": 0.205,
+      "step": 11048
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6289355699964738e-05,
+      "loss": 0.2603,
+      "step": 11049
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6289019864658373e-05,
+      "loss": 0.3964,
+      "step": 11050
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6288684029352007e-05,
+      "loss": 0.4365,
+      "step": 11051
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6288348194045642e-05,
+      "loss": 0.2881,
+      "step": 11052
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6288012358739277e-05,
+      "loss": 0.2641,
+      "step": 11053
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6287676523432908e-05,
+      "loss": 0.2438,
+      "step": 11054
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6287340688126546e-05,
+      "loss": 0.2808,
+      "step": 11055
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6287004852820177e-05,
+      "loss": 0.4344,
+      "step": 11056
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6286669017513812e-05,
+      "loss": 0.289,
+      "step": 11057
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6286333182207447e-05,
+      "loss": 0.2722,
+      "step": 11058
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.628599734690108e-05,
+      "loss": 0.2942,
+      "step": 11059
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6285661511594716e-05,
+      "loss": 0.3296,
+      "step": 11060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.628532567628835e-05,
+      "loss": 0.5229,
+      "step": 11061
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6284989840981985e-05,
+      "loss": 0.2911,
+      "step": 11062
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6284654005675616e-05,
+      "loss": 0.3231,
+      "step": 11063
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6284318170369254e-05,
+      "loss": 0.3564,
+      "step": 11064
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6283982335062886e-05,
+      "loss": 0.4198,
+      "step": 11065
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.628364649975652e-05,
+      "loss": 0.3022,
+      "step": 11066
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6283310664450155e-05,
+      "loss": 0.3833,
+      "step": 11067
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.628297482914379e-05,
+      "loss": 0.4696,
+      "step": 11068
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6282638993837424e-05,
+      "loss": 0.1868,
+      "step": 11069
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.628230315853106e-05,
+      "loss": 0.274,
+      "step": 11070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6281967323224694e-05,
+      "loss": 0.3094,
+      "step": 11071
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6281631487918325e-05,
+      "loss": 0.193,
+      "step": 11072
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6281295652611963e-05,
+      "loss": 0.1748,
+      "step": 11073
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6280959817305594e-05,
+      "loss": 0.2553,
+      "step": 11074
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.628062398199923e-05,
+      "loss": 0.2844,
+      "step": 11075
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6280288146692863e-05,
+      "loss": 0.2465,
+      "step": 11076
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6279952311386498e-05,
+      "loss": 0.4548,
+      "step": 11077
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6279616476080133e-05,
+      "loss": 0.4385,
+      "step": 11078
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6279280640773767e-05,
+      "loss": 0.1417,
+      "step": 11079
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6278944805467402e-05,
+      "loss": 0.2716,
+      "step": 11080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6278608970161033e-05,
+      "loss": 0.2765,
+      "step": 11081
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.627827313485467e-05,
+      "loss": 0.2646,
+      "step": 11082
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6277937299548303e-05,
+      "loss": 0.4197,
+      "step": 11083
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6277601464241937e-05,
+      "loss": 0.3462,
+      "step": 11084
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6277265628935572e-05,
+      "loss": 0.4079,
+      "step": 11085
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6276929793629206e-05,
+      "loss": 0.3704,
+      "step": 11086
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6276593958322838e-05,
+      "loss": 0.4295,
+      "step": 11087
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6276258123016476e-05,
+      "loss": 0.3382,
+      "step": 11088
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6275922287710107e-05,
+      "loss": 0.2861,
+      "step": 11089
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6275586452403742e-05,
+      "loss": 0.2356,
+      "step": 11090
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6275250617097376e-05,
+      "loss": 0.3973,
+      "step": 11091
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.627491478179101e-05,
+      "loss": 0.3029,
+      "step": 11092
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6274578946484646e-05,
+      "loss": 0.227,
+      "step": 11093
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.627424311117828e-05,
+      "loss": 0.3172,
+      "step": 11094
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6273907275871915e-05,
+      "loss": 0.4595,
+      "step": 11095
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6273571440565546e-05,
+      "loss": 0.4115,
+      "step": 11096
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6273235605259184e-05,
+      "loss": 0.3087,
+      "step": 11097
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6272899769952815e-05,
+      "loss": 0.5393,
+      "step": 11098
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.627256393464645e-05,
+      "loss": 0.2127,
+      "step": 11099
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6272228099340085e-05,
+      "loss": 0.3438,
+      "step": 11100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.627189226403372e-05,
+      "loss": 0.4223,
+      "step": 11101
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6271556428727354e-05,
+      "loss": 0.3264,
+      "step": 11102
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6271220593420985e-05,
+      "loss": 0.3715,
+      "step": 11103
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6270884758114623e-05,
+      "loss": 0.3281,
+      "step": 11104
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6270548922808255e-05,
+      "loss": 0.3838,
+      "step": 11105
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.627021308750189e-05,
+      "loss": 0.2344,
+      "step": 11106
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6269877252195524e-05,
+      "loss": 0.4234,
+      "step": 11107
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.626954141688916e-05,
+      "loss": 0.2976,
+      "step": 11108
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6269205581582793e-05,
+      "loss": 0.3837,
+      "step": 11109
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6268869746276428e-05,
+      "loss": 0.3047,
+      "step": 11110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6268533910970063e-05,
+      "loss": 0.1792,
+      "step": 11111
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6268198075663694e-05,
+      "loss": 0.4714,
+      "step": 11112
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6267862240357332e-05,
+      "loss": 0.5445,
+      "step": 11113
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6267526405050963e-05,
+      "loss": 0.2861,
+      "step": 11114
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6267190569744598e-05,
+      "loss": 0.2283,
+      "step": 11115
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6266854734438232e-05,
+      "loss": 0.2754,
+      "step": 11116
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6266518899131867e-05,
+      "loss": 0.517,
+      "step": 11117
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.62661830638255e-05,
+      "loss": 0.2676,
+      "step": 11118
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6265847228519136e-05,
+      "loss": 0.2716,
+      "step": 11119
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.626551139321277e-05,
+      "loss": 0.4105,
+      "step": 11120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6265175557906402e-05,
+      "loss": 0.2382,
+      "step": 11121
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.626483972260004e-05,
+      "loss": 0.2484,
+      "step": 11122
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.626450388729367e-05,
+      "loss": 0.4032,
+      "step": 11123
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6264168051987306e-05,
+      "loss": 0.218,
+      "step": 11124
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.626383221668094e-05,
+      "loss": 0.2402,
+      "step": 11125
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6263496381374575e-05,
+      "loss": 0.4065,
+      "step": 11126
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.626316054606821e-05,
+      "loss": 0.2902,
+      "step": 11127
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6262824710761845e-05,
+      "loss": 0.4121,
+      "step": 11128
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.626248887545548e-05,
+      "loss": 0.3271,
+      "step": 11129
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.626215304014911e-05,
+      "loss": 0.2831,
+      "step": 11130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6261817204842745e-05,
+      "loss": 0.2589,
+      "step": 11131
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.626148136953638e-05,
+      "loss": 0.2907,
+      "step": 11132
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6261145534230015e-05,
+      "loss": 0.1886,
+      "step": 11133
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.626080969892365e-05,
+      "loss": 0.262,
+      "step": 11134
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6260473863617284e-05,
+      "loss": 0.5138,
+      "step": 11135
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6260138028310915e-05,
+      "loss": 0.3809,
+      "step": 11136
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6259802193004553e-05,
+      "loss": 0.2609,
+      "step": 11137
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6259466357698184e-05,
+      "loss": 0.1782,
+      "step": 11138
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.625913052239182e-05,
+      "loss": 0.3122,
+      "step": 11139
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6258794687085454e-05,
+      "loss": 0.2095,
+      "step": 11140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.625845885177909e-05,
+      "loss": 0.2801,
+      "step": 11141
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6258123016472723e-05,
+      "loss": 0.2418,
+      "step": 11142
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6257787181166358e-05,
+      "loss": 0.3066,
+      "step": 11143
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6257451345859992e-05,
+      "loss": 0.3607,
+      "step": 11144
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6257115510553624e-05,
+      "loss": 0.4871,
+      "step": 11145
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.625677967524726e-05,
+      "loss": 0.3281,
+      "step": 11146
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6256443839940893e-05,
+      "loss": 0.5237,
+      "step": 11147
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6256108004634528e-05,
+      "loss": 0.2933,
+      "step": 11148
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6255772169328162e-05,
+      "loss": 0.253,
+      "step": 11149
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6255436334021797e-05,
+      "loss": 0.3045,
+      "step": 11150
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.625510049871543e-05,
+      "loss": 0.2306,
+      "step": 11151
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6254764663409066e-05,
+      "loss": 0.2491,
+      "step": 11152
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.62544288281027e-05,
+      "loss": 0.3196,
+      "step": 11153
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6254092992796332e-05,
+      "loss": 0.3086,
+      "step": 11154
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.625375715748997e-05,
+      "loss": 0.2229,
+      "step": 11155
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.62534213221836e-05,
+      "loss": 0.287,
+      "step": 11156
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6253085486877236e-05,
+      "loss": 0.3829,
+      "step": 11157
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.625274965157087e-05,
+      "loss": 0.3103,
+      "step": 11158
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6252413816264505e-05,
+      "loss": 0.3766,
+      "step": 11159
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.625207798095814e-05,
+      "loss": 0.2935,
+      "step": 11160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6251742145651775e-05,
+      "loss": 0.2363,
+      "step": 11161
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.625140631034541e-05,
+      "loss": 0.1625,
+      "step": 11162
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.625107047503904e-05,
+      "loss": 0.2685,
+      "step": 11163
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.625073463973268e-05,
+      "loss": 0.3789,
+      "step": 11164
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.625039880442631e-05,
+      "loss": 0.5397,
+      "step": 11165
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6250062969119944e-05,
+      "loss": 0.383,
+      "step": 11166
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.624972713381358e-05,
+      "loss": 0.4014,
+      "step": 11167
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6249391298507214e-05,
+      "loss": 0.3808,
+      "step": 11168
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.624905546320085e-05,
+      "loss": 0.3336,
+      "step": 11169
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6248719627894483e-05,
+      "loss": 0.3292,
+      "step": 11170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6248383792588118e-05,
+      "loss": 0.1878,
+      "step": 11171
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.624804795728175e-05,
+      "loss": 0.1631,
+      "step": 11172
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6247712121975387e-05,
+      "loss": 0.3475,
+      "step": 11173
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6247376286669018e-05,
+      "loss": 0.3278,
+      "step": 11174
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6247040451362653e-05,
+      "loss": 0.3213,
+      "step": 11175
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6246704616056287e-05,
+      "loss": 0.4267,
+      "step": 11176
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6246368780749922e-05,
+      "loss": 0.2167,
+      "step": 11177
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6246032945443553e-05,
+      "loss": 0.4003,
+      "step": 11178
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.624569711013719e-05,
+      "loss": 0.23,
+      "step": 11179
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6245361274830823e-05,
+      "loss": 0.2499,
+      "step": 11180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6245025439524457e-05,
+      "loss": 0.3906,
+      "step": 11181
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6244689604218092e-05,
+      "loss": 0.358,
+      "step": 11182
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6244353768911727e-05,
+      "loss": 0.3315,
+      "step": 11183
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.624401793360536e-05,
+      "loss": 0.4194,
+      "step": 11184
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6243682098298996e-05,
+      "loss": 0.3145,
+      "step": 11185
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.624334626299263e-05,
+      "loss": 0.3287,
+      "step": 11186
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6243010427686262e-05,
+      "loss": 0.3367,
+      "step": 11187
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.62426745923799e-05,
+      "loss": 0.4328,
+      "step": 11188
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.624233875707353e-05,
+      "loss": 0.3053,
+      "step": 11189
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6242002921767166e-05,
+      "loss": 0.4264,
+      "step": 11190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.62416670864608e-05,
+      "loss": 0.3816,
+      "step": 11191
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6241331251154435e-05,
+      "loss": 0.1815,
+      "step": 11192
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.624099541584807e-05,
+      "loss": 0.4535,
+      "step": 11193
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6240659580541704e-05,
+      "loss": 0.3305,
+      "step": 11194
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.624032374523534e-05,
+      "loss": 0.5338,
+      "step": 11195
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.623998790992897e-05,
+      "loss": 0.3606,
+      "step": 11196
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.623965207462261e-05,
+      "loss": 0.3498,
+      "step": 11197
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.623931623931624e-05,
+      "loss": 0.3283,
+      "step": 11198
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6238980404009874e-05,
+      "loss": 0.2561,
+      "step": 11199
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.623864456870351e-05,
+      "loss": 0.2373,
+      "step": 11200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6238308733397144e-05,
+      "loss": 0.3678,
+      "step": 11201
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6237972898090778e-05,
+      "loss": 0.322,
+      "step": 11202
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6237637062784413e-05,
+      "loss": 0.3748,
+      "step": 11203
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6237301227478047e-05,
+      "loss": 0.3747,
+      "step": 11204
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.623696539217168e-05,
+      "loss": 0.3554,
+      "step": 11205
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6236629556865317e-05,
+      "loss": 0.3016,
+      "step": 11206
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6236293721558948e-05,
+      "loss": 0.3169,
+      "step": 11207
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6235957886252583e-05,
+      "loss": 0.2695,
+      "step": 11208
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6235622050946217e-05,
+      "loss": 0.2641,
+      "step": 11209
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6235286215639852e-05,
+      "loss": 0.4274,
+      "step": 11210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6234950380333487e-05,
+      "loss": 0.3026,
+      "step": 11211
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.623461454502712e-05,
+      "loss": 0.3833,
+      "step": 11212
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6234278709720756e-05,
+      "loss": 0.2957,
+      "step": 11213
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6233942874414387e-05,
+      "loss": 0.2966,
+      "step": 11214
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.6233607039108025e-05,
+      "loss": 0.2037,
+      "step": 11215
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6233271203801656e-05,
+      "loss": 0.1678,
+      "step": 11216
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.623293536849529e-05,
+      "loss": 0.1807,
+      "step": 11217
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6232599533188926e-05,
+      "loss": 0.3506,
+      "step": 11218
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.623226369788256e-05,
+      "loss": 0.405,
+      "step": 11219
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6231927862576195e-05,
+      "loss": 0.3975,
+      "step": 11220
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.623159202726983e-05,
+      "loss": 0.2647,
+      "step": 11221
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.623125619196346e-05,
+      "loss": 0.4903,
+      "step": 11222
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6230920356657096e-05,
+      "loss": 0.2475,
+      "step": 11223
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.623058452135073e-05,
+      "loss": 0.2853,
+      "step": 11224
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6230248686044365e-05,
+      "loss": 0.369,
+      "step": 11225
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6229912850738e-05,
+      "loss": 0.1524,
+      "step": 11226
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6229577015431634e-05,
+      "loss": 0.2883,
+      "step": 11227
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.622924118012527e-05,
+      "loss": 0.2493,
+      "step": 11228
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.62289053448189e-05,
+      "loss": 0.1826,
+      "step": 11229
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6228569509512538e-05,
+      "loss": 0.4636,
+      "step": 11230
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.622823367420617e-05,
+      "loss": 0.4235,
+      "step": 11231
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6227897838899804e-05,
+      "loss": 0.244,
+      "step": 11232
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.622756200359344e-05,
+      "loss": 0.2761,
+      "step": 11233
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6227226168287073e-05,
+      "loss": 0.3257,
+      "step": 11234
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6226890332980708e-05,
+      "loss": 0.4891,
+      "step": 11235
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6226554497674343e-05,
+      "loss": 0.3588,
+      "step": 11236
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6226218662367977e-05,
+      "loss": 0.2211,
+      "step": 11237
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.622588282706161e-05,
+      "loss": 0.4172,
+      "step": 11238
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6225546991755247e-05,
+      "loss": 0.2493,
+      "step": 11239
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6225211156448878e-05,
+      "loss": 0.2703,
+      "step": 11240
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6224875321142512e-05,
+      "loss": 0.3929,
+      "step": 11241
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6224539485836147e-05,
+      "loss": 0.3621,
+      "step": 11242
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6224203650529782e-05,
+      "loss": 0.4492,
+      "step": 11243
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6223867815223416e-05,
+      "loss": 0.4179,
+      "step": 11244
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.622353197991705e-05,
+      "loss": 0.2956,
+      "step": 11245
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6223196144610686e-05,
+      "loss": 0.2502,
+      "step": 11246
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6222860309304317e-05,
+      "loss": 0.2176,
+      "step": 11247
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6222524473997955e-05,
+      "loss": 0.3688,
+      "step": 11248
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6222188638691586e-05,
+      "loss": 0.3216,
+      "step": 11249
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.622185280338522e-05,
+      "loss": 0.31,
+      "step": 11250
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6221516968078856e-05,
+      "loss": 0.2439,
+      "step": 11251
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.622118113277249e-05,
+      "loss": 0.3802,
+      "step": 11252
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6220845297466125e-05,
+      "loss": 0.2749,
+      "step": 11253
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.622050946215976e-05,
+      "loss": 0.2672,
+      "step": 11254
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6220173626853394e-05,
+      "loss": 0.4584,
+      "step": 11255
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6219837791547025e-05,
+      "loss": 0.1917,
+      "step": 11256
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6219501956240663e-05,
+      "loss": 0.3864,
+      "step": 11257
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6219166120934295e-05,
+      "loss": 0.3265,
+      "step": 11258
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.621883028562793e-05,
+      "loss": 0.2561,
+      "step": 11259
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6218494450321564e-05,
+      "loss": 0.1884,
+      "step": 11260
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.62181586150152e-05,
+      "loss": 0.1585,
+      "step": 11261
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6217822779708833e-05,
+      "loss": 0.3027,
+      "step": 11262
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6217486944402468e-05,
+      "loss": 0.3419,
+      "step": 11263
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6217151109096103e-05,
+      "loss": 0.2751,
+      "step": 11264
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6216815273789734e-05,
+      "loss": 0.2848,
+      "step": 11265
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.621647943848337e-05,
+      "loss": 0.4276,
+      "step": 11266
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6216143603177003e-05,
+      "loss": 0.3841,
+      "step": 11267
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6215807767870638e-05,
+      "loss": 0.435,
+      "step": 11268
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6215471932564272e-05,
+      "loss": 0.2751,
+      "step": 11269
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6215136097257907e-05,
+      "loss": 0.2554,
+      "step": 11270
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.621480026195154e-05,
+      "loss": 0.3396,
+      "step": 11271
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6214464426645176e-05,
+      "loss": 0.2765,
+      "step": 11272
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6214128591338808e-05,
+      "loss": 0.3788,
+      "step": 11273
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6213792756032442e-05,
+      "loss": 0.3445,
+      "step": 11274
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6213456920726077e-05,
+      "loss": 0.2568,
+      "step": 11275
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.621312108541971e-05,
+      "loss": 0.3054,
+      "step": 11276
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6212785250113346e-05,
+      "loss": 0.3531,
+      "step": 11277
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.621244941480698e-05,
+      "loss": 0.3913,
+      "step": 11278
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6212113579500616e-05,
+      "loss": 0.3182,
+      "step": 11279
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6211777744194247e-05,
+      "loss": 0.3805,
+      "step": 11280
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.621144190888788e-05,
+      "loss": 0.5901,
+      "step": 11281
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6211106073581516e-05,
+      "loss": 0.3119,
+      "step": 11282
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.621077023827515e-05,
+      "loss": 0.3873,
+      "step": 11283
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6210434402968785e-05,
+      "loss": 0.2531,
+      "step": 11284
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.621009856766242e-05,
+      "loss": 0.3354,
+      "step": 11285
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6209762732356055e-05,
+      "loss": 0.3086,
+      "step": 11286
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6209426897049686e-05,
+      "loss": 0.4609,
+      "step": 11287
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6209091061743324e-05,
+      "loss": 0.2975,
+      "step": 11288
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6208755226436955e-05,
+      "loss": 0.3743,
+      "step": 11289
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.620841939113059e-05,
+      "loss": 0.2449,
+      "step": 11290
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6208083555824225e-05,
+      "loss": 0.4712,
+      "step": 11291
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.620774772051786e-05,
+      "loss": 0.3008,
+      "step": 11292
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6207411885211494e-05,
+      "loss": 0.3376,
+      "step": 11293
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.620707604990513e-05,
+      "loss": 0.5861,
+      "step": 11294
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6206740214598763e-05,
+      "loss": 0.3917,
+      "step": 11295
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6206404379292394e-05,
+      "loss": 0.2798,
+      "step": 11296
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6206068543986032e-05,
+      "loss": 0.3564,
+      "step": 11297
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6205732708679664e-05,
+      "loss": 0.2166,
+      "step": 11298
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6205396873373298e-05,
+      "loss": 0.4886,
+      "step": 11299
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6205061038066933e-05,
+      "loss": 0.2611,
+      "step": 11300
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6204725202760568e-05,
+      "loss": 0.3271,
+      "step": 11301
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6204389367454202e-05,
+      "loss": 0.2469,
+      "step": 11302
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6204053532147837e-05,
+      "loss": 0.2384,
+      "step": 11303
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.620371769684147e-05,
+      "loss": 0.388,
+      "step": 11304
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6203381861535103e-05,
+      "loss": 0.1581,
+      "step": 11305
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.620304602622874e-05,
+      "loss": 0.3671,
+      "step": 11306
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6202710190922372e-05,
+      "loss": 0.3497,
+      "step": 11307
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6202374355616007e-05,
+      "loss": 0.4702,
+      "step": 11308
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.620203852030964e-05,
+      "loss": 0.2661,
+      "step": 11309
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6201702685003276e-05,
+      "loss": 0.3016,
+      "step": 11310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6201366849696907e-05,
+      "loss": 0.266,
+      "step": 11311
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6201031014390545e-05,
+      "loss": 0.3145,
+      "step": 11312
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6200695179084177e-05,
+      "loss": 0.4223,
+      "step": 11313
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.620035934377781e-05,
+      "loss": 0.2446,
+      "step": 11314
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6200023508471446e-05,
+      "loss": 0.2063,
+      "step": 11315
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.619968767316508e-05,
+      "loss": 0.4463,
+      "step": 11316
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6199351837858715e-05,
+      "loss": 0.2786,
+      "step": 11317
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.619901600255235e-05,
+      "loss": 0.2866,
+      "step": 11318
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6198680167245984e-05,
+      "loss": 0.3494,
+      "step": 11319
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6198344331939616e-05,
+      "loss": 0.4382,
+      "step": 11320
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6198008496633254e-05,
+      "loss": 0.3403,
+      "step": 11321
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6197672661326885e-05,
+      "loss": 0.4019,
+      "step": 11322
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.619733682602052e-05,
+      "loss": 0.3073,
+      "step": 11323
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6197000990714154e-05,
+      "loss": 0.3909,
+      "step": 11324
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.619666515540779e-05,
+      "loss": 0.4178,
+      "step": 11325
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6196329320101424e-05,
+      "loss": 0.2673,
+      "step": 11326
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6195993484795058e-05,
+      "loss": 0.4323,
+      "step": 11327
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6195657649488693e-05,
+      "loss": 0.5001,
+      "step": 11328
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6195321814182324e-05,
+      "loss": 0.2093,
+      "step": 11329
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6194985978875962e-05,
+      "loss": 0.3426,
+      "step": 11330
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6194650143569593e-05,
+      "loss": 0.4104,
+      "step": 11331
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6194314308263228e-05,
+      "loss": 0.3565,
+      "step": 11332
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6193978472956863e-05,
+      "loss": 0.2376,
+      "step": 11333
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6193642637650497e-05,
+      "loss": 0.354,
+      "step": 11334
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6193306802344132e-05,
+      "loss": 0.2576,
+      "step": 11335
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6192970967037767e-05,
+      "loss": 0.2724,
+      "step": 11336
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.61926351317314e-05,
+      "loss": 0.2029,
+      "step": 11337
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6192299296425033e-05,
+      "loss": 0.3666,
+      "step": 11338
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.619196346111867e-05,
+      "loss": 0.3067,
+      "step": 11339
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6191627625812302e-05,
+      "loss": 0.2703,
+      "step": 11340
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6191291790505937e-05,
+      "loss": 0.2684,
+      "step": 11341
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.619095595519957e-05,
+      "loss": 0.3388,
+      "step": 11342
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6190620119893206e-05,
+      "loss": 0.1996,
+      "step": 11343
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.619028428458684e-05,
+      "loss": 0.1979,
+      "step": 11344
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6189948449280475e-05,
+      "loss": 0.3035,
+      "step": 11345
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.618961261397411e-05,
+      "loss": 0.303,
+      "step": 11346
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.618927677866774e-05,
+      "loss": 0.5244,
+      "step": 11347
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.618894094336138e-05,
+      "loss": 0.2464,
+      "step": 11348
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.618860510805501e-05,
+      "loss": 0.2935,
+      "step": 11349
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6188269272748645e-05,
+      "loss": 0.2595,
+      "step": 11350
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.618793343744228e-05,
+      "loss": 0.3624,
+      "step": 11351
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6187597602135914e-05,
+      "loss": 0.3439,
+      "step": 11352
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.618726176682955e-05,
+      "loss": 0.4226,
+      "step": 11353
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6186925931523184e-05,
+      "loss": 0.126,
+      "step": 11354
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6186590096216818e-05,
+      "loss": 0.3134,
+      "step": 11355
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.618625426091045e-05,
+      "loss": 0.2899,
+      "step": 11356
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6185918425604084e-05,
+      "loss": 0.337,
+      "step": 11357
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.618558259029772e-05,
+      "loss": 0.1705,
+      "step": 11358
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6185246754991353e-05,
+      "loss": 0.5805,
+      "step": 11359
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6184910919684988e-05,
+      "loss": 0.2037,
+      "step": 11360
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6184575084378623e-05,
+      "loss": 0.3012,
+      "step": 11361
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6184239249072254e-05,
+      "loss": 0.4452,
+      "step": 11362
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6183903413765892e-05,
+      "loss": 0.2251,
+      "step": 11363
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6183567578459523e-05,
+      "loss": 0.2565,
+      "step": 11364
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6183231743153158e-05,
+      "loss": 0.2626,
+      "step": 11365
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6182895907846793e-05,
+      "loss": 0.4442,
+      "step": 11366
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6182560072540427e-05,
+      "loss": 0.2743,
+      "step": 11367
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6182224237234062e-05,
+      "loss": 0.2652,
+      "step": 11368
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6181888401927697e-05,
+      "loss": 0.4894,
+      "step": 11369
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.618155256662133e-05,
+      "loss": 0.5686,
+      "step": 11370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6181216731314962e-05,
+      "loss": 0.238,
+      "step": 11371
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.61808808960086e-05,
+      "loss": 0.2386,
+      "step": 11372
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6180545060702232e-05,
+      "loss": 0.2717,
+      "step": 11373
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6180209225395866e-05,
+      "loss": 0.2521,
+      "step": 11374
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.61798733900895e-05,
+      "loss": 0.1984,
+      "step": 11375
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6179537554783136e-05,
+      "loss": 0.4447,
+      "step": 11376
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.617920171947677e-05,
+      "loss": 0.348,
+      "step": 11377
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6178865884170405e-05,
+      "loss": 0.3256,
+      "step": 11378
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.617853004886404e-05,
+      "loss": 0.4235,
+      "step": 11379
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.617819421355767e-05,
+      "loss": 0.4151,
+      "step": 11380
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.617785837825131e-05,
+      "loss": 0.3759,
+      "step": 11381
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.617752254294494e-05,
+      "loss": 0.3741,
+      "step": 11382
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6177186707638575e-05,
+      "loss": 0.343,
+      "step": 11383
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.617685087233221e-05,
+      "loss": 0.3117,
+      "step": 11384
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6176515037025844e-05,
+      "loss": 0.313,
+      "step": 11385
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.617617920171948e-05,
+      "loss": 0.198,
+      "step": 11386
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6175843366413113e-05,
+      "loss": 0.2388,
+      "step": 11387
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6175507531106748e-05,
+      "loss": 0.2241,
+      "step": 11388
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.617517169580038e-05,
+      "loss": 0.3065,
+      "step": 11389
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6174835860494017e-05,
+      "loss": 0.4031,
+      "step": 11390
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.617450002518765e-05,
+      "loss": 0.2497,
+      "step": 11391
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6174164189881283e-05,
+      "loss": 0.3159,
+      "step": 11392
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6173828354574918e-05,
+      "loss": 0.5598,
+      "step": 11393
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6173492519268553e-05,
+      "loss": 0.2562,
+      "step": 11394
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6173156683962187e-05,
+      "loss": 0.2592,
+      "step": 11395
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6172820848655822e-05,
+      "loss": 0.2678,
+      "step": 11396
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6172485013349456e-05,
+      "loss": 0.4703,
+      "step": 11397
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6172149178043088e-05,
+      "loss": 0.184,
+      "step": 11398
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6171813342736726e-05,
+      "loss": 0.4058,
+      "step": 11399
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6171477507430357e-05,
+      "loss": 0.391,
+      "step": 11400
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.617114167212399e-05,
+      "loss": 0.4857,
+      "step": 11401
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6170805836817626e-05,
+      "loss": 0.3054,
+      "step": 11402
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.617047000151126e-05,
+      "loss": 0.2204,
+      "step": 11403
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6170134166204892e-05,
+      "loss": 0.3942,
+      "step": 11404
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.616979833089853e-05,
+      "loss": 0.3767,
+      "step": 11405
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.616946249559216e-05,
+      "loss": 0.2982,
+      "step": 11406
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6169126660285796e-05,
+      "loss": 0.2461,
+      "step": 11407
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.616879082497943e-05,
+      "loss": 0.3981,
+      "step": 11408
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6168454989673065e-05,
+      "loss": 0.4043,
+      "step": 11409
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.61681191543667e-05,
+      "loss": 0.2944,
+      "step": 11410
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.6167783319060335e-05,
+      "loss": 0.3234,
+      "step": 11411
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.616744748375397e-05,
+      "loss": 0.2705,
+      "step": 11412
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.61671116484476e-05,
+      "loss": 0.4016,
+      "step": 11413
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.616677581314124e-05,
+      "loss": 0.4089,
+      "step": 11414
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.616643997783487e-05,
+      "loss": 0.41,
+      "step": 11415
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6166104142528505e-05,
+      "loss": 0.4451,
+      "step": 11416
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.616576830722214e-05,
+      "loss": 0.3362,
+      "step": 11417
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6165432471915774e-05,
+      "loss": 0.2417,
+      "step": 11418
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.616509663660941e-05,
+      "loss": 0.4064,
+      "step": 11419
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6164760801303043e-05,
+      "loss": 0.2941,
+      "step": 11420
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6164424965996678e-05,
+      "loss": 0.468,
+      "step": 11421
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.616408913069031e-05,
+      "loss": 0.3076,
+      "step": 11422
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6163753295383947e-05,
+      "loss": 0.3119,
+      "step": 11423
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.616341746007758e-05,
+      "loss": 0.3275,
+      "step": 11424
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6163081624771213e-05,
+      "loss": 0.2646,
+      "step": 11425
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6162745789464848e-05,
+      "loss": 0.349,
+      "step": 11426
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6162409954158482e-05,
+      "loss": 0.2856,
+      "step": 11427
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6162074118852117e-05,
+      "loss": 0.4267,
+      "step": 11428
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.616173828354575e-05,
+      "loss": 0.36,
+      "step": 11429
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6161402448239386e-05,
+      "loss": 0.2345,
+      "step": 11430
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6161066612933018e-05,
+      "loss": 0.3153,
+      "step": 11431
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6160730777626656e-05,
+      "loss": 0.3729,
+      "step": 11432
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6160394942320287e-05,
+      "loss": 0.378,
+      "step": 11433
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.616005910701392e-05,
+      "loss": 0.3673,
+      "step": 11434
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6159723271707556e-05,
+      "loss": 0.3463,
+      "step": 11435
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.615938743640119e-05,
+      "loss": 0.3541,
+      "step": 11436
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6159051601094825e-05,
+      "loss": 0.3771,
+      "step": 11437
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.615871576578846e-05,
+      "loss": 0.3932,
+      "step": 11438
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6158379930482095e-05,
+      "loss": 0.2086,
+      "step": 11439
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6158044095175726e-05,
+      "loss": 0.4611,
+      "step": 11440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6157708259869364e-05,
+      "loss": 0.3085,
+      "step": 11441
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6157372424562995e-05,
+      "loss": 0.5774,
+      "step": 11442
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.615703658925663e-05,
+      "loss": 0.4487,
+      "step": 11443
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6156700753950265e-05,
+      "loss": 0.3521,
+      "step": 11444
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.61563649186439e-05,
+      "loss": 0.3599,
+      "step": 11445
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.615602908333753e-05,
+      "loss": 0.2961,
+      "step": 11446
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.615569324803117e-05,
+      "loss": 0.4743,
+      "step": 11447
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.61553574127248e-05,
+      "loss": 0.3762,
+      "step": 11448
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6155021577418434e-05,
+      "loss": 0.2145,
+      "step": 11449
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.615468574211207e-05,
+      "loss": 0.254,
+      "step": 11450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6154349906805704e-05,
+      "loss": 0.2657,
+      "step": 11451
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.615401407149934e-05,
+      "loss": 0.3884,
+      "step": 11452
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6153678236192973e-05,
+      "loss": 0.3059,
+      "step": 11453
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6153342400886608e-05,
+      "loss": 0.4182,
+      "step": 11454
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.615300656558024e-05,
+      "loss": 0.3886,
+      "step": 11455
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6152670730273877e-05,
+      "loss": 0.29,
+      "step": 11456
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6152334894967508e-05,
+      "loss": 0.2473,
+      "step": 11457
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6151999059661143e-05,
+      "loss": 0.2619,
+      "step": 11458
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6151663224354778e-05,
+      "loss": 0.394,
+      "step": 11459
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6151327389048412e-05,
+      "loss": 0.3902,
+      "step": 11460
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6150991553742047e-05,
+      "loss": 0.299,
+      "step": 11461
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6150655718435678e-05,
+      "loss": 0.1652,
+      "step": 11462
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6150319883129316e-05,
+      "loss": 0.2499,
+      "step": 11463
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6149984047822947e-05,
+      "loss": 0.3275,
+      "step": 11464
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6149648212516582e-05,
+      "loss": 0.3665,
+      "step": 11465
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6149312377210217e-05,
+      "loss": 0.3907,
+      "step": 11466
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.614897654190385e-05,
+      "loss": 0.2595,
+      "step": 11467
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6148640706597486e-05,
+      "loss": 0.4851,
+      "step": 11468
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.614830487129112e-05,
+      "loss": 0.5124,
+      "step": 11469
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6147969035984755e-05,
+      "loss": 0.2848,
+      "step": 11470
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6147633200678387e-05,
+      "loss": 0.2048,
+      "step": 11471
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6147297365372025e-05,
+      "loss": 0.18,
+      "step": 11472
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6146961530065656e-05,
+      "loss": 0.4443,
+      "step": 11473
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.614662569475929e-05,
+      "loss": 0.4274,
+      "step": 11474
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6146289859452925e-05,
+      "loss": 0.3944,
+      "step": 11475
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.614595402414656e-05,
+      "loss": 0.4034,
+      "step": 11476
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6145618188840194e-05,
+      "loss": 0.3256,
+      "step": 11477
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.614528235353383e-05,
+      "loss": 0.3282,
+      "step": 11478
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6144946518227464e-05,
+      "loss": 0.3647,
+      "step": 11479
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6144610682921095e-05,
+      "loss": 0.2513,
+      "step": 11480
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6144274847614733e-05,
+      "loss": 0.367,
+      "step": 11481
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6143939012308364e-05,
+      "loss": 0.3944,
+      "step": 11482
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6143603177002e-05,
+      "loss": 0.2144,
+      "step": 11483
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6143267341695634e-05,
+      "loss": 0.1992,
+      "step": 11484
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6142931506389268e-05,
+      "loss": 0.3598,
+      "step": 11485
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6142595671082903e-05,
+      "loss": 0.2971,
+      "step": 11486
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6142259835776537e-05,
+      "loss": 0.2722,
+      "step": 11487
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6141924000470172e-05,
+      "loss": 0.3281,
+      "step": 11488
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6141588165163803e-05,
+      "loss": 0.2329,
+      "step": 11489
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.614125232985744e-05,
+      "loss": 0.3644,
+      "step": 11490
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6140916494551073e-05,
+      "loss": 0.5981,
+      "step": 11491
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6140580659244707e-05,
+      "loss": 0.3025,
+      "step": 11492
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6140244823938342e-05,
+      "loss": 0.3961,
+      "step": 11493
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6139908988631977e-05,
+      "loss": 0.3441,
+      "step": 11494
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6139573153325608e-05,
+      "loss": 0.349,
+      "step": 11495
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6139237318019246e-05,
+      "loss": 0.228,
+      "step": 11496
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6138901482712877e-05,
+      "loss": 0.2971,
+      "step": 11497
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6138565647406512e-05,
+      "loss": 0.1786,
+      "step": 11498
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6138229812100146e-05,
+      "loss": 0.3181,
+      "step": 11499
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.613789397679378e-05,
+      "loss": 0.3336,
+      "step": 11500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6137558141487416e-05,
+      "loss": 0.3252,
+      "step": 11501
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.613722230618105e-05,
+      "loss": 0.3491,
+      "step": 11502
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6136886470874685e-05,
+      "loss": 0.2302,
+      "step": 11503
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6136550635568316e-05,
+      "loss": 0.2964,
+      "step": 11504
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6136214800261954e-05,
+      "loss": 0.3141,
+      "step": 11505
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6135878964955586e-05,
+      "loss": 0.27,
+      "step": 11506
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.613554312964922e-05,
+      "loss": 0.5451,
+      "step": 11507
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6135207294342855e-05,
+      "loss": 0.2556,
+      "step": 11508
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.613487145903649e-05,
+      "loss": 0.302,
+      "step": 11509
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6134535623730124e-05,
+      "loss": 0.1999,
+      "step": 11510
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.613419978842376e-05,
+      "loss": 0.3257,
+      "step": 11511
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6133863953117393e-05,
+      "loss": 0.2768,
+      "step": 11512
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6133528117811025e-05,
+      "loss": 0.471,
+      "step": 11513
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6133192282504663e-05,
+      "loss": 0.2806,
+      "step": 11514
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6132856447198294e-05,
+      "loss": 0.3048,
+      "step": 11515
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.613252061189193e-05,
+      "loss": 0.3205,
+      "step": 11516
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6132184776585563e-05,
+      "loss": 0.3246,
+      "step": 11517
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6131848941279198e-05,
+      "loss": 0.1862,
+      "step": 11518
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6131513105972833e-05,
+      "loss": 0.2725,
+      "step": 11519
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6131177270666467e-05,
+      "loss": 0.328,
+      "step": 11520
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6130841435360102e-05,
+      "loss": 0.283,
+      "step": 11521
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6130505600053733e-05,
+      "loss": 0.3761,
+      "step": 11522
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.613016976474737e-05,
+      "loss": 0.239,
+      "step": 11523
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6129833929441002e-05,
+      "loss": 0.2162,
+      "step": 11524
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6129498094134637e-05,
+      "loss": 0.3584,
+      "step": 11525
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6129162258828272e-05,
+      "loss": 0.3009,
+      "step": 11526
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6128826423521906e-05,
+      "loss": 0.3106,
+      "step": 11527
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.612849058821554e-05,
+      "loss": 0.2975,
+      "step": 11528
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6128154752909176e-05,
+      "loss": 0.2287,
+      "step": 11529
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.612781891760281e-05,
+      "loss": 0.2878,
+      "step": 11530
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.612748308229644e-05,
+      "loss": 0.3219,
+      "step": 11531
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.612714724699008e-05,
+      "loss": 0.188,
+      "step": 11532
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.612681141168371e-05,
+      "loss": 0.435,
+      "step": 11533
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6126475576377346e-05,
+      "loss": 0.3283,
+      "step": 11534
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.612613974107098e-05,
+      "loss": 0.4399,
+      "step": 11535
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6125803905764615e-05,
+      "loss": 0.2609,
+      "step": 11536
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6125468070458246e-05,
+      "loss": 0.2465,
+      "step": 11537
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6125132235151884e-05,
+      "loss": 0.2183,
+      "step": 11538
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6124796399845515e-05,
+      "loss": 0.307,
+      "step": 11539
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.612446056453915e-05,
+      "loss": 0.2345,
+      "step": 11540
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6124124729232785e-05,
+      "loss": 0.2414,
+      "step": 11541
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.612378889392642e-05,
+      "loss": 0.759,
+      "step": 11542
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6123453058620054e-05,
+      "loss": 0.3924,
+      "step": 11543
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.612311722331369e-05,
+      "loss": 0.3046,
+      "step": 11544
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6122781388007323e-05,
+      "loss": 0.326,
+      "step": 11545
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6122445552700955e-05,
+      "loss": 0.2531,
+      "step": 11546
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6122109717394593e-05,
+      "loss": 0.3736,
+      "step": 11547
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6121773882088224e-05,
+      "loss": 0.2726,
+      "step": 11548
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.612143804678186e-05,
+      "loss": 0.2004,
+      "step": 11549
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6121102211475493e-05,
+      "loss": 0.5119,
+      "step": 11550
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6120766376169128e-05,
+      "loss": 0.1489,
+      "step": 11551
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6120430540862762e-05,
+      "loss": 0.1478,
+      "step": 11552
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6120094705556397e-05,
+      "loss": 0.354,
+      "step": 11553
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6119758870250032e-05,
+      "loss": 0.2829,
+      "step": 11554
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6119423034943663e-05,
+      "loss": 0.3333,
+      "step": 11555
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.61190871996373e-05,
+      "loss": 0.1997,
+      "step": 11556
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6118751364330932e-05,
+      "loss": 0.2711,
+      "step": 11557
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6118415529024567e-05,
+      "loss": 0.2321,
+      "step": 11558
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.61180796937182e-05,
+      "loss": 0.4665,
+      "step": 11559
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6117743858411836e-05,
+      "loss": 0.2154,
+      "step": 11560
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.611740802310547e-05,
+      "loss": 0.2216,
+      "step": 11561
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6117072187799106e-05,
+      "loss": 0.3155,
+      "step": 11562
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.611673635249274e-05,
+      "loss": 0.2167,
+      "step": 11563
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.611640051718637e-05,
+      "loss": 0.3241,
+      "step": 11564
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.611606468188001e-05,
+      "loss": 0.235,
+      "step": 11565
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.611572884657364e-05,
+      "loss": 0.2748,
+      "step": 11566
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6115393011267275e-05,
+      "loss": 0.6697,
+      "step": 11567
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.611505717596091e-05,
+      "loss": 0.3875,
+      "step": 11568
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6114721340654545e-05,
+      "loss": 0.2923,
+      "step": 11569
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.611438550534818e-05,
+      "loss": 0.2406,
+      "step": 11570
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6114049670041814e-05,
+      "loss": 0.4232,
+      "step": 11571
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.611371383473545e-05,
+      "loss": 0.4324,
+      "step": 11572
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.611337799942908e-05,
+      "loss": 0.3462,
+      "step": 11573
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6113042164122718e-05,
+      "loss": 0.2717,
+      "step": 11574
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.611270632881635e-05,
+      "loss": 0.3207,
+      "step": 11575
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6112370493509984e-05,
+      "loss": 0.1848,
+      "step": 11576
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.611203465820362e-05,
+      "loss": 0.2521,
+      "step": 11577
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6111698822897253e-05,
+      "loss": 0.2815,
+      "step": 11578
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6111362987590888e-05,
+      "loss": 0.4446,
+      "step": 11579
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6111027152284522e-05,
+      "loss": 0.2013,
+      "step": 11580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6110691316978154e-05,
+      "loss": 0.4495,
+      "step": 11581
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.611035548167179e-05,
+      "loss": 0.3738,
+      "step": 11582
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6110019646365423e-05,
+      "loss": 0.5439,
+      "step": 11583
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6109683811059058e-05,
+      "loss": 0.3821,
+      "step": 11584
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6109347975752692e-05,
+      "loss": 0.2496,
+      "step": 11585
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6109012140446327e-05,
+      "loss": 0.2525,
+      "step": 11586
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.610867630513996e-05,
+      "loss": 0.2797,
+      "step": 11587
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6108340469833593e-05,
+      "loss": 0.2609,
+      "step": 11588
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.610800463452723e-05,
+      "loss": 0.4574,
+      "step": 11589
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6107668799220862e-05,
+      "loss": 0.2264,
+      "step": 11590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6107332963914497e-05,
+      "loss": 0.2264,
+      "step": 11591
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.610699712860813e-05,
+      "loss": 0.2088,
+      "step": 11592
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6106661293301766e-05,
+      "loss": 0.2516,
+      "step": 11593
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.61063254579954e-05,
+      "loss": 0.2332,
+      "step": 11594
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6105989622689035e-05,
+      "loss": 0.1796,
+      "step": 11595
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.610565378738267e-05,
+      "loss": 0.345,
+      "step": 11596
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.61053179520763e-05,
+      "loss": 0.4,
+      "step": 11597
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.610498211676994e-05,
+      "loss": 0.1873,
+      "step": 11598
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.610464628146357e-05,
+      "loss": 0.2493,
+      "step": 11599
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6104310446157205e-05,
+      "loss": 0.4049,
+      "step": 11600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.610397461085084e-05,
+      "loss": 0.4786,
+      "step": 11601
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6103638775544474e-05,
+      "loss": 0.2713,
+      "step": 11602
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.610330294023811e-05,
+      "loss": 0.2209,
+      "step": 11603
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6102967104931744e-05,
+      "loss": 0.2634,
+      "step": 11604
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.610263126962538e-05,
+      "loss": 0.2497,
+      "step": 11605
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.610229543431901e-05,
+      "loss": 0.2984,
+      "step": 11606
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6101959599012648e-05,
+      "loss": 0.1933,
+      "step": 11607
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.610162376370628e-05,
+      "loss": 0.2454,
+      "step": 11608
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6101287928399914e-05,
+      "loss": 0.2695,
+      "step": 11609
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6100952093093548e-05,
+      "loss": 0.3229,
+      "step": 11610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6100616257787183e-05,
+      "loss": 0.3063,
+      "step": 11611
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.6100280422480818e-05,
+      "loss": 0.3802,
+      "step": 11612
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6099944587174452e-05,
+      "loss": 0.2767,
+      "step": 11613
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6099608751868087e-05,
+      "loss": 0.3845,
+      "step": 11614
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6099272916561718e-05,
+      "loss": 0.273,
+      "step": 11615
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6098937081255356e-05,
+      "loss": 0.5036,
+      "step": 11616
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6098601245948987e-05,
+      "loss": 0.2452,
+      "step": 11617
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6098265410642622e-05,
+      "loss": 0.2777,
+      "step": 11618
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6097929575336257e-05,
+      "loss": 0.343,
+      "step": 11619
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.609759374002989e-05,
+      "loss": 0.2671,
+      "step": 11620
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6097257904723526e-05,
+      "loss": 0.3374,
+      "step": 11621
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.609692206941716e-05,
+      "loss": 0.2442,
+      "step": 11622
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6096586234110795e-05,
+      "loss": 0.3775,
+      "step": 11623
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6096250398804427e-05,
+      "loss": 0.4058,
+      "step": 11624
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6095914563498065e-05,
+      "loss": 0.3544,
+      "step": 11625
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6095578728191696e-05,
+      "loss": 0.2631,
+      "step": 11626
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.609524289288533e-05,
+      "loss": 0.3348,
+      "step": 11627
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6094907057578965e-05,
+      "loss": 0.5793,
+      "step": 11628
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.60945712222726e-05,
+      "loss": 0.2646,
+      "step": 11629
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.609423538696623e-05,
+      "loss": 0.4045,
+      "step": 11630
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.609389955165987e-05,
+      "loss": 0.45,
+      "step": 11631
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.60935637163535e-05,
+      "loss": 0.5571,
+      "step": 11632
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6093227881047135e-05,
+      "loss": 0.3276,
+      "step": 11633
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.609289204574077e-05,
+      "loss": 0.393,
+      "step": 11634
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6092556210434404e-05,
+      "loss": 0.3308,
+      "step": 11635
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.609222037512804e-05,
+      "loss": 0.4337,
+      "step": 11636
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6091884539821674e-05,
+      "loss": 0.2918,
+      "step": 11637
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6091548704515308e-05,
+      "loss": 0.2173,
+      "step": 11638
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.609121286920894e-05,
+      "loss": 0.3189,
+      "step": 11639
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6090877033902574e-05,
+      "loss": 0.357,
+      "step": 11640
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.609054119859621e-05,
+      "loss": 0.2922,
+      "step": 11641
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6090205363289843e-05,
+      "loss": 0.2126,
+      "step": 11642
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6089869527983478e-05,
+      "loss": 0.3288,
+      "step": 11643
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6089533692677113e-05,
+      "loss": 0.3296,
+      "step": 11644
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6089197857370747e-05,
+      "loss": 0.2681,
+      "step": 11645
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.608886202206438e-05,
+      "loss": 0.2507,
+      "step": 11646
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6088526186758017e-05,
+      "loss": 0.236,
+      "step": 11647
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6088190351451648e-05,
+      "loss": 0.1729,
+      "step": 11648
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6087854516145283e-05,
+      "loss": 0.307,
+      "step": 11649
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6087518680838917e-05,
+      "loss": 0.2344,
+      "step": 11650
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6087182845532552e-05,
+      "loss": 0.2966,
+      "step": 11651
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6086847010226187e-05,
+      "loss": 0.2703,
+      "step": 11652
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.608651117491982e-05,
+      "loss": 0.4325,
+      "step": 11653
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6086175339613456e-05,
+      "loss": 0.1106,
+      "step": 11654
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6085839504307087e-05,
+      "loss": 0.3,
+      "step": 11655
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6085503669000725e-05,
+      "loss": 0.348,
+      "step": 11656
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6085167833694356e-05,
+      "loss": 0.3583,
+      "step": 11657
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.608483199838799e-05,
+      "loss": 0.4428,
+      "step": 11658
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6084496163081626e-05,
+      "loss": 0.3943,
+      "step": 11659
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.608416032777526e-05,
+      "loss": 0.6448,
+      "step": 11660
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6083824492468895e-05,
+      "loss": 0.3486,
+      "step": 11661
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.608348865716253e-05,
+      "loss": 0.193,
+      "step": 11662
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6083152821856164e-05,
+      "loss": 0.1524,
+      "step": 11663
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6082816986549796e-05,
+      "loss": 0.4882,
+      "step": 11664
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6082481151243434e-05,
+      "loss": 0.4266,
+      "step": 11665
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6082145315937065e-05,
+      "loss": 0.3942,
+      "step": 11666
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.60818094806307e-05,
+      "loss": 0.2303,
+      "step": 11667
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6081473645324334e-05,
+      "loss": 0.4164,
+      "step": 11668
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.608113781001797e-05,
+      "loss": 0.3643,
+      "step": 11669
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6080801974711603e-05,
+      "loss": 0.3016,
+      "step": 11670
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6080466139405238e-05,
+      "loss": 0.397,
+      "step": 11671
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.608013030409887e-05,
+      "loss": 0.2921,
+      "step": 11672
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6079794468792504e-05,
+      "loss": 0.4206,
+      "step": 11673
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.607945863348614e-05,
+      "loss": 0.3841,
+      "step": 11674
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6079122798179773e-05,
+      "loss": 0.399,
+      "step": 11675
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6078786962873408e-05,
+      "loss": 0.4174,
+      "step": 11676
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6078451127567043e-05,
+      "loss": 0.3815,
+      "step": 11677
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6078115292260677e-05,
+      "loss": 0.4379,
+      "step": 11678
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.607777945695431e-05,
+      "loss": 0.4002,
+      "step": 11679
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6077443621647946e-05,
+      "loss": 0.2577,
+      "step": 11680
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6077107786341578e-05,
+      "loss": 0.3791,
+      "step": 11681
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6076771951035212e-05,
+      "loss": 0.4674,
+      "step": 11682
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6076436115728847e-05,
+      "loss": 0.2725,
+      "step": 11683
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.607610028042248e-05,
+      "loss": 0.3926,
+      "step": 11684
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6075764445116116e-05,
+      "loss": 0.2683,
+      "step": 11685
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.607542860980975e-05,
+      "loss": 0.2899,
+      "step": 11686
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6075092774503386e-05,
+      "loss": 0.22,
+      "step": 11687
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6074756939197017e-05,
+      "loss": 0.272,
+      "step": 11688
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6074421103890655e-05,
+      "loss": 0.2117,
+      "step": 11689
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6074085268584286e-05,
+      "loss": 0.2824,
+      "step": 11690
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.607374943327792e-05,
+      "loss": 0.3948,
+      "step": 11691
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6073413597971555e-05,
+      "loss": 0.2682,
+      "step": 11692
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.607307776266519e-05,
+      "loss": 0.3484,
+      "step": 11693
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6072741927358825e-05,
+      "loss": 0.4443,
+      "step": 11694
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.607240609205246e-05,
+      "loss": 0.246,
+      "step": 11695
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6072070256746094e-05,
+      "loss": 0.3781,
+      "step": 11696
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6071734421439725e-05,
+      "loss": 0.2163,
+      "step": 11697
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6071398586133363e-05,
+      "loss": 0.3054,
+      "step": 11698
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6071062750826995e-05,
+      "loss": 0.2237,
+      "step": 11699
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.607072691552063e-05,
+      "loss": 0.3205,
+      "step": 11700
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6070391080214264e-05,
+      "loss": 0.391,
+      "step": 11701
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.60700552449079e-05,
+      "loss": 0.5595,
+      "step": 11702
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6069719409601533e-05,
+      "loss": 0.1813,
+      "step": 11703
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6069383574295168e-05,
+      "loss": 0.4238,
+      "step": 11704
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6069047738988803e-05,
+      "loss": 0.316,
+      "step": 11705
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6068711903682434e-05,
+      "loss": 0.4957,
+      "step": 11706
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6068376068376072e-05,
+      "loss": 0.2585,
+      "step": 11707
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6068040233069703e-05,
+      "loss": 0.3642,
+      "step": 11708
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6067704397763338e-05,
+      "loss": 0.3601,
+      "step": 11709
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6067368562456972e-05,
+      "loss": 0.3814,
+      "step": 11710
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6067032727150607e-05,
+      "loss": 0.3551,
+      "step": 11711
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.606669689184424e-05,
+      "loss": 0.2852,
+      "step": 11712
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6066361056537876e-05,
+      "loss": 0.3507,
+      "step": 11713
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.606602522123151e-05,
+      "loss": 0.2071,
+      "step": 11714
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6065689385925142e-05,
+      "loss": 0.439,
+      "step": 11715
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6065353550618777e-05,
+      "loss": 0.3484,
+      "step": 11716
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.606501771531241e-05,
+      "loss": 0.408,
+      "step": 11717
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6064681880006046e-05,
+      "loss": 0.3523,
+      "step": 11718
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.606434604469968e-05,
+      "loss": 0.434,
+      "step": 11719
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6064010209393315e-05,
+      "loss": 0.1937,
+      "step": 11720
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6063674374086947e-05,
+      "loss": 0.3156,
+      "step": 11721
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6063338538780585e-05,
+      "loss": 0.1579,
+      "step": 11722
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6063002703474216e-05,
+      "loss": 0.4114,
+      "step": 11723
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.606266686816785e-05,
+      "loss": 0.3757,
+      "step": 11724
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6062331032861485e-05,
+      "loss": 0.3701,
+      "step": 11725
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.606199519755512e-05,
+      "loss": 0.4159,
+      "step": 11726
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6061659362248755e-05,
+      "loss": 0.3601,
+      "step": 11727
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.606132352694239e-05,
+      "loss": 0.3617,
+      "step": 11728
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6060987691636024e-05,
+      "loss": 0.3016,
+      "step": 11729
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6060651856329655e-05,
+      "loss": 0.2146,
+      "step": 11730
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6060316021023293e-05,
+      "loss": 0.463,
+      "step": 11731
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6059980185716924e-05,
+      "loss": 0.2142,
+      "step": 11732
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.605964435041056e-05,
+      "loss": 0.3718,
+      "step": 11733
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6059308515104194e-05,
+      "loss": 0.4764,
+      "step": 11734
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.605897267979783e-05,
+      "loss": 0.2984,
+      "step": 11735
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6058636844491463e-05,
+      "loss": 0.4557,
+      "step": 11736
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6058301009185098e-05,
+      "loss": 0.4279,
+      "step": 11737
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6057965173878732e-05,
+      "loss": 0.3067,
+      "step": 11738
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6057629338572364e-05,
+      "loss": 0.2272,
+      "step": 11739
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6057293503266e-05,
+      "loss": 0.3529,
+      "step": 11740
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6056957667959633e-05,
+      "loss": 0.334,
+      "step": 11741
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6056621832653268e-05,
+      "loss": 0.3935,
+      "step": 11742
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6056285997346902e-05,
+      "loss": 0.2776,
+      "step": 11743
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6055950162040537e-05,
+      "loss": 0.4113,
+      "step": 11744
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.605561432673417e-05,
+      "loss": 0.214,
+      "step": 11745
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6055278491427806e-05,
+      "loss": 0.4557,
+      "step": 11746
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.605494265612144e-05,
+      "loss": 0.3047,
+      "step": 11747
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6054606820815072e-05,
+      "loss": 0.2562,
+      "step": 11748
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.605427098550871e-05,
+      "loss": 0.2307,
+      "step": 11749
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.605393515020234e-05,
+      "loss": 0.2978,
+      "step": 11750
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6053599314895976e-05,
+      "loss": 0.4124,
+      "step": 11751
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.605326347958961e-05,
+      "loss": 0.493,
+      "step": 11752
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6052927644283245e-05,
+      "loss": 0.3624,
+      "step": 11753
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.605259180897688e-05,
+      "loss": 0.2533,
+      "step": 11754
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6052255973670515e-05,
+      "loss": 0.3982,
+      "step": 11755
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.605192013836415e-05,
+      "loss": 0.2886,
+      "step": 11756
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.605158430305778e-05,
+      "loss": 0.4391,
+      "step": 11757
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.605124846775142e-05,
+      "loss": 0.2612,
+      "step": 11758
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.605091263244505e-05,
+      "loss": 0.2184,
+      "step": 11759
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6050576797138684e-05,
+      "loss": 0.3521,
+      "step": 11760
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.605024096183232e-05,
+      "loss": 0.2997,
+      "step": 11761
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6049905126525954e-05,
+      "loss": 0.3451,
+      "step": 11762
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6049569291219585e-05,
+      "loss": 0.279,
+      "step": 11763
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6049233455913223e-05,
+      "loss": 0.214,
+      "step": 11764
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6048897620606854e-05,
+      "loss": 0.337,
+      "step": 11765
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.604856178530049e-05,
+      "loss": 0.3349,
+      "step": 11766
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6048225949994124e-05,
+      "loss": 0.2914,
+      "step": 11767
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6047890114687758e-05,
+      "loss": 0.2894,
+      "step": 11768
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6047554279381393e-05,
+      "loss": 0.3456,
+      "step": 11769
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6047218444075027e-05,
+      "loss": 0.358,
+      "step": 11770
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6046882608768662e-05,
+      "loss": 0.3175,
+      "step": 11771
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6046546773462293e-05,
+      "loss": 0.2564,
+      "step": 11772
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.604621093815593e-05,
+      "loss": 0.1907,
+      "step": 11773
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6045875102849563e-05,
+      "loss": 0.2594,
+      "step": 11774
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6045539267543197e-05,
+      "loss": 0.2278,
+      "step": 11775
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6045203432236832e-05,
+      "loss": 0.2386,
+      "step": 11776
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6044867596930467e-05,
+      "loss": 0.3665,
+      "step": 11777
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.60445317616241e-05,
+      "loss": 0.3248,
+      "step": 11778
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6044195926317736e-05,
+      "loss": 0.2989,
+      "step": 11779
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.604386009101137e-05,
+      "loss": 0.447,
+      "step": 11780
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6043524255705002e-05,
+      "loss": 0.3409,
+      "step": 11781
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.604318842039864e-05,
+      "loss": 0.293,
+      "step": 11782
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.604285258509227e-05,
+      "loss": 0.3618,
+      "step": 11783
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6042516749785906e-05,
+      "loss": 0.181,
+      "step": 11784
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.604218091447954e-05,
+      "loss": 0.2355,
+      "step": 11785
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6041845079173175e-05,
+      "loss": 0.3634,
+      "step": 11786
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.604150924386681e-05,
+      "loss": 0.2217,
+      "step": 11787
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6041173408560444e-05,
+      "loss": 0.2492,
+      "step": 11788
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.604083757325408e-05,
+      "loss": 0.2991,
+      "step": 11789
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.604050173794771e-05,
+      "loss": 0.3549,
+      "step": 11790
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6040165902641348e-05,
+      "loss": 0.2492,
+      "step": 11791
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.603983006733498e-05,
+      "loss": 0.3772,
+      "step": 11792
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6039494232028614e-05,
+      "loss": 0.3592,
+      "step": 11793
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.603915839672225e-05,
+      "loss": 0.2767,
+      "step": 11794
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6038822561415884e-05,
+      "loss": 0.3875,
+      "step": 11795
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6038486726109518e-05,
+      "loss": 0.3545,
+      "step": 11796
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6038150890803153e-05,
+      "loss": 0.474,
+      "step": 11797
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6037815055496787e-05,
+      "loss": 0.452,
+      "step": 11798
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.603747922019042e-05,
+      "loss": 0.3987,
+      "step": 11799
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6037143384884057e-05,
+      "loss": 0.2076,
+      "step": 11800
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6036807549577688e-05,
+      "loss": 0.2414,
+      "step": 11801
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6036471714271323e-05,
+      "loss": 0.3767,
+      "step": 11802
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6036135878964957e-05,
+      "loss": 0.2289,
+      "step": 11803
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6035800043658592e-05,
+      "loss": 0.4805,
+      "step": 11804
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6035464208352227e-05,
+      "loss": 0.4519,
+      "step": 11805
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.603512837304586e-05,
+      "loss": 0.4044,
+      "step": 11806
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6034792537739492e-05,
+      "loss": 0.3321,
+      "step": 11807
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6034456702433127e-05,
+      "loss": 0.464,
+      "step": 11808
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6034120867126762e-05,
+      "loss": 0.2865,
+      "step": 11809
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.6033785031820396e-05,
+      "loss": 0.3148,
+      "step": 11810
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.603344919651403e-05,
+      "loss": 0.2485,
+      "step": 11811
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6033113361207666e-05,
+      "loss": 0.2912,
+      "step": 11812
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.60327775259013e-05,
+      "loss": 0.2331,
+      "step": 11813
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.603244169059493e-05,
+      "loss": 0.4535,
+      "step": 11814
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.603210585528857e-05,
+      "loss": 0.2946,
+      "step": 11815
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.60317700199822e-05,
+      "loss": 0.353,
+      "step": 11816
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6031434184675836e-05,
+      "loss": 0.3216,
+      "step": 11817
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.603109834936947e-05,
+      "loss": 0.2763,
+      "step": 11818
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6030762514063105e-05,
+      "loss": 0.2276,
+      "step": 11819
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.603042667875674e-05,
+      "loss": 0.4053,
+      "step": 11820
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.603009084345037e-05,
+      "loss": 0.2919,
+      "step": 11821
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.602975500814401e-05,
+      "loss": 0.5057,
+      "step": 11822
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.602941917283764e-05,
+      "loss": 0.4978,
+      "step": 11823
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6029083337531275e-05,
+      "loss": 0.2975,
+      "step": 11824
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.602874750222491e-05,
+      "loss": 0.4132,
+      "step": 11825
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6028411666918544e-05,
+      "loss": 0.4619,
+      "step": 11826
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.602807583161218e-05,
+      "loss": 0.2801,
+      "step": 11827
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6027739996305813e-05,
+      "loss": 0.3226,
+      "step": 11828
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6027404160999448e-05,
+      "loss": 0.2873,
+      "step": 11829
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.602706832569308e-05,
+      "loss": 0.241,
+      "step": 11830
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6026732490386717e-05,
+      "loss": 0.1669,
+      "step": 11831
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.602639665508035e-05,
+      "loss": 0.2308,
+      "step": 11832
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6026060819773983e-05,
+      "loss": 0.4134,
+      "step": 11833
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6025724984467618e-05,
+      "loss": 0.2845,
+      "step": 11834
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6025389149161252e-05,
+      "loss": 0.2413,
+      "step": 11835
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6025053313854887e-05,
+      "loss": 0.216,
+      "step": 11836
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6024717478548522e-05,
+      "loss": 0.2946,
+      "step": 11837
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6024381643242156e-05,
+      "loss": 0.2329,
+      "step": 11838
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6024045807935788e-05,
+      "loss": 0.3303,
+      "step": 11839
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6023709972629426e-05,
+      "loss": 0.4117,
+      "step": 11840
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6023374137323057e-05,
+      "loss": 0.2538,
+      "step": 11841
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.602303830201669e-05,
+      "loss": 0.3197,
+      "step": 11842
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6022702466710326e-05,
+      "loss": 0.2572,
+      "step": 11843
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.602236663140396e-05,
+      "loss": 0.35,
+      "step": 11844
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6022030796097596e-05,
+      "loss": 0.2815,
+      "step": 11845
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.602169496079123e-05,
+      "loss": 0.4525,
+      "step": 11846
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6021359125484865e-05,
+      "loss": 0.2458,
+      "step": 11847
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6021023290178496e-05,
+      "loss": 0.2727,
+      "step": 11848
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6020687454872134e-05,
+      "loss": 0.4452,
+      "step": 11849
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6020351619565765e-05,
+      "loss": 0.3826,
+      "step": 11850
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.60200157842594e-05,
+      "loss": 0.233,
+      "step": 11851
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6019679948953035e-05,
+      "loss": 0.4589,
+      "step": 11852
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.601934411364667e-05,
+      "loss": 0.6166,
+      "step": 11853
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.60190082783403e-05,
+      "loss": 0.2962,
+      "step": 11854
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.601867244303394e-05,
+      "loss": 0.3333,
+      "step": 11855
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.601833660772757e-05,
+      "loss": 0.1847,
+      "step": 11856
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6018000772421205e-05,
+      "loss": 0.3802,
+      "step": 11857
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.601766493711484e-05,
+      "loss": 0.284,
+      "step": 11858
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6017329101808474e-05,
+      "loss": 0.3184,
+      "step": 11859
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.601699326650211e-05,
+      "loss": 0.2107,
+      "step": 11860
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6016657431195743e-05,
+      "loss": 0.1632,
+      "step": 11861
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6016321595889378e-05,
+      "loss": 0.2959,
+      "step": 11862
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.601598576058301e-05,
+      "loss": 0.3361,
+      "step": 11863
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6015649925276647e-05,
+      "loss": 0.4983,
+      "step": 11864
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.601531408997028e-05,
+      "loss": 0.3466,
+      "step": 11865
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6014978254663913e-05,
+      "loss": 0.2949,
+      "step": 11866
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6014642419357548e-05,
+      "loss": 0.3398,
+      "step": 11867
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6014306584051182e-05,
+      "loss": 0.5582,
+      "step": 11868
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6013970748744817e-05,
+      "loss": 0.3301,
+      "step": 11869
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.601363491343845e-05,
+      "loss": 0.2934,
+      "step": 11870
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6013299078132086e-05,
+      "loss": 0.2212,
+      "step": 11871
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6012963242825717e-05,
+      "loss": 0.3764,
+      "step": 11872
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6012627407519356e-05,
+      "loss": 0.3126,
+      "step": 11873
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6012291572212987e-05,
+      "loss": 0.4046,
+      "step": 11874
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.601195573690662e-05,
+      "loss": 0.5924,
+      "step": 11875
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6011619901600256e-05,
+      "loss": 0.4047,
+      "step": 11876
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.601128406629389e-05,
+      "loss": 0.3372,
+      "step": 11877
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6010948230987525e-05,
+      "loss": 0.3257,
+      "step": 11878
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.601061239568116e-05,
+      "loss": 0.279,
+      "step": 11879
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6010276560374795e-05,
+      "loss": 0.311,
+      "step": 11880
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6009940725068426e-05,
+      "loss": 0.2668,
+      "step": 11881
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6009604889762064e-05,
+      "loss": 0.4344,
+      "step": 11882
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6009269054455695e-05,
+      "loss": 0.149,
+      "step": 11883
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.600893321914933e-05,
+      "loss": 0.4018,
+      "step": 11884
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6008597383842965e-05,
+      "loss": 0.2473,
+      "step": 11885
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.60082615485366e-05,
+      "loss": 0.3098,
+      "step": 11886
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6007925713230234e-05,
+      "loss": 0.3243,
+      "step": 11887
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.600758987792387e-05,
+      "loss": 0.4911,
+      "step": 11888
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6007254042617503e-05,
+      "loss": 0.3792,
+      "step": 11889
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6006918207311134e-05,
+      "loss": 0.4276,
+      "step": 11890
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6006582372004772e-05,
+      "loss": 0.2701,
+      "step": 11891
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6006246536698404e-05,
+      "loss": 0.3812,
+      "step": 11892
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6005910701392038e-05,
+      "loss": 0.2872,
+      "step": 11893
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6005574866085673e-05,
+      "loss": 0.3512,
+      "step": 11894
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6005239030779308e-05,
+      "loss": 0.535,
+      "step": 11895
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6004903195472942e-05,
+      "loss": 0.3026,
+      "step": 11896
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6004567360166577e-05,
+      "loss": 0.3032,
+      "step": 11897
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6004231524860208e-05,
+      "loss": 0.4647,
+      "step": 11898
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6003895689553843e-05,
+      "loss": 0.4474,
+      "step": 11899
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6003559854247477e-05,
+      "loss": 0.2776,
+      "step": 11900
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6003224018941112e-05,
+      "loss": 0.2785,
+      "step": 11901
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6002888183634747e-05,
+      "loss": 0.3115,
+      "step": 11902
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.600255234832838e-05,
+      "loss": 0.1242,
+      "step": 11903
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6002216513022016e-05,
+      "loss": 0.2848,
+      "step": 11904
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6001880677715647e-05,
+      "loss": 0.4964,
+      "step": 11905
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6001544842409285e-05,
+      "loss": 0.4491,
+      "step": 11906
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6001209007102917e-05,
+      "loss": 0.2873,
+      "step": 11907
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.600087317179655e-05,
+      "loss": 0.2195,
+      "step": 11908
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.6000537336490186e-05,
+      "loss": 0.4413,
+      "step": 11909
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.600020150118382e-05,
+      "loss": 0.1888,
+      "step": 11910
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5999865665877455e-05,
+      "loss": 0.2991,
+      "step": 11911
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.599952983057109e-05,
+      "loss": 0.2746,
+      "step": 11912
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5999193995264724e-05,
+      "loss": 0.3327,
+      "step": 11913
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5998858159958356e-05,
+      "loss": 0.4086,
+      "step": 11914
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5998522324651994e-05,
+      "loss": 0.2431,
+      "step": 11915
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5998186489345625e-05,
+      "loss": 0.3059,
+      "step": 11916
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.599785065403926e-05,
+      "loss": 0.5542,
+      "step": 11917
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5997514818732894e-05,
+      "loss": 0.3957,
+      "step": 11918
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.599717898342653e-05,
+      "loss": 0.3836,
+      "step": 11919
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5996843148120164e-05,
+      "loss": 0.3025,
+      "step": 11920
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5996507312813798e-05,
+      "loss": 0.401,
+      "step": 11921
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5996171477507433e-05,
+      "loss": 0.2809,
+      "step": 11922
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5995835642201064e-05,
+      "loss": 0.2301,
+      "step": 11923
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5995499806894702e-05,
+      "loss": 0.46,
+      "step": 11924
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5995163971588333e-05,
+      "loss": 0.3262,
+      "step": 11925
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5994828136281968e-05,
+      "loss": 0.2295,
+      "step": 11926
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5994492300975603e-05,
+      "loss": 0.4871,
+      "step": 11927
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5994156465669237e-05,
+      "loss": 0.4262,
+      "step": 11928
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5993820630362872e-05,
+      "loss": 0.3227,
+      "step": 11929
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5993484795056507e-05,
+      "loss": 0.3886,
+      "step": 11930
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.599314895975014e-05,
+      "loss": 0.2178,
+      "step": 11931
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5992813124443773e-05,
+      "loss": 0.2835,
+      "step": 11932
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.599247728913741e-05,
+      "loss": 0.2874,
+      "step": 11933
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5992141453831042e-05,
+      "loss": 0.4188,
+      "step": 11934
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5991805618524677e-05,
+      "loss": 0.3054,
+      "step": 11935
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.599146978321831e-05,
+      "loss": 0.3305,
+      "step": 11936
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5991133947911946e-05,
+      "loss": 0.296,
+      "step": 11937
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.599079811260558e-05,
+      "loss": 0.4323,
+      "step": 11938
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5990462277299215e-05,
+      "loss": 0.3795,
+      "step": 11939
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.599012644199285e-05,
+      "loss": 0.2602,
+      "step": 11940
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.598979060668648e-05,
+      "loss": 0.3062,
+      "step": 11941
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5989454771380116e-05,
+      "loss": 0.2162,
+      "step": 11942
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.598911893607375e-05,
+      "loss": 0.2656,
+      "step": 11943
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5988783100767385e-05,
+      "loss": 0.2545,
+      "step": 11944
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.598844726546102e-05,
+      "loss": 0.3008,
+      "step": 11945
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5988111430154654e-05,
+      "loss": 0.1545,
+      "step": 11946
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5987775594848286e-05,
+      "loss": 0.4589,
+      "step": 11947
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5987439759541924e-05,
+      "loss": 0.363,
+      "step": 11948
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5987103924235555e-05,
+      "loss": 0.393,
+      "step": 11949
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.598676808892919e-05,
+      "loss": 0.509,
+      "step": 11950
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5986432253622824e-05,
+      "loss": 0.276,
+      "step": 11951
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.598609641831646e-05,
+      "loss": 0.4467,
+      "step": 11952
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5985760583010093e-05,
+      "loss": 0.6042,
+      "step": 11953
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5985424747703728e-05,
+      "loss": 0.3674,
+      "step": 11954
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5985088912397363e-05,
+      "loss": 0.2899,
+      "step": 11955
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5984753077090994e-05,
+      "loss": 0.3672,
+      "step": 11956
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5984417241784632e-05,
+      "loss": 0.4154,
+      "step": 11957
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5984081406478263e-05,
+      "loss": 0.3116,
+      "step": 11958
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5983745571171898e-05,
+      "loss": 0.303,
+      "step": 11959
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5983409735865533e-05,
+      "loss": 0.3067,
+      "step": 11960
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5983073900559167e-05,
+      "loss": 0.2982,
+      "step": 11961
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5982738065252802e-05,
+      "loss": 0.2441,
+      "step": 11962
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5982402229946437e-05,
+      "loss": 0.2642,
+      "step": 11963
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.598206639464007e-05,
+      "loss": 0.4673,
+      "step": 11964
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5981730559333702e-05,
+      "loss": 0.2202,
+      "step": 11965
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.598139472402734e-05,
+      "loss": 0.3971,
+      "step": 11966
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5981058888720972e-05,
+      "loss": 0.2312,
+      "step": 11967
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5980723053414606e-05,
+      "loss": 0.3571,
+      "step": 11968
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.598038721810824e-05,
+      "loss": 0.3245,
+      "step": 11969
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5980051382801876e-05,
+      "loss": 0.4806,
+      "step": 11970
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.597971554749551e-05,
+      "loss": 0.4446,
+      "step": 11971
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5979379712189145e-05,
+      "loss": 0.4701,
+      "step": 11972
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.597904387688278e-05,
+      "loss": 0.3467,
+      "step": 11973
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.597870804157641e-05,
+      "loss": 0.2795,
+      "step": 11974
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.597837220627005e-05,
+      "loss": 0.2259,
+      "step": 11975
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.597803637096368e-05,
+      "loss": 0.1935,
+      "step": 11976
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5977700535657315e-05,
+      "loss": 0.2023,
+      "step": 11977
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.597736470035095e-05,
+      "loss": 0.29,
+      "step": 11978
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5977028865044584e-05,
+      "loss": 0.2206,
+      "step": 11979
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.597669302973822e-05,
+      "loss": 0.3199,
+      "step": 11980
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5976357194431853e-05,
+      "loss": 0.3728,
+      "step": 11981
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5976021359125488e-05,
+      "loss": 0.2424,
+      "step": 11982
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.597568552381912e-05,
+      "loss": 0.4587,
+      "step": 11983
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5975349688512757e-05,
+      "loss": 0.2395,
+      "step": 11984
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.597501385320639e-05,
+      "loss": 0.1809,
+      "step": 11985
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5974678017900023e-05,
+      "loss": 0.3143,
+      "step": 11986
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5974342182593658e-05,
+      "loss": 0.3302,
+      "step": 11987
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5974006347287293e-05,
+      "loss": 0.2791,
+      "step": 11988
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5973670511980924e-05,
+      "loss": 0.3776,
+      "step": 11989
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5973334676674562e-05,
+      "loss": 0.4332,
+      "step": 11990
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5972998841368193e-05,
+      "loss": 0.2596,
+      "step": 11991
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5972663006061828e-05,
+      "loss": 0.3822,
+      "step": 11992
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5972327170755462e-05,
+      "loss": 0.1829,
+      "step": 11993
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5971991335449097e-05,
+      "loss": 0.2522,
+      "step": 11994
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.597165550014273e-05,
+      "loss": 0.2413,
+      "step": 11995
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5971319664836366e-05,
+      "loss": 0.2046,
+      "step": 11996
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.597098382953e-05,
+      "loss": 0.3578,
+      "step": 11997
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5970647994223632e-05,
+      "loss": 0.2926,
+      "step": 11998
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.597031215891727e-05,
+      "loss": 0.256,
+      "step": 11999
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.59699763236109e-05,
+      "loss": 0.3525,
+      "step": 12000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5969640488304536e-05,
+      "loss": 0.3156,
+      "step": 12001
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.596930465299817e-05,
+      "loss": 0.3294,
+      "step": 12002
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5968968817691805e-05,
+      "loss": 0.3372,
+      "step": 12003
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.596863298238544e-05,
+      "loss": 0.2852,
+      "step": 12004
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.596829714707907e-05,
+      "loss": 0.3749,
+      "step": 12005
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.596796131177271e-05,
+      "loss": 0.3946,
+      "step": 12006
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.596762547646634e-05,
+      "loss": 0.421,
+      "step": 12007
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.5967289641159975e-05,
+      "loss": 0.2669,
+      "step": 12008
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.596695380585361e-05,
+      "loss": 0.2241,
+      "step": 12009
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5966617970547245e-05,
+      "loss": 0.4285,
+      "step": 12010
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.596628213524088e-05,
+      "loss": 0.1977,
+      "step": 12011
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5965946299934514e-05,
+      "loss": 0.3289,
+      "step": 12012
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.596561046462815e-05,
+      "loss": 0.3679,
+      "step": 12013
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.596527462932178e-05,
+      "loss": 0.5527,
+      "step": 12014
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5964938794015418e-05,
+      "loss": 0.2897,
+      "step": 12015
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.596460295870905e-05,
+      "loss": 0.267,
+      "step": 12016
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5964267123402684e-05,
+      "loss": 0.4194,
+      "step": 12017
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.596393128809632e-05,
+      "loss": 0.3272,
+      "step": 12018
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5963595452789953e-05,
+      "loss": 0.3178,
+      "step": 12019
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5963259617483588e-05,
+      "loss": 0.4882,
+      "step": 12020
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5962923782177222e-05,
+      "loss": 0.1824,
+      "step": 12021
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5962587946870857e-05,
+      "loss": 0.3233,
+      "step": 12022
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5962252111564488e-05,
+      "loss": 0.3532,
+      "step": 12023
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5961916276258126e-05,
+      "loss": 0.449,
+      "step": 12024
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5961580440951758e-05,
+      "loss": 0.3294,
+      "step": 12025
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5961244605645392e-05,
+      "loss": 0.2787,
+      "step": 12026
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5960908770339027e-05,
+      "loss": 0.3588,
+      "step": 12027
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.596057293503266e-05,
+      "loss": 0.2076,
+      "step": 12028
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5960237099726296e-05,
+      "loss": 0.2378,
+      "step": 12029
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.595990126441993e-05,
+      "loss": 0.2718,
+      "step": 12030
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5959565429113565e-05,
+      "loss": 0.2231,
+      "step": 12031
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5959229593807197e-05,
+      "loss": 0.2405,
+      "step": 12032
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.595889375850083e-05,
+      "loss": 0.3064,
+      "step": 12033
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5958557923194466e-05,
+      "loss": 0.3569,
+      "step": 12034
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.59582220878881e-05,
+      "loss": 0.4885,
+      "step": 12035
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5957886252581735e-05,
+      "loss": 0.3714,
+      "step": 12036
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.595755041727537e-05,
+      "loss": 0.2465,
+      "step": 12037
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5957214581969e-05,
+      "loss": 0.23,
+      "step": 12038
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.595687874666264e-05,
+      "loss": 0.357,
+      "step": 12039
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.595654291135627e-05,
+      "loss": 0.2993,
+      "step": 12040
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5956207076049905e-05,
+      "loss": 0.2817,
+      "step": 12041
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.595587124074354e-05,
+      "loss": 0.2688,
+      "step": 12042
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5955535405437174e-05,
+      "loss": 0.2766,
+      "step": 12043
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.595519957013081e-05,
+      "loss": 0.4686,
+      "step": 12044
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5954863734824444e-05,
+      "loss": 0.2921,
+      "step": 12045
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.595452789951808e-05,
+      "loss": 0.3225,
+      "step": 12046
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.595419206421171e-05,
+      "loss": 0.3469,
+      "step": 12047
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5953856228905348e-05,
+      "loss": 0.4577,
+      "step": 12048
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.595352039359898e-05,
+      "loss": 0.5014,
+      "step": 12049
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5953184558292614e-05,
+      "loss": 0.3409,
+      "step": 12050
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5952848722986248e-05,
+      "loss": 0.3796,
+      "step": 12051
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5952512887679883e-05,
+      "loss": 0.2883,
+      "step": 12052
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5952177052373517e-05,
+      "loss": 0.1872,
+      "step": 12053
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5951841217067152e-05,
+      "loss": 0.4197,
+      "step": 12054
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5951505381760787e-05,
+      "loss": 0.3441,
+      "step": 12055
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5951169546454418e-05,
+      "loss": 0.4222,
+      "step": 12056
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5950833711148056e-05,
+      "loss": 0.4189,
+      "step": 12057
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5950497875841687e-05,
+      "loss": 0.3671,
+      "step": 12058
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5950162040535322e-05,
+      "loss": 0.3936,
+      "step": 12059
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5949826205228957e-05,
+      "loss": 0.3756,
+      "step": 12060
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.594949036992259e-05,
+      "loss": 0.4159,
+      "step": 12061
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5949154534616226e-05,
+      "loss": 0.2306,
+      "step": 12062
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.594881869930986e-05,
+      "loss": 0.2097,
+      "step": 12063
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5948482864003495e-05,
+      "loss": 0.4584,
+      "step": 12064
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5948147028697126e-05,
+      "loss": 0.5208,
+      "step": 12065
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5947811193390765e-05,
+      "loss": 0.2544,
+      "step": 12066
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5947475358084396e-05,
+      "loss": 0.3369,
+      "step": 12067
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.594713952277803e-05,
+      "loss": 0.3042,
+      "step": 12068
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5946803687471665e-05,
+      "loss": 0.2737,
+      "step": 12069
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.59464678521653e-05,
+      "loss": 0.4071,
+      "step": 12070
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5946132016858934e-05,
+      "loss": 0.2351,
+      "step": 12071
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.594579618155257e-05,
+      "loss": 0.3699,
+      "step": 12072
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5945460346246204e-05,
+      "loss": 0.3997,
+      "step": 12073
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5945124510939835e-05,
+      "loss": 0.1875,
+      "step": 12074
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5944788675633473e-05,
+      "loss": 0.2209,
+      "step": 12075
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5944452840327104e-05,
+      "loss": 0.3238,
+      "step": 12076
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.594411700502074e-05,
+      "loss": 0.3069,
+      "step": 12077
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5943781169714374e-05,
+      "loss": 0.3157,
+      "step": 12078
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5943445334408008e-05,
+      "loss": 0.2899,
+      "step": 12079
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.594310949910164e-05,
+      "loss": 0.3392,
+      "step": 12080
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5942773663795277e-05,
+      "loss": 0.3331,
+      "step": 12081
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.594243782848891e-05,
+      "loss": 0.293,
+      "step": 12082
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5942101993182543e-05,
+      "loss": 0.1706,
+      "step": 12083
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5941766157876178e-05,
+      "loss": 0.2405,
+      "step": 12084
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5941430322569813e-05,
+      "loss": 0.3833,
+      "step": 12085
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5941094487263447e-05,
+      "loss": 0.2269,
+      "step": 12086
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5940758651957082e-05,
+      "loss": 0.43,
+      "step": 12087
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5940422816650717e-05,
+      "loss": 0.3072,
+      "step": 12088
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5940086981344348e-05,
+      "loss": 0.3688,
+      "step": 12089
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5939751146037986e-05,
+      "loss": 0.2526,
+      "step": 12090
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5939415310731617e-05,
+      "loss": 0.2349,
+      "step": 12091
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5939079475425252e-05,
+      "loss": 0.2189,
+      "step": 12092
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5938743640118886e-05,
+      "loss": 0.6453,
+      "step": 12093
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.593840780481252e-05,
+      "loss": 0.2879,
+      "step": 12094
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5938071969506156e-05,
+      "loss": 0.4318,
+      "step": 12095
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.593773613419979e-05,
+      "loss": 0.3822,
+      "step": 12096
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5937400298893425e-05,
+      "loss": 0.2856,
+      "step": 12097
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5937064463587056e-05,
+      "loss": 0.3489,
+      "step": 12098
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5936728628280694e-05,
+      "loss": 0.2756,
+      "step": 12099
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5936392792974326e-05,
+      "loss": 0.3215,
+      "step": 12100
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.593605695766796e-05,
+      "loss": 0.2027,
+      "step": 12101
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5935721122361595e-05,
+      "loss": 0.2921,
+      "step": 12102
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.593538528705523e-05,
+      "loss": 0.3187,
+      "step": 12103
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5935049451748864e-05,
+      "loss": 0.2039,
+      "step": 12104
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.59347136164425e-05,
+      "loss": 0.2646,
+      "step": 12105
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5934377781136133e-05,
+      "loss": 0.1662,
+      "step": 12106
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5934041945829765e-05,
+      "loss": 0.261,
+      "step": 12107
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5933706110523403e-05,
+      "loss": 0.4092,
+      "step": 12108
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5933370275217034e-05,
+      "loss": 0.2733,
+      "step": 12109
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.593303443991067e-05,
+      "loss": 0.3888,
+      "step": 12110
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5932698604604303e-05,
+      "loss": 0.3072,
+      "step": 12111
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5932362769297938e-05,
+      "loss": 0.2418,
+      "step": 12112
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5932026933991573e-05,
+      "loss": 0.3859,
+      "step": 12113
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5931691098685207e-05,
+      "loss": 0.2484,
+      "step": 12114
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5931355263378842e-05,
+      "loss": 0.2776,
+      "step": 12115
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5931019428072473e-05,
+      "loss": 0.3068,
+      "step": 12116
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.593068359276611e-05,
+      "loss": 0.4599,
+      "step": 12117
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5930347757459742e-05,
+      "loss": 0.2946,
+      "step": 12118
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5930011922153377e-05,
+      "loss": 0.4046,
+      "step": 12119
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5929676086847012e-05,
+      "loss": 0.3684,
+      "step": 12120
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5929340251540646e-05,
+      "loss": 0.2587,
+      "step": 12121
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.592900441623428e-05,
+      "loss": 0.247,
+      "step": 12122
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5928668580927916e-05,
+      "loss": 0.4186,
+      "step": 12123
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5928332745621547e-05,
+      "loss": 0.338,
+      "step": 12124
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.592799691031518e-05,
+      "loss": 0.2346,
+      "step": 12125
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5927661075008816e-05,
+      "loss": 0.3788,
+      "step": 12126
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.592732523970245e-05,
+      "loss": 0.2884,
+      "step": 12127
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5926989404396086e-05,
+      "loss": 0.3081,
+      "step": 12128
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.592665356908972e-05,
+      "loss": 0.2558,
+      "step": 12129
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5926317733783355e-05,
+      "loss": 0.3048,
+      "step": 12130
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5925981898476986e-05,
+      "loss": 0.3727,
+      "step": 12131
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5925646063170624e-05,
+      "loss": 0.3431,
+      "step": 12132
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5925310227864255e-05,
+      "loss": 0.3156,
+      "step": 12133
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.592497439255789e-05,
+      "loss": 0.1965,
+      "step": 12134
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5924638557251525e-05,
+      "loss": 0.2218,
+      "step": 12135
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.592430272194516e-05,
+      "loss": 0.4256,
+      "step": 12136
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5923966886638794e-05,
+      "loss": 0.287,
+      "step": 12137
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.592363105133243e-05,
+      "loss": 0.3218,
+      "step": 12138
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5923295216026063e-05,
+      "loss": 0.165,
+      "step": 12139
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5922959380719695e-05,
+      "loss": 0.3541,
+      "step": 12140
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5922623545413333e-05,
+      "loss": 0.1752,
+      "step": 12141
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5922287710106964e-05,
+      "loss": 0.1751,
+      "step": 12142
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.59219518748006e-05,
+      "loss": 0.4594,
+      "step": 12143
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5921616039494233e-05,
+      "loss": 0.4367,
+      "step": 12144
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5921280204187868e-05,
+      "loss": 0.264,
+      "step": 12145
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5920944368881502e-05,
+      "loss": 0.3048,
+      "step": 12146
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5920608533575137e-05,
+      "loss": 0.2815,
+      "step": 12147
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5920272698268772e-05,
+      "loss": 0.256,
+      "step": 12148
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5919936862962403e-05,
+      "loss": 0.513,
+      "step": 12149
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.591960102765604e-05,
+      "loss": 0.3497,
+      "step": 12150
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5919265192349672e-05,
+      "loss": 0.4752,
+      "step": 12151
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5918929357043307e-05,
+      "loss": 0.3316,
+      "step": 12152
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.591859352173694e-05,
+      "loss": 0.3199,
+      "step": 12153
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5918257686430576e-05,
+      "loss": 0.1646,
+      "step": 12154
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.591792185112421e-05,
+      "loss": 0.3539,
+      "step": 12155
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5917586015817846e-05,
+      "loss": 0.1974,
+      "step": 12156
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.591725018051148e-05,
+      "loss": 0.2889,
+      "step": 12157
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.591691434520511e-05,
+      "loss": 0.3734,
+      "step": 12158
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.591657850989875e-05,
+      "loss": 0.4988,
+      "step": 12159
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.591624267459238e-05,
+      "loss": 0.4752,
+      "step": 12160
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5915906839286015e-05,
+      "loss": 0.2607,
+      "step": 12161
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.591557100397965e-05,
+      "loss": 0.3299,
+      "step": 12162
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5915235168673285e-05,
+      "loss": 0.3009,
+      "step": 12163
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.591489933336692e-05,
+      "loss": 0.3376,
+      "step": 12164
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5914563498060554e-05,
+      "loss": 0.3406,
+      "step": 12165
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.591422766275419e-05,
+      "loss": 0.2795,
+      "step": 12166
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.591389182744782e-05,
+      "loss": 0.2686,
+      "step": 12167
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5913555992141455e-05,
+      "loss": 0.3351,
+      "step": 12168
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.591322015683509e-05,
+      "loss": 0.4265,
+      "step": 12169
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5912884321528724e-05,
+      "loss": 0.508,
+      "step": 12170
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.591254848622236e-05,
+      "loss": 0.3008,
+      "step": 12171
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5912212650915993e-05,
+      "loss": 0.323,
+      "step": 12172
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5911876815609624e-05,
+      "loss": 0.3844,
+      "step": 12173
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5911540980303262e-05,
+      "loss": 0.339,
+      "step": 12174
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5911205144996894e-05,
+      "loss": 0.2087,
+      "step": 12175
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5910869309690528e-05,
+      "loss": 0.432,
+      "step": 12176
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5910533474384163e-05,
+      "loss": 0.4796,
+      "step": 12177
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5910197639077798e-05,
+      "loss": 0.2008,
+      "step": 12178
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5909861803771432e-05,
+      "loss": 0.3937,
+      "step": 12179
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5909525968465067e-05,
+      "loss": 0.3514,
+      "step": 12180
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.59091901331587e-05,
+      "loss": 0.3111,
+      "step": 12181
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5908854297852333e-05,
+      "loss": 0.247,
+      "step": 12182
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5908518462545967e-05,
+      "loss": 0.3115,
+      "step": 12183
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5908182627239602e-05,
+      "loss": 0.4654,
+      "step": 12184
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5907846791933237e-05,
+      "loss": 0.3275,
+      "step": 12185
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.590751095662687e-05,
+      "loss": 0.3665,
+      "step": 12186
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5907175121320506e-05,
+      "loss": 0.3119,
+      "step": 12187
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.590683928601414e-05,
+      "loss": 0.3958,
+      "step": 12188
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5906503450707772e-05,
+      "loss": 0.3735,
+      "step": 12189
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.590616761540141e-05,
+      "loss": 0.351,
+      "step": 12190
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.590583178009504e-05,
+      "loss": 0.2625,
+      "step": 12191
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5905495944788676e-05,
+      "loss": 0.4269,
+      "step": 12192
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.590516010948231e-05,
+      "loss": 0.2263,
+      "step": 12193
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5904824274175945e-05,
+      "loss": 0.3666,
+      "step": 12194
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.590448843886958e-05,
+      "loss": 0.4215,
+      "step": 12195
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5904152603563214e-05,
+      "loss": 0.1948,
+      "step": 12196
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.590381676825685e-05,
+      "loss": 0.3246,
+      "step": 12197
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.590348093295048e-05,
+      "loss": 0.3748,
+      "step": 12198
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.590314509764412e-05,
+      "loss": 0.3477,
+      "step": 12199
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.590280926233775e-05,
+      "loss": 0.2492,
+      "step": 12200
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5902473427031384e-05,
+      "loss": 0.3953,
+      "step": 12201
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.590213759172502e-05,
+      "loss": 0.2787,
+      "step": 12202
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5901801756418654e-05,
+      "loss": 0.2289,
+      "step": 12203
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5901465921112288e-05,
+      "loss": 0.2494,
+      "step": 12204
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5901130085805923e-05,
+      "loss": 0.3105,
+      "step": 12205
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5900794250499558e-05,
+      "loss": 0.4571,
+      "step": 12206
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.590045841519319e-05,
+      "loss": 0.2377,
+      "step": 12207
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.5900122579886827e-05,
+      "loss": 0.3515,
+      "step": 12208
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5899786744580458e-05,
+      "loss": 0.2625,
+      "step": 12209
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5899450909274093e-05,
+      "loss": 0.2968,
+      "step": 12210
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5899115073967727e-05,
+      "loss": 0.3652,
+      "step": 12211
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5898779238661362e-05,
+      "loss": 0.3733,
+      "step": 12212
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5898443403354997e-05,
+      "loss": 0.3654,
+      "step": 12213
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.589810756804863e-05,
+      "loss": 0.3528,
+      "step": 12214
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5897771732742263e-05,
+      "loss": 0.2658,
+      "step": 12215
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5897435897435897e-05,
+      "loss": 0.4366,
+      "step": 12216
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5897100062129532e-05,
+      "loss": 0.245,
+      "step": 12217
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5896764226823167e-05,
+      "loss": 0.242,
+      "step": 12218
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.58964283915168e-05,
+      "loss": 0.3705,
+      "step": 12219
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5896092556210436e-05,
+      "loss": 0.3178,
+      "step": 12220
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.589575672090407e-05,
+      "loss": 0.1873,
+      "step": 12221
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5895420885597702e-05,
+      "loss": 0.3036,
+      "step": 12222
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.589508505029134e-05,
+      "loss": 0.2223,
+      "step": 12223
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.589474921498497e-05,
+      "loss": 0.2999,
+      "step": 12224
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5894413379678606e-05,
+      "loss": 0.222,
+      "step": 12225
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.589407754437224e-05,
+      "loss": 0.2547,
+      "step": 12226
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5893741709065875e-05,
+      "loss": 0.3306,
+      "step": 12227
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.589340587375951e-05,
+      "loss": 0.5447,
+      "step": 12228
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5893070038453144e-05,
+      "loss": 0.1994,
+      "step": 12229
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.589273420314678e-05,
+      "loss": 0.2565,
+      "step": 12230
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.589239836784041e-05,
+      "loss": 0.2921,
+      "step": 12231
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5892062532534048e-05,
+      "loss": 0.2405,
+      "step": 12232
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.589172669722768e-05,
+      "loss": 0.2761,
+      "step": 12233
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5891390861921314e-05,
+      "loss": 0.3735,
+      "step": 12234
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.589105502661495e-05,
+      "loss": 0.3134,
+      "step": 12235
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5890719191308583e-05,
+      "loss": 0.3845,
+      "step": 12236
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5890383356002218e-05,
+      "loss": 0.3158,
+      "step": 12237
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5890047520695853e-05,
+      "loss": 0.5188,
+      "step": 12238
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5889711685389487e-05,
+      "loss": 0.2609,
+      "step": 12239
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.588937585008312e-05,
+      "loss": 0.3019,
+      "step": 12240
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5889040014776757e-05,
+      "loss": 0.2332,
+      "step": 12241
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5888704179470388e-05,
+      "loss": 0.4378,
+      "step": 12242
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5888368344164023e-05,
+      "loss": 0.3692,
+      "step": 12243
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5888032508857657e-05,
+      "loss": 0.3223,
+      "step": 12244
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5887696673551292e-05,
+      "loss": 0.1831,
+      "step": 12245
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5887360838244927e-05,
+      "loss": 0.3359,
+      "step": 12246
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.588702500293856e-05,
+      "loss": 0.2381,
+      "step": 12247
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5886689167632196e-05,
+      "loss": 0.3037,
+      "step": 12248
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5886353332325827e-05,
+      "loss": 0.223,
+      "step": 12249
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5886017497019465e-05,
+      "loss": 0.3849,
+      "step": 12250
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5885681661713096e-05,
+      "loss": 0.3831,
+      "step": 12251
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.588534582640673e-05,
+      "loss": 0.2189,
+      "step": 12252
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5885009991100366e-05,
+      "loss": 0.3903,
+      "step": 12253
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5884674155794e-05,
+      "loss": 0.402,
+      "step": 12254
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5884338320487635e-05,
+      "loss": 0.3164,
+      "step": 12255
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.588400248518127e-05,
+      "loss": 0.2393,
+      "step": 12256
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5883666649874904e-05,
+      "loss": 0.2625,
+      "step": 12257
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5883330814568536e-05,
+      "loss": 0.2491,
+      "step": 12258
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.588299497926217e-05,
+      "loss": 0.3377,
+      "step": 12259
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5882659143955805e-05,
+      "loss": 0.2918,
+      "step": 12260
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.588232330864944e-05,
+      "loss": 0.3717,
+      "step": 12261
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5881987473343074e-05,
+      "loss": 0.3081,
+      "step": 12262
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.588165163803671e-05,
+      "loss": 0.2763,
+      "step": 12263
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.588131580273034e-05,
+      "loss": 0.4386,
+      "step": 12264
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5880979967423978e-05,
+      "loss": 0.234,
+      "step": 12265
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.588064413211761e-05,
+      "loss": 0.3266,
+      "step": 12266
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5880308296811244e-05,
+      "loss": 0.4221,
+      "step": 12267
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.587997246150488e-05,
+      "loss": 0.431,
+      "step": 12268
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5879636626198513e-05,
+      "loss": 0.2571,
+      "step": 12269
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5879300790892148e-05,
+      "loss": 0.153,
+      "step": 12270
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5878964955585783e-05,
+      "loss": 0.3923,
+      "step": 12271
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5878629120279417e-05,
+      "loss": 0.3856,
+      "step": 12272
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.587829328497305e-05,
+      "loss": 0.3235,
+      "step": 12273
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5877957449666686e-05,
+      "loss": 0.274,
+      "step": 12274
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5877621614360318e-05,
+      "loss": 0.4321,
+      "step": 12275
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5877285779053952e-05,
+      "loss": 0.2614,
+      "step": 12276
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5876949943747587e-05,
+      "loss": 0.3119,
+      "step": 12277
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.587661410844122e-05,
+      "loss": 0.3618,
+      "step": 12278
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5876278273134856e-05,
+      "loss": 0.2501,
+      "step": 12279
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.587594243782849e-05,
+      "loss": 0.1196,
+      "step": 12280
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5875606602522126e-05,
+      "loss": 0.3662,
+      "step": 12281
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5875270767215757e-05,
+      "loss": 0.3326,
+      "step": 12282
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5874934931909395e-05,
+      "loss": 0.2569,
+      "step": 12283
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5874599096603026e-05,
+      "loss": 0.3087,
+      "step": 12284
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.587426326129666e-05,
+      "loss": 0.329,
+      "step": 12285
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5873927425990295e-05,
+      "loss": 0.2367,
+      "step": 12286
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.587359159068393e-05,
+      "loss": 0.4298,
+      "step": 12287
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5873255755377565e-05,
+      "loss": 0.4228,
+      "step": 12288
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.58729199200712e-05,
+      "loss": 0.2548,
+      "step": 12289
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5872584084764834e-05,
+      "loss": 0.3327,
+      "step": 12290
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5872248249458465e-05,
+      "loss": 0.3071,
+      "step": 12291
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5871912414152103e-05,
+      "loss": 0.2867,
+      "step": 12292
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5871576578845735e-05,
+      "loss": 0.3543,
+      "step": 12293
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.587124074353937e-05,
+      "loss": 0.5882,
+      "step": 12294
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5870904908233004e-05,
+      "loss": 0.2901,
+      "step": 12295
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.587056907292664e-05,
+      "loss": 0.2629,
+      "step": 12296
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5870233237620273e-05,
+      "loss": 0.2366,
+      "step": 12297
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5869897402313908e-05,
+      "loss": 0.2253,
+      "step": 12298
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5869561567007542e-05,
+      "loss": 0.4942,
+      "step": 12299
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5869225731701174e-05,
+      "loss": 0.3016,
+      "step": 12300
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5868889896394812e-05,
+      "loss": 0.5103,
+      "step": 12301
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5868554061088443e-05,
+      "loss": 0.3277,
+      "step": 12302
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5868218225782078e-05,
+      "loss": 0.2575,
+      "step": 12303
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5867882390475712e-05,
+      "loss": 0.3372,
+      "step": 12304
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5867546555169347e-05,
+      "loss": 0.2848,
+      "step": 12305
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5867210719862978e-05,
+      "loss": 0.3467,
+      "step": 12306
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5866874884556616e-05,
+      "loss": 0.2939,
+      "step": 12307
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5866539049250248e-05,
+      "loss": 0.3257,
+      "step": 12308
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5866203213943882e-05,
+      "loss": 0.2497,
+      "step": 12309
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5865867378637517e-05,
+      "loss": 0.4745,
+      "step": 12310
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.586553154333115e-05,
+      "loss": 0.381,
+      "step": 12311
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5865195708024786e-05,
+      "loss": 0.1868,
+      "step": 12312
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.586485987271842e-05,
+      "loss": 0.3259,
+      "step": 12313
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5864524037412055e-05,
+      "loss": 0.3563,
+      "step": 12314
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5864188202105687e-05,
+      "loss": 0.3189,
+      "step": 12315
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5863852366799325e-05,
+      "loss": 0.4331,
+      "step": 12316
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5863516531492956e-05,
+      "loss": 0.3478,
+      "step": 12317
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.586318069618659e-05,
+      "loss": 0.5258,
+      "step": 12318
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5862844860880225e-05,
+      "loss": 0.3306,
+      "step": 12319
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.586250902557386e-05,
+      "loss": 0.18,
+      "step": 12320
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5862173190267495e-05,
+      "loss": 0.2649,
+      "step": 12321
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.586183735496113e-05,
+      "loss": 0.4255,
+      "step": 12322
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5861501519654764e-05,
+      "loss": 0.2153,
+      "step": 12323
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5861165684348395e-05,
+      "loss": 0.311,
+      "step": 12324
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5860829849042033e-05,
+      "loss": 0.295,
+      "step": 12325
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5860494013735664e-05,
+      "loss": 0.2297,
+      "step": 12326
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.58601581784293e-05,
+      "loss": 0.4266,
+      "step": 12327
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5859822343122934e-05,
+      "loss": 0.4124,
+      "step": 12328
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.585948650781657e-05,
+      "loss": 0.2886,
+      "step": 12329
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5859150672510203e-05,
+      "loss": 0.3122,
+      "step": 12330
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5858814837203838e-05,
+      "loss": 0.3075,
+      "step": 12331
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5858479001897472e-05,
+      "loss": 0.3671,
+      "step": 12332
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5858143166591104e-05,
+      "loss": 0.4317,
+      "step": 12333
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.585780733128474e-05,
+      "loss": 0.2128,
+      "step": 12334
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5857471495978373e-05,
+      "loss": 0.4204,
+      "step": 12335
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5857135660672008e-05,
+      "loss": 0.3425,
+      "step": 12336
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5856799825365642e-05,
+      "loss": 0.1968,
+      "step": 12337
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5856463990059277e-05,
+      "loss": 0.1592,
+      "step": 12338
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.585612815475291e-05,
+      "loss": 0.27,
+      "step": 12339
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5855792319446546e-05,
+      "loss": 0.3301,
+      "step": 12340
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.585545648414018e-05,
+      "loss": 0.3779,
+      "step": 12341
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5855120648833812e-05,
+      "loss": 0.1885,
+      "step": 12342
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.585478481352745e-05,
+      "loss": 0.2437,
+      "step": 12343
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.585444897822108e-05,
+      "loss": 0.6793,
+      "step": 12344
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5854113142914716e-05,
+      "loss": 0.3122,
+      "step": 12345
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.585377730760835e-05,
+      "loss": 0.2619,
+      "step": 12346
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5853441472301985e-05,
+      "loss": 0.473,
+      "step": 12347
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.585310563699562e-05,
+      "loss": 0.2909,
+      "step": 12348
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5852769801689255e-05,
+      "loss": 0.3333,
+      "step": 12349
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5852433966382886e-05,
+      "loss": 0.395,
+      "step": 12350
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.585209813107652e-05,
+      "loss": 0.3339,
+      "step": 12351
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5851762295770155e-05,
+      "loss": 0.3882,
+      "step": 12352
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.585142646046379e-05,
+      "loss": 0.239,
+      "step": 12353
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5851090625157424e-05,
+      "loss": 0.4355,
+      "step": 12354
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.585075478985106e-05,
+      "loss": 0.6348,
+      "step": 12355
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5850418954544694e-05,
+      "loss": 0.4298,
+      "step": 12356
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5850083119238325e-05,
+      "loss": 0.47,
+      "step": 12357
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5849747283931963e-05,
+      "loss": 0.2743,
+      "step": 12358
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5849411448625594e-05,
+      "loss": 0.2398,
+      "step": 12359
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.584907561331923e-05,
+      "loss": 0.3783,
+      "step": 12360
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5848739778012864e-05,
+      "loss": 0.5209,
+      "step": 12361
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5848403942706498e-05,
+      "loss": 0.2694,
+      "step": 12362
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5848068107400133e-05,
+      "loss": 0.3366,
+      "step": 12363
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5847732272093764e-05,
+      "loss": 0.16,
+      "step": 12364
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5847396436787402e-05,
+      "loss": 0.3171,
+      "step": 12365
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5847060601481033e-05,
+      "loss": 0.3212,
+      "step": 12366
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5846724766174668e-05,
+      "loss": 0.4504,
+      "step": 12367
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5846388930868303e-05,
+      "loss": 0.3304,
+      "step": 12368
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5846053095561937e-05,
+      "loss": 0.5863,
+      "step": 12369
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5845717260255572e-05,
+      "loss": 0.2605,
+      "step": 12370
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5845381424949207e-05,
+      "loss": 0.3755,
+      "step": 12371
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.584504558964284e-05,
+      "loss": 0.3176,
+      "step": 12372
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5844709754336473e-05,
+      "loss": 0.3567,
+      "step": 12373
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.584437391903011e-05,
+      "loss": 0.2306,
+      "step": 12374
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5844038083723742e-05,
+      "loss": 0.2414,
+      "step": 12375
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5843702248417376e-05,
+      "loss": 0.2868,
+      "step": 12376
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.584336641311101e-05,
+      "loss": 0.4714,
+      "step": 12377
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5843030577804646e-05,
+      "loss": 0.191,
+      "step": 12378
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.584269474249828e-05,
+      "loss": 0.4552,
+      "step": 12379
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5842358907191915e-05,
+      "loss": 0.1751,
+      "step": 12380
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.584202307188555e-05,
+      "loss": 0.4911,
+      "step": 12381
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.584168723657918e-05,
+      "loss": 0.2461,
+      "step": 12382
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.584135140127282e-05,
+      "loss": 0.352,
+      "step": 12383
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.584101556596645e-05,
+      "loss": 0.2434,
+      "step": 12384
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5840679730660085e-05,
+      "loss": 0.1998,
+      "step": 12385
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.584034389535372e-05,
+      "loss": 0.3494,
+      "step": 12386
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5840008060047354e-05,
+      "loss": 0.4699,
+      "step": 12387
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.583967222474099e-05,
+      "loss": 0.4668,
+      "step": 12388
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5839336389434623e-05,
+      "loss": 0.3316,
+      "step": 12389
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5839000554128258e-05,
+      "loss": 0.3596,
+      "step": 12390
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.583866471882189e-05,
+      "loss": 0.5476,
+      "step": 12391
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5838328883515527e-05,
+      "loss": 0.2537,
+      "step": 12392
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.583799304820916e-05,
+      "loss": 0.2643,
+      "step": 12393
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5837657212902793e-05,
+      "loss": 0.3839,
+      "step": 12394
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5837321377596428e-05,
+      "loss": 0.3311,
+      "step": 12395
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5836985542290063e-05,
+      "loss": 0.3082,
+      "step": 12396
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5836649706983694e-05,
+      "loss": 0.3989,
+      "step": 12397
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5836313871677332e-05,
+      "loss": 0.2399,
+      "step": 12398
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5835978036370963e-05,
+      "loss": 0.4395,
+      "step": 12399
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5835642201064598e-05,
+      "loss": 0.5334,
+      "step": 12400
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5835306365758232e-05,
+      "loss": 0.2834,
+      "step": 12401
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5834970530451867e-05,
+      "loss": 0.3867,
+      "step": 12402
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5834634695145502e-05,
+      "loss": 0.4236,
+      "step": 12403
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5834298859839136e-05,
+      "loss": 0.2829,
+      "step": 12404
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.583396302453277e-05,
+      "loss": 0.3633,
+      "step": 12405
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.5833627189226402e-05,
+      "loss": 0.2042,
+      "step": 12406
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.583329135392004e-05,
+      "loss": 0.2743,
+      "step": 12407
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.583295551861367e-05,
+      "loss": 0.3905,
+      "step": 12408
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5832619683307306e-05,
+      "loss": 0.3853,
+      "step": 12409
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.583228384800094e-05,
+      "loss": 0.301,
+      "step": 12410
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5831948012694576e-05,
+      "loss": 0.2351,
+      "step": 12411
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.583161217738821e-05,
+      "loss": 0.303,
+      "step": 12412
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5831276342081845e-05,
+      "loss": 0.2064,
+      "step": 12413
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.583094050677548e-05,
+      "loss": 0.418,
+      "step": 12414
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.583060467146911e-05,
+      "loss": 0.3913,
+      "step": 12415
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.583026883616275e-05,
+      "loss": 0.2231,
+      "step": 12416
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.582993300085638e-05,
+      "loss": 0.313,
+      "step": 12417
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5829597165550015e-05,
+      "loss": 0.2952,
+      "step": 12418
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.582926133024365e-05,
+      "loss": 0.2801,
+      "step": 12419
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5828925494937284e-05,
+      "loss": 0.2748,
+      "step": 12420
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.582858965963092e-05,
+      "loss": 0.406,
+      "step": 12421
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5828253824324553e-05,
+      "loss": 0.2353,
+      "step": 12422
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5827917989018188e-05,
+      "loss": 0.3993,
+      "step": 12423
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.582758215371182e-05,
+      "loss": 0.3092,
+      "step": 12424
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5827246318405457e-05,
+      "loss": 0.3258,
+      "step": 12425
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.582691048309909e-05,
+      "loss": 0.4069,
+      "step": 12426
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5826574647792723e-05,
+      "loss": 0.2681,
+      "step": 12427
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5826238812486358e-05,
+      "loss": 0.2799,
+      "step": 12428
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5825902977179992e-05,
+      "loss": 0.2471,
+      "step": 12429
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5825567141873627e-05,
+      "loss": 0.37,
+      "step": 12430
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5825231306567262e-05,
+      "loss": 0.288,
+      "step": 12431
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5824895471260896e-05,
+      "loss": 0.2389,
+      "step": 12432
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5824559635954528e-05,
+      "loss": 0.1977,
+      "step": 12433
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5824223800648166e-05,
+      "loss": 0.3739,
+      "step": 12434
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5823887965341797e-05,
+      "loss": 0.2486,
+      "step": 12435
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.582355213003543e-05,
+      "loss": 0.3761,
+      "step": 12436
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5823216294729066e-05,
+      "loss": 0.3674,
+      "step": 12437
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.58228804594227e-05,
+      "loss": 0.3445,
+      "step": 12438
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5822544624116336e-05,
+      "loss": 0.2122,
+      "step": 12439
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.582220878880997e-05,
+      "loss": 0.1898,
+      "step": 12440
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.58218729535036e-05,
+      "loss": 0.3012,
+      "step": 12441
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5821537118197236e-05,
+      "loss": 0.2366,
+      "step": 12442
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.582120128289087e-05,
+      "loss": 0.2036,
+      "step": 12443
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5820865447584505e-05,
+      "loss": 0.2296,
+      "step": 12444
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.582052961227814e-05,
+      "loss": 0.5755,
+      "step": 12445
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5820193776971775e-05,
+      "loss": 0.4784,
+      "step": 12446
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.581985794166541e-05,
+      "loss": 0.198,
+      "step": 12447
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.581952210635904e-05,
+      "loss": 0.41,
+      "step": 12448
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.581918627105268e-05,
+      "loss": 0.2808,
+      "step": 12449
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.581885043574631e-05,
+      "loss": 0.4496,
+      "step": 12450
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5818514600439945e-05,
+      "loss": 0.2078,
+      "step": 12451
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.581817876513358e-05,
+      "loss": 0.4355,
+      "step": 12452
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5817842929827214e-05,
+      "loss": 0.3378,
+      "step": 12453
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.581750709452085e-05,
+      "loss": 0.4153,
+      "step": 12454
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5817171259214483e-05,
+      "loss": 0.234,
+      "step": 12455
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5816835423908118e-05,
+      "loss": 0.3334,
+      "step": 12456
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.581649958860175e-05,
+      "loss": 0.2629,
+      "step": 12457
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5816163753295387e-05,
+      "loss": 0.5334,
+      "step": 12458
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.581582791798902e-05,
+      "loss": 0.5397,
+      "step": 12459
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5815492082682653e-05,
+      "loss": 0.1811,
+      "step": 12460
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5815156247376288e-05,
+      "loss": 0.3103,
+      "step": 12461
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5814820412069922e-05,
+      "loss": 0.2587,
+      "step": 12462
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5814484576763557e-05,
+      "loss": 0.4003,
+      "step": 12463
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.581414874145719e-05,
+      "loss": 0.3015,
+      "step": 12464
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5813812906150826e-05,
+      "loss": 0.3791,
+      "step": 12465
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5813477070844457e-05,
+      "loss": 0.2717,
+      "step": 12466
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5813141235538095e-05,
+      "loss": 0.1464,
+      "step": 12467
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5812805400231727e-05,
+      "loss": 0.3837,
+      "step": 12468
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.581246956492536e-05,
+      "loss": 0.4656,
+      "step": 12469
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5812133729618996e-05,
+      "loss": 0.1455,
+      "step": 12470
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.581179789431263e-05,
+      "loss": 0.3088,
+      "step": 12471
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5811462059006265e-05,
+      "loss": 0.3725,
+      "step": 12472
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.58111262236999e-05,
+      "loss": 0.3212,
+      "step": 12473
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5810790388393535e-05,
+      "loss": 0.2221,
+      "step": 12474
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5810454553087166e-05,
+      "loss": 0.2763,
+      "step": 12475
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5810118717780804e-05,
+      "loss": 0.2067,
+      "step": 12476
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5809782882474435e-05,
+      "loss": 0.2528,
+      "step": 12477
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.580944704716807e-05,
+      "loss": 0.3667,
+      "step": 12478
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5809111211861704e-05,
+      "loss": 0.349,
+      "step": 12479
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.580877537655534e-05,
+      "loss": 0.2305,
+      "step": 12480
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5808439541248974e-05,
+      "loss": 0.334,
+      "step": 12481
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.580810370594261e-05,
+      "loss": 0.4103,
+      "step": 12482
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5807767870636243e-05,
+      "loss": 0.2578,
+      "step": 12483
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5807432035329874e-05,
+      "loss": 0.4325,
+      "step": 12484
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.580709620002351e-05,
+      "loss": 0.3055,
+      "step": 12485
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5806760364717144e-05,
+      "loss": 0.3059,
+      "step": 12486
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5806424529410778e-05,
+      "loss": 0.2772,
+      "step": 12487
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5806088694104413e-05,
+      "loss": 0.328,
+      "step": 12488
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5805752858798048e-05,
+      "loss": 0.2316,
+      "step": 12489
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.580541702349168e-05,
+      "loss": 0.2993,
+      "step": 12490
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5805081188185317e-05,
+      "loss": 0.4249,
+      "step": 12491
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5804745352878948e-05,
+      "loss": 0.376,
+      "step": 12492
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5804409517572583e-05,
+      "loss": 0.2586,
+      "step": 12493
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5804073682266217e-05,
+      "loss": 0.5363,
+      "step": 12494
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5803737846959852e-05,
+      "loss": 0.2644,
+      "step": 12495
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5803402011653487e-05,
+      "loss": 0.3183,
+      "step": 12496
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.580306617634712e-05,
+      "loss": 0.3195,
+      "step": 12497
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5802730341040756e-05,
+      "loss": 0.3097,
+      "step": 12498
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5802394505734387e-05,
+      "loss": 0.426,
+      "step": 12499
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5802058670428025e-05,
+      "loss": 0.3693,
+      "step": 12500
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5801722835121657e-05,
+      "loss": 0.2563,
+      "step": 12501
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.580138699981529e-05,
+      "loss": 0.2658,
+      "step": 12502
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5801051164508926e-05,
+      "loss": 0.248,
+      "step": 12503
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.580071532920256e-05,
+      "loss": 0.2175,
+      "step": 12504
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5800379493896195e-05,
+      "loss": 0.2546,
+      "step": 12505
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.580004365858983e-05,
+      "loss": 0.3067,
+      "step": 12506
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5799707823283464e-05,
+      "loss": 0.1848,
+      "step": 12507
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5799371987977096e-05,
+      "loss": 0.2999,
+      "step": 12508
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5799036152670734e-05,
+      "loss": 0.2903,
+      "step": 12509
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5798700317364365e-05,
+      "loss": 0.3235,
+      "step": 12510
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5798364482058e-05,
+      "loss": 0.4453,
+      "step": 12511
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5798028646751634e-05,
+      "loss": 0.5026,
+      "step": 12512
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.579769281144527e-05,
+      "loss": 0.3179,
+      "step": 12513
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5797356976138904e-05,
+      "loss": 0.4041,
+      "step": 12514
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5797021140832538e-05,
+      "loss": 0.3301,
+      "step": 12515
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5796685305526173e-05,
+      "loss": 0.2668,
+      "step": 12516
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5796349470219804e-05,
+      "loss": 0.4303,
+      "step": 12517
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5796013634913442e-05,
+      "loss": 0.1547,
+      "step": 12518
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5795677799607073e-05,
+      "loss": 0.2994,
+      "step": 12519
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5795341964300708e-05,
+      "loss": 0.2296,
+      "step": 12520
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5795006128994343e-05,
+      "loss": 0.3341,
+      "step": 12521
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5794670293687977e-05,
+      "loss": 0.396,
+      "step": 12522
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5794334458381612e-05,
+      "loss": 0.2736,
+      "step": 12523
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5793998623075247e-05,
+      "loss": 0.3536,
+      "step": 12524
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.579366278776888e-05,
+      "loss": 0.3448,
+      "step": 12525
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5793326952462513e-05,
+      "loss": 0.4535,
+      "step": 12526
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.579299111715615e-05,
+      "loss": 0.2539,
+      "step": 12527
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5792655281849782e-05,
+      "loss": 0.2696,
+      "step": 12528
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5792319446543417e-05,
+      "loss": 0.3502,
+      "step": 12529
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.579198361123705e-05,
+      "loss": 0.4767,
+      "step": 12530
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5791647775930686e-05,
+      "loss": 0.4987,
+      "step": 12531
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5791311940624317e-05,
+      "loss": 0.2124,
+      "step": 12532
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5790976105317955e-05,
+      "loss": 0.2948,
+      "step": 12533
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5790640270011586e-05,
+      "loss": 0.3724,
+      "step": 12534
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.579030443470522e-05,
+      "loss": 0.3395,
+      "step": 12535
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5789968599398856e-05,
+      "loss": 0.1883,
+      "step": 12536
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.578963276409249e-05,
+      "loss": 0.3664,
+      "step": 12537
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5789296928786125e-05,
+      "loss": 0.3137,
+      "step": 12538
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.578896109347976e-05,
+      "loss": 0.2542,
+      "step": 12539
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5788625258173394e-05,
+      "loss": 0.2531,
+      "step": 12540
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5788289422867026e-05,
+      "loss": 0.4336,
+      "step": 12541
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5787953587560664e-05,
+      "loss": 0.197,
+      "step": 12542
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5787617752254295e-05,
+      "loss": 0.3763,
+      "step": 12543
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.578728191694793e-05,
+      "loss": 0.5042,
+      "step": 12544
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5786946081641564e-05,
+      "loss": 0.4025,
+      "step": 12545
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.57866102463352e-05,
+      "loss": 0.325,
+      "step": 12546
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5786274411028833e-05,
+      "loss": 0.2574,
+      "step": 12547
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5785938575722465e-05,
+      "loss": 0.2501,
+      "step": 12548
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5785602740416103e-05,
+      "loss": 0.2896,
+      "step": 12549
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5785266905109734e-05,
+      "loss": 0.2782,
+      "step": 12550
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.578493106980337e-05,
+      "loss": 0.208,
+      "step": 12551
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5784595234497003e-05,
+      "loss": 0.4327,
+      "step": 12552
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5784259399190638e-05,
+      "loss": 0.4588,
+      "step": 12553
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5783923563884273e-05,
+      "loss": 0.2457,
+      "step": 12554
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5783587728577907e-05,
+      "loss": 0.3213,
+      "step": 12555
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5783251893271542e-05,
+      "loss": 0.3965,
+      "step": 12556
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5782916057965173e-05,
+      "loss": 0.2528,
+      "step": 12557
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.578258022265881e-05,
+      "loss": 0.3114,
+      "step": 12558
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5782244387352442e-05,
+      "loss": 0.319,
+      "step": 12559
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5781908552046077e-05,
+      "loss": 0.3397,
+      "step": 12560
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.578157271673971e-05,
+      "loss": 0.2598,
+      "step": 12561
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5781236881433346e-05,
+      "loss": 0.415,
+      "step": 12562
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.578090104612698e-05,
+      "loss": 0.3432,
+      "step": 12563
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5780565210820616e-05,
+      "loss": 0.3221,
+      "step": 12564
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.578022937551425e-05,
+      "loss": 0.3103,
+      "step": 12565
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.577989354020788e-05,
+      "loss": 0.3677,
+      "step": 12566
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.577955770490152e-05,
+      "loss": 0.4212,
+      "step": 12567
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.577922186959515e-05,
+      "loss": 0.2801,
+      "step": 12568
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5778886034288785e-05,
+      "loss": 0.2642,
+      "step": 12569
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.577855019898242e-05,
+      "loss": 0.3847,
+      "step": 12570
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5778214363676055e-05,
+      "loss": 0.3461,
+      "step": 12571
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.577787852836969e-05,
+      "loss": 0.1956,
+      "step": 12572
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5777542693063324e-05,
+      "loss": 0.3508,
+      "step": 12573
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.577720685775696e-05,
+      "loss": 0.268,
+      "step": 12574
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.577687102245059e-05,
+      "loss": 0.2428,
+      "step": 12575
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5776535187144225e-05,
+      "loss": 0.4831,
+      "step": 12576
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.577619935183786e-05,
+      "loss": 0.4606,
+      "step": 12577
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5775863516531494e-05,
+      "loss": 0.3319,
+      "step": 12578
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.577552768122513e-05,
+      "loss": 0.2953,
+      "step": 12579
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5775191845918763e-05,
+      "loss": 0.3691,
+      "step": 12580
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5774856010612394e-05,
+      "loss": 0.2326,
+      "step": 12581
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5774520175306033e-05,
+      "loss": 0.2731,
+      "step": 12582
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5774184339999664e-05,
+      "loss": 0.3072,
+      "step": 12583
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.57738485046933e-05,
+      "loss": 0.3188,
+      "step": 12584
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5773512669386933e-05,
+      "loss": 0.3306,
+      "step": 12585
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5773176834080568e-05,
+      "loss": 0.2163,
+      "step": 12586
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5772840998774202e-05,
+      "loss": 0.5333,
+      "step": 12587
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5772505163467837e-05,
+      "loss": 0.1489,
+      "step": 12588
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.577216932816147e-05,
+      "loss": 0.2633,
+      "step": 12589
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5771833492855103e-05,
+      "loss": 0.3627,
+      "step": 12590
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.577149765754874e-05,
+      "loss": 0.3315,
+      "step": 12591
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5771161822242372e-05,
+      "loss": 0.3382,
+      "step": 12592
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5770825986936007e-05,
+      "loss": 0.3379,
+      "step": 12593
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.577049015162964e-05,
+      "loss": 0.2408,
+      "step": 12594
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5770154316323276e-05,
+      "loss": 0.3756,
+      "step": 12595
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.576981848101691e-05,
+      "loss": 0.2984,
+      "step": 12596
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5769482645710545e-05,
+      "loss": 0.3532,
+      "step": 12597
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.576914681040418e-05,
+      "loss": 0.4126,
+      "step": 12598
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.576881097509781e-05,
+      "loss": 0.218,
+      "step": 12599
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.576847513979145e-05,
+      "loss": 0.2483,
+      "step": 12600
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.576813930448508e-05,
+      "loss": 0.3204,
+      "step": 12601
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5767803469178715e-05,
+      "loss": 0.3089,
+      "step": 12602
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.576746763387235e-05,
+      "loss": 0.2363,
+      "step": 12603
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.5767131798565985e-05,
+      "loss": 0.5362,
+      "step": 12604
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.576679596325962e-05,
+      "loss": 0.349,
+      "step": 12605
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5766460127953254e-05,
+      "loss": 0.3498,
+      "step": 12606
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.576612429264689e-05,
+      "loss": 0.3066,
+      "step": 12607
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.576578845734052e-05,
+      "loss": 0.3161,
+      "step": 12608
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5765452622034158e-05,
+      "loss": 0.5273,
+      "step": 12609
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.576511678672779e-05,
+      "loss": 0.4436,
+      "step": 12610
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5764780951421424e-05,
+      "loss": 0.4155,
+      "step": 12611
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.576444511611506e-05,
+      "loss": 0.2642,
+      "step": 12612
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5764109280808693e-05,
+      "loss": 0.3623,
+      "step": 12613
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5763773445502328e-05,
+      "loss": 0.4889,
+      "step": 12614
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5763437610195962e-05,
+      "loss": 0.317,
+      "step": 12615
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5763101774889597e-05,
+      "loss": 0.3441,
+      "step": 12616
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5762765939583228e-05,
+      "loss": 0.3802,
+      "step": 12617
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5762430104276866e-05,
+      "loss": 0.2817,
+      "step": 12618
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5762094268970498e-05,
+      "loss": 0.3407,
+      "step": 12619
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5761758433664132e-05,
+      "loss": 0.5068,
+      "step": 12620
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5761422598357767e-05,
+      "loss": 0.4078,
+      "step": 12621
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.57610867630514e-05,
+      "loss": 0.3937,
+      "step": 12622
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5760750927745033e-05,
+      "loss": 0.3268,
+      "step": 12623
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.576041509243867e-05,
+      "loss": 0.4559,
+      "step": 12624
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5760079257132302e-05,
+      "loss": 0.2389,
+      "step": 12625
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5759743421825937e-05,
+      "loss": 0.3058,
+      "step": 12626
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.575940758651957e-05,
+      "loss": 0.2915,
+      "step": 12627
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5759071751213206e-05,
+      "loss": 0.3496,
+      "step": 12628
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.575873591590684e-05,
+      "loss": 0.2116,
+      "step": 12629
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5758400080600475e-05,
+      "loss": 0.2359,
+      "step": 12630
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.575806424529411e-05,
+      "loss": 0.2068,
+      "step": 12631
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.575772840998774e-05,
+      "loss": 0.1241,
+      "step": 12632
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.575739257468138e-05,
+      "loss": 0.4621,
+      "step": 12633
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.575705673937501e-05,
+      "loss": 0.3969,
+      "step": 12634
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5756720904068645e-05,
+      "loss": 0.4194,
+      "step": 12635
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.575638506876228e-05,
+      "loss": 0.2248,
+      "step": 12636
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5756049233455914e-05,
+      "loss": 0.2572,
+      "step": 12637
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.575571339814955e-05,
+      "loss": 0.4742,
+      "step": 12638
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5755377562843184e-05,
+      "loss": 0.2377,
+      "step": 12639
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.575504172753682e-05,
+      "loss": 0.3692,
+      "step": 12640
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.575470589223045e-05,
+      "loss": 0.274,
+      "step": 12641
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5754370056924088e-05,
+      "loss": 0.3398,
+      "step": 12642
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.575403422161772e-05,
+      "loss": 0.4147,
+      "step": 12643
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5753698386311354e-05,
+      "loss": 0.3486,
+      "step": 12644
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5753362551004988e-05,
+      "loss": 0.4448,
+      "step": 12645
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5753026715698623e-05,
+      "loss": 0.2617,
+      "step": 12646
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5752690880392257e-05,
+      "loss": 0.229,
+      "step": 12647
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5752355045085892e-05,
+      "loss": 0.341,
+      "step": 12648
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5752019209779527e-05,
+      "loss": 0.2593,
+      "step": 12649
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5751683374473158e-05,
+      "loss": 0.1858,
+      "step": 12650
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5751347539166796e-05,
+      "loss": 0.2585,
+      "step": 12651
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5751011703860427e-05,
+      "loss": 0.4225,
+      "step": 12652
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5750675868554062e-05,
+      "loss": 0.259,
+      "step": 12653
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5750340033247697e-05,
+      "loss": 0.3089,
+      "step": 12654
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.575000419794133e-05,
+      "loss": 0.2517,
+      "step": 12655
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5749668362634966e-05,
+      "loss": 0.3402,
+      "step": 12656
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.57493325273286e-05,
+      "loss": 0.406,
+      "step": 12657
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5748996692022235e-05,
+      "loss": 0.305,
+      "step": 12658
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5748660856715866e-05,
+      "loss": 0.3943,
+      "step": 12659
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5748325021409505e-05,
+      "loss": 0.4237,
+      "step": 12660
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5747989186103136e-05,
+      "loss": 0.4449,
+      "step": 12661
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.574765335079677e-05,
+      "loss": 0.186,
+      "step": 12662
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5747317515490405e-05,
+      "loss": 0.2101,
+      "step": 12663
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.574698168018404e-05,
+      "loss": 0.3221,
+      "step": 12664
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5746645844877674e-05,
+      "loss": 0.2895,
+      "step": 12665
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.574631000957131e-05,
+      "loss": 0.2479,
+      "step": 12666
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.574597417426494e-05,
+      "loss": 0.4382,
+      "step": 12667
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5745638338958575e-05,
+      "loss": 0.4815,
+      "step": 12668
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.574530250365221e-05,
+      "loss": 0.2605,
+      "step": 12669
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5744966668345844e-05,
+      "loss": 0.3871,
+      "step": 12670
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.574463083303948e-05,
+      "loss": 0.143,
+      "step": 12671
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5744294997733114e-05,
+      "loss": 0.3775,
+      "step": 12672
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5743959162426748e-05,
+      "loss": 0.3125,
+      "step": 12673
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.574362332712038e-05,
+      "loss": 0.3203,
+      "step": 12674
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5743287491814017e-05,
+      "loss": 0.4199,
+      "step": 12675
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.574295165650765e-05,
+      "loss": 0.4612,
+      "step": 12676
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5742615821201283e-05,
+      "loss": 0.3013,
+      "step": 12677
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5742279985894918e-05,
+      "loss": 0.2449,
+      "step": 12678
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5741944150588553e-05,
+      "loss": 0.3077,
+      "step": 12679
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5741608315282187e-05,
+      "loss": 0.3404,
+      "step": 12680
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5741272479975822e-05,
+      "loss": 0.2792,
+      "step": 12681
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5740936644669457e-05,
+      "loss": 0.2762,
+      "step": 12682
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5740600809363088e-05,
+      "loss": 0.118,
+      "step": 12683
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5740264974056726e-05,
+      "loss": 0.2542,
+      "step": 12684
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5739929138750357e-05,
+      "loss": 0.3817,
+      "step": 12685
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5739593303443992e-05,
+      "loss": 0.4635,
+      "step": 12686
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5739257468137626e-05,
+      "loss": 0.1879,
+      "step": 12687
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.573892163283126e-05,
+      "loss": 0.3977,
+      "step": 12688
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5738585797524896e-05,
+      "loss": 0.4267,
+      "step": 12689
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.573824996221853e-05,
+      "loss": 0.4059,
+      "step": 12690
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5737914126912165e-05,
+      "loss": 0.1466,
+      "step": 12691
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5737578291605796e-05,
+      "loss": 0.3397,
+      "step": 12692
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5737242456299434e-05,
+      "loss": 0.4378,
+      "step": 12693
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5736906620993066e-05,
+      "loss": 0.2677,
+      "step": 12694
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.57365707856867e-05,
+      "loss": 0.3559,
+      "step": 12695
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5736234950380335e-05,
+      "loss": 0.362,
+      "step": 12696
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.573589911507397e-05,
+      "loss": 0.3725,
+      "step": 12697
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5735563279767604e-05,
+      "loss": 0.5386,
+      "step": 12698
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.573522744446124e-05,
+      "loss": 0.3044,
+      "step": 12699
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5734891609154873e-05,
+      "loss": 0.3081,
+      "step": 12700
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5734555773848505e-05,
+      "loss": 0.2633,
+      "step": 12701
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5734219938542143e-05,
+      "loss": 0.4354,
+      "step": 12702
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5733884103235774e-05,
+      "loss": 0.2825,
+      "step": 12703
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.573354826792941e-05,
+      "loss": 0.3644,
+      "step": 12704
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5733212432623043e-05,
+      "loss": 0.5104,
+      "step": 12705
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5732876597316678e-05,
+      "loss": 0.2728,
+      "step": 12706
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5732540762010313e-05,
+      "loss": 0.2578,
+      "step": 12707
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5732204926703947e-05,
+      "loss": 0.2411,
+      "step": 12708
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5731869091397582e-05,
+      "loss": 0.1928,
+      "step": 12709
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5731533256091213e-05,
+      "loss": 0.3471,
+      "step": 12710
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5731197420784848e-05,
+      "loss": 0.2862,
+      "step": 12711
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5730861585478482e-05,
+      "loss": 0.3222,
+      "step": 12712
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5730525750172117e-05,
+      "loss": 0.2957,
+      "step": 12713
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5730189914865752e-05,
+      "loss": 0.5412,
+      "step": 12714
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5729854079559386e-05,
+      "loss": 0.2907,
+      "step": 12715
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5729518244253018e-05,
+      "loss": 0.2854,
+      "step": 12716
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5729182408946656e-05,
+      "loss": 0.2449,
+      "step": 12717
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5728846573640287e-05,
+      "loss": 0.1548,
+      "step": 12718
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.572851073833392e-05,
+      "loss": 0.2925,
+      "step": 12719
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5728174903027556e-05,
+      "loss": 0.3818,
+      "step": 12720
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.572783906772119e-05,
+      "loss": 0.4529,
+      "step": 12721
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5727503232414826e-05,
+      "loss": 0.1914,
+      "step": 12722
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5727167397108457e-05,
+      "loss": 0.2561,
+      "step": 12723
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5726831561802095e-05,
+      "loss": 0.2407,
+      "step": 12724
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5726495726495726e-05,
+      "loss": 0.4077,
+      "step": 12725
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.572615989118936e-05,
+      "loss": 0.3234,
+      "step": 12726
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5725824055882995e-05,
+      "loss": 0.3008,
+      "step": 12727
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.572548822057663e-05,
+      "loss": 0.2393,
+      "step": 12728
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5725152385270265e-05,
+      "loss": 0.4088,
+      "step": 12729
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.57248165499639e-05,
+      "loss": 0.3529,
+      "step": 12730
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5724480714657534e-05,
+      "loss": 0.2541,
+      "step": 12731
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5724144879351165e-05,
+      "loss": 0.299,
+      "step": 12732
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5723809044044803e-05,
+      "loss": 0.3207,
+      "step": 12733
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5723473208738435e-05,
+      "loss": 0.3093,
+      "step": 12734
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.572313737343207e-05,
+      "loss": 0.2926,
+      "step": 12735
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5722801538125704e-05,
+      "loss": 0.2871,
+      "step": 12736
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.572246570281934e-05,
+      "loss": 0.2878,
+      "step": 12737
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5722129867512973e-05,
+      "loss": 0.3272,
+      "step": 12738
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5721794032206608e-05,
+      "loss": 0.4153,
+      "step": 12739
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5721458196900242e-05,
+      "loss": 0.3775,
+      "step": 12740
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5721122361593874e-05,
+      "loss": 0.3163,
+      "step": 12741
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5720786526287512e-05,
+      "loss": 0.3281,
+      "step": 12742
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5720450690981143e-05,
+      "loss": 0.255,
+      "step": 12743
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5720114855674778e-05,
+      "loss": 0.2091,
+      "step": 12744
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5719779020368412e-05,
+      "loss": 0.3185,
+      "step": 12745
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5719443185062047e-05,
+      "loss": 0.2203,
+      "step": 12746
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.571910734975568e-05,
+      "loss": 0.4193,
+      "step": 12747
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5718771514449316e-05,
+      "loss": 0.3515,
+      "step": 12748
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.571843567914295e-05,
+      "loss": 0.3388,
+      "step": 12749
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5718099843836582e-05,
+      "loss": 0.3876,
+      "step": 12750
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.571776400853022e-05,
+      "loss": 0.2876,
+      "step": 12751
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.571742817322385e-05,
+      "loss": 0.4757,
+      "step": 12752
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5717092337917486e-05,
+      "loss": 0.2906,
+      "step": 12753
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.571675650261112e-05,
+      "loss": 0.3134,
+      "step": 12754
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5716420667304755e-05,
+      "loss": 0.3007,
+      "step": 12755
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5716084831998387e-05,
+      "loss": 0.4423,
+      "step": 12756
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5715748996692025e-05,
+      "loss": 0.4298,
+      "step": 12757
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5715413161385656e-05,
+      "loss": 0.3965,
+      "step": 12758
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.571507732607929e-05,
+      "loss": 0.2461,
+      "step": 12759
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5714741490772925e-05,
+      "loss": 0.4874,
+      "step": 12760
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.571440565546656e-05,
+      "loss": 0.4679,
+      "step": 12761
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5714069820160195e-05,
+      "loss": 0.376,
+      "step": 12762
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.571373398485383e-05,
+      "loss": 0.5113,
+      "step": 12763
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5713398149547464e-05,
+      "loss": 0.1621,
+      "step": 12764
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5713062314241095e-05,
+      "loss": 0.2686,
+      "step": 12765
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5712726478934733e-05,
+      "loss": 0.2797,
+      "step": 12766
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5712390643628364e-05,
+      "loss": 0.2938,
+      "step": 12767
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5712054808322e-05,
+      "loss": 0.3747,
+      "step": 12768
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5711718973015634e-05,
+      "loss": 0.2807,
+      "step": 12769
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5711383137709268e-05,
+      "loss": 0.4458,
+      "step": 12770
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5711047302402903e-05,
+      "loss": 0.3269,
+      "step": 12771
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5710711467096538e-05,
+      "loss": 0.357,
+      "step": 12772
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5710375631790172e-05,
+      "loss": 0.3608,
+      "step": 12773
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5710039796483803e-05,
+      "loss": 0.415,
+      "step": 12774
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.570970396117744e-05,
+      "loss": 0.412,
+      "step": 12775
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5709368125871073e-05,
+      "loss": 0.4187,
+      "step": 12776
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5709032290564707e-05,
+      "loss": 0.3203,
+      "step": 12777
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5708696455258342e-05,
+      "loss": 0.3395,
+      "step": 12778
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5708360619951977e-05,
+      "loss": 0.3605,
+      "step": 12779
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.570802478464561e-05,
+      "loss": 0.3791,
+      "step": 12780
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5707688949339246e-05,
+      "loss": 0.2512,
+      "step": 12781
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.570735311403288e-05,
+      "loss": 0.47,
+      "step": 12782
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5707017278726512e-05,
+      "loss": 0.4359,
+      "step": 12783
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.570668144342015e-05,
+      "loss": 0.2399,
+      "step": 12784
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.570634560811378e-05,
+      "loss": 0.3136,
+      "step": 12785
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5706009772807416e-05,
+      "loss": 0.3253,
+      "step": 12786
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.570567393750105e-05,
+      "loss": 0.5019,
+      "step": 12787
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5705338102194685e-05,
+      "loss": 0.2568,
+      "step": 12788
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.570500226688832e-05,
+      "loss": 0.3643,
+      "step": 12789
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5704666431581954e-05,
+      "loss": 0.2816,
+      "step": 12790
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.570433059627559e-05,
+      "loss": 0.2064,
+      "step": 12791
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.570399476096922e-05,
+      "loss": 0.2881,
+      "step": 12792
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.570365892566286e-05,
+      "loss": 0.2226,
+      "step": 12793
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.570332309035649e-05,
+      "loss": 0.4326,
+      "step": 12794
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5702987255050124e-05,
+      "loss": 0.3045,
+      "step": 12795
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.570265141974376e-05,
+      "loss": 0.423,
+      "step": 12796
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5702315584437394e-05,
+      "loss": 0.3971,
+      "step": 12797
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5701979749131028e-05,
+      "loss": 0.3201,
+      "step": 12798
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5701643913824663e-05,
+      "loss": 0.2183,
+      "step": 12799
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5701308078518298e-05,
+      "loss": 0.2571,
+      "step": 12800
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.570097224321193e-05,
+      "loss": 0.2964,
+      "step": 12801
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5700636407905563e-05,
+      "loss": 0.2632,
+      "step": 12802
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.5700300572599198e-05,
+      "loss": 0.3041,
+      "step": 12803
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5699964737292833e-05,
+      "loss": 0.318,
+      "step": 12804
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5699628901986467e-05,
+      "loss": 0.1501,
+      "step": 12805
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5699293066680102e-05,
+      "loss": 0.2625,
+      "step": 12806
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5698957231373733e-05,
+      "loss": 0.2833,
+      "step": 12807
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.569862139606737e-05,
+      "loss": 0.277,
+      "step": 12808
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5698285560761003e-05,
+      "loss": 0.3525,
+      "step": 12809
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5697949725454637e-05,
+      "loss": 0.2085,
+      "step": 12810
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5697613890148272e-05,
+      "loss": 0.3563,
+      "step": 12811
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5697278054841907e-05,
+      "loss": 0.1687,
+      "step": 12812
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.569694221953554e-05,
+      "loss": 0.1409,
+      "step": 12813
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5696606384229176e-05,
+      "loss": 0.2254,
+      "step": 12814
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.569627054892281e-05,
+      "loss": 0.3061,
+      "step": 12815
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5695934713616442e-05,
+      "loss": 0.3037,
+      "step": 12816
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.569559887831008e-05,
+      "loss": 0.2968,
+      "step": 12817
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.569526304300371e-05,
+      "loss": 0.3506,
+      "step": 12818
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5694927207697346e-05,
+      "loss": 0.49,
+      "step": 12819
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.569459137239098e-05,
+      "loss": 0.321,
+      "step": 12820
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5694255537084615e-05,
+      "loss": 0.2164,
+      "step": 12821
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.569391970177825e-05,
+      "loss": 0.2629,
+      "step": 12822
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5693583866471884e-05,
+      "loss": 0.3688,
+      "step": 12823
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.569324803116552e-05,
+      "loss": 0.4798,
+      "step": 12824
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.569291219585915e-05,
+      "loss": 0.26,
+      "step": 12825
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5692576360552788e-05,
+      "loss": 0.1781,
+      "step": 12826
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.569224052524642e-05,
+      "loss": 0.2195,
+      "step": 12827
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5691904689940054e-05,
+      "loss": 0.3418,
+      "step": 12828
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.569156885463369e-05,
+      "loss": 0.3928,
+      "step": 12829
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5691233019327323e-05,
+      "loss": 0.2879,
+      "step": 12830
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5690897184020958e-05,
+      "loss": 0.2747,
+      "step": 12831
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5690561348714593e-05,
+      "loss": 0.265,
+      "step": 12832
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5690225513408227e-05,
+      "loss": 0.3199,
+      "step": 12833
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.568988967810186e-05,
+      "loss": 0.2174,
+      "step": 12834
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5689553842795497e-05,
+      "loss": 0.2891,
+      "step": 12835
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5689218007489128e-05,
+      "loss": 0.2241,
+      "step": 12836
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5688882172182763e-05,
+      "loss": 0.3515,
+      "step": 12837
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5688546336876397e-05,
+      "loss": 0.3189,
+      "step": 12838
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5688210501570032e-05,
+      "loss": 0.2817,
+      "step": 12839
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5687874666263667e-05,
+      "loss": 0.1918,
+      "step": 12840
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.56875388309573e-05,
+      "loss": 0.343,
+      "step": 12841
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5687202995650936e-05,
+      "loss": 0.3,
+      "step": 12842
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5686867160344567e-05,
+      "loss": 0.2072,
+      "step": 12843
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5686531325038205e-05,
+      "loss": 0.2703,
+      "step": 12844
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5686195489731836e-05,
+      "loss": 0.2529,
+      "step": 12845
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.568585965442547e-05,
+      "loss": 0.2571,
+      "step": 12846
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5685523819119106e-05,
+      "loss": 0.4139,
+      "step": 12847
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.568518798381274e-05,
+      "loss": 0.2718,
+      "step": 12848
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.568485214850637e-05,
+      "loss": 0.3706,
+      "step": 12849
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.568451631320001e-05,
+      "loss": 0.5274,
+      "step": 12850
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.568418047789364e-05,
+      "loss": 0.2266,
+      "step": 12851
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5683844642587275e-05,
+      "loss": 0.3893,
+      "step": 12852
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.568350880728091e-05,
+      "loss": 0.3412,
+      "step": 12853
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5683172971974545e-05,
+      "loss": 0.429,
+      "step": 12854
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.568283713666818e-05,
+      "loss": 0.3886,
+      "step": 12855
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5682501301361814e-05,
+      "loss": 0.5123,
+      "step": 12856
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.568216546605545e-05,
+      "loss": 0.2779,
+      "step": 12857
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.568182963074908e-05,
+      "loss": 0.2778,
+      "step": 12858
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5681493795442718e-05,
+      "loss": 0.2871,
+      "step": 12859
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.568115796013635e-05,
+      "loss": 0.3483,
+      "step": 12860
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5680822124829984e-05,
+      "loss": 0.2951,
+      "step": 12861
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.568048628952362e-05,
+      "loss": 0.3196,
+      "step": 12862
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5680150454217253e-05,
+      "loss": 0.2752,
+      "step": 12863
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5679814618910888e-05,
+      "loss": 0.5344,
+      "step": 12864
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5679478783604523e-05,
+      "loss": 0.2465,
+      "step": 12865
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5679142948298157e-05,
+      "loss": 0.406,
+      "step": 12866
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.567880711299179e-05,
+      "loss": 0.158,
+      "step": 12867
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5678471277685426e-05,
+      "loss": 0.4006,
+      "step": 12868
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5678135442379058e-05,
+      "loss": 0.4876,
+      "step": 12869
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5677799607072692e-05,
+      "loss": 0.4664,
+      "step": 12870
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5677463771766327e-05,
+      "loss": 0.2775,
+      "step": 12871
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.567712793645996e-05,
+      "loss": 0.3743,
+      "step": 12872
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5676792101153596e-05,
+      "loss": 0.4767,
+      "step": 12873
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.567645626584723e-05,
+      "loss": 0.4913,
+      "step": 12874
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5676120430540866e-05,
+      "loss": 0.3018,
+      "step": 12875
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5675784595234497e-05,
+      "loss": 0.2838,
+      "step": 12876
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5675448759928135e-05,
+      "loss": 0.3807,
+      "step": 12877
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5675112924621766e-05,
+      "loss": 0.4806,
+      "step": 12878
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.56747770893154e-05,
+      "loss": 0.3573,
+      "step": 12879
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5674441254009035e-05,
+      "loss": 0.2286,
+      "step": 12880
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.567410541870267e-05,
+      "loss": 0.3409,
+      "step": 12881
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5673769583396305e-05,
+      "loss": 0.2636,
+      "step": 12882
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.567343374808994e-05,
+      "loss": 0.3911,
+      "step": 12883
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5673097912783574e-05,
+      "loss": 0.2083,
+      "step": 12884
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5672762077477205e-05,
+      "loss": 0.429,
+      "step": 12885
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5672426242170843e-05,
+      "loss": 0.3854,
+      "step": 12886
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5672090406864475e-05,
+      "loss": 0.4112,
+      "step": 12887
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.567175457155811e-05,
+      "loss": 0.3594,
+      "step": 12888
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5671418736251744e-05,
+      "loss": 0.3457,
+      "step": 12889
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.567108290094538e-05,
+      "loss": 0.5448,
+      "step": 12890
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.567074706563901e-05,
+      "loss": 0.328,
+      "step": 12891
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5670411230332648e-05,
+      "loss": 0.1934,
+      "step": 12892
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.567007539502628e-05,
+      "loss": 0.256,
+      "step": 12893
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5669739559719914e-05,
+      "loss": 0.2352,
+      "step": 12894
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.566940372441355e-05,
+      "loss": 0.4044,
+      "step": 12895
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5669067889107183e-05,
+      "loss": 0.2416,
+      "step": 12896
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5668732053800818e-05,
+      "loss": 0.2871,
+      "step": 12897
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5668396218494452e-05,
+      "loss": 0.2193,
+      "step": 12898
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5668060383188087e-05,
+      "loss": 0.3577,
+      "step": 12899
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5667724547881718e-05,
+      "loss": 0.4285,
+      "step": 12900
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5667388712575356e-05,
+      "loss": 0.3832,
+      "step": 12901
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5667052877268988e-05,
+      "loss": 0.3148,
+      "step": 12902
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5666717041962622e-05,
+      "loss": 0.2744,
+      "step": 12903
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5666381206656257e-05,
+      "loss": 0.238,
+      "step": 12904
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.566604537134989e-05,
+      "loss": 0.2889,
+      "step": 12905
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5665709536043526e-05,
+      "loss": 0.427,
+      "step": 12906
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5665373700737157e-05,
+      "loss": 0.3245,
+      "step": 12907
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5665037865430795e-05,
+      "loss": 0.1959,
+      "step": 12908
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5664702030124427e-05,
+      "loss": 0.2931,
+      "step": 12909
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.566436619481806e-05,
+      "loss": 0.324,
+      "step": 12910
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5664030359511696e-05,
+      "loss": 0.2909,
+      "step": 12911
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.566369452420533e-05,
+      "loss": 0.2587,
+      "step": 12912
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5663358688898965e-05,
+      "loss": 0.4435,
+      "step": 12913
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.56630228535926e-05,
+      "loss": 0.2946,
+      "step": 12914
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5662687018286235e-05,
+      "loss": 0.4055,
+      "step": 12915
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5662351182979866e-05,
+      "loss": 0.2445,
+      "step": 12916
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5662015347673504e-05,
+      "loss": 0.4323,
+      "step": 12917
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5661679512367135e-05,
+      "loss": 0.2799,
+      "step": 12918
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.566134367706077e-05,
+      "loss": 0.2645,
+      "step": 12919
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5661007841754404e-05,
+      "loss": 0.2554,
+      "step": 12920
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.566067200644804e-05,
+      "loss": 0.2884,
+      "step": 12921
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5660336171141674e-05,
+      "loss": 0.2114,
+      "step": 12922
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.566000033583531e-05,
+      "loss": 0.3337,
+      "step": 12923
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5659664500528943e-05,
+      "loss": 0.4055,
+      "step": 12924
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5659328665222574e-05,
+      "loss": 0.2487,
+      "step": 12925
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5658992829916212e-05,
+      "loss": 0.2968,
+      "step": 12926
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5658656994609844e-05,
+      "loss": 0.3528,
+      "step": 12927
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5658321159303478e-05,
+      "loss": 0.3625,
+      "step": 12928
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5657985323997113e-05,
+      "loss": 0.3229,
+      "step": 12929
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5657649488690748e-05,
+      "loss": 0.2208,
+      "step": 12930
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5657313653384382e-05,
+      "loss": 0.2369,
+      "step": 12931
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5656977818078017e-05,
+      "loss": 0.3998,
+      "step": 12932
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.565664198277165e-05,
+      "loss": 0.3335,
+      "step": 12933
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5656306147465283e-05,
+      "loss": 0.2442,
+      "step": 12934
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.565597031215892e-05,
+      "loss": 0.431,
+      "step": 12935
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5655634476852552e-05,
+      "loss": 0.2965,
+      "step": 12936
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5655298641546187e-05,
+      "loss": 0.1834,
+      "step": 12937
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.565496280623982e-05,
+      "loss": 0.3124,
+      "step": 12938
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5654626970933456e-05,
+      "loss": 0.355,
+      "step": 12939
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5654291135627087e-05,
+      "loss": 0.2691,
+      "step": 12940
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5653955300320725e-05,
+      "loss": 0.3398,
+      "step": 12941
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5653619465014356e-05,
+      "loss": 0.2282,
+      "step": 12942
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.565328362970799e-05,
+      "loss": 0.458,
+      "step": 12943
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5652947794401626e-05,
+      "loss": 0.3515,
+      "step": 12944
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.565261195909526e-05,
+      "loss": 0.2156,
+      "step": 12945
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5652276123788895e-05,
+      "loss": 0.2378,
+      "step": 12946
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.565194028848253e-05,
+      "loss": 0.3196,
+      "step": 12947
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5651604453176164e-05,
+      "loss": 0.3987,
+      "step": 12948
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5651268617869796e-05,
+      "loss": 0.2421,
+      "step": 12949
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5650932782563434e-05,
+      "loss": 0.2876,
+      "step": 12950
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5650596947257065e-05,
+      "loss": 0.3565,
+      "step": 12951
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.56502611119507e-05,
+      "loss": 0.2572,
+      "step": 12952
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5649925276644334e-05,
+      "loss": 0.184,
+      "step": 12953
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.564958944133797e-05,
+      "loss": 0.4206,
+      "step": 12954
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5649253606031604e-05,
+      "loss": 0.222,
+      "step": 12955
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5648917770725238e-05,
+      "loss": 0.3932,
+      "step": 12956
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5648581935418873e-05,
+      "loss": 0.6499,
+      "step": 12957
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5648246100112504e-05,
+      "loss": 0.2885,
+      "step": 12958
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5647910264806142e-05,
+      "loss": 0.272,
+      "step": 12959
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5647574429499773e-05,
+      "loss": 0.3517,
+      "step": 12960
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5647238594193408e-05,
+      "loss": 0.2443,
+      "step": 12961
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5646902758887043e-05,
+      "loss": 0.2182,
+      "step": 12962
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5646566923580677e-05,
+      "loss": 0.3039,
+      "step": 12963
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5646231088274312e-05,
+      "loss": 0.3887,
+      "step": 12964
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5645895252967947e-05,
+      "loss": 0.2058,
+      "step": 12965
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.564555941766158e-05,
+      "loss": 0.3484,
+      "step": 12966
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5645223582355213e-05,
+      "loss": 0.2792,
+      "step": 12967
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.564488774704885e-05,
+      "loss": 0.3062,
+      "step": 12968
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5644551911742482e-05,
+      "loss": 0.3308,
+      "step": 12969
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5644216076436116e-05,
+      "loss": 0.2359,
+      "step": 12970
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.564388024112975e-05,
+      "loss": 0.362,
+      "step": 12971
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5643544405823386e-05,
+      "loss": 0.3754,
+      "step": 12972
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.564320857051702e-05,
+      "loss": 0.2817,
+      "step": 12973
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5642872735210655e-05,
+      "loss": 0.2325,
+      "step": 12974
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.564253689990429e-05,
+      "loss": 0.3991,
+      "step": 12975
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.564220106459792e-05,
+      "loss": 0.2428,
+      "step": 12976
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.564186522929156e-05,
+      "loss": 0.2712,
+      "step": 12977
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.564152939398519e-05,
+      "loss": 0.3222,
+      "step": 12978
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5641193558678825e-05,
+      "loss": 0.3539,
+      "step": 12979
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.564085772337246e-05,
+      "loss": 0.3033,
+      "step": 12980
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5640521888066094e-05,
+      "loss": 0.3014,
+      "step": 12981
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5640186052759725e-05,
+      "loss": 0.3777,
+      "step": 12982
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5639850217453363e-05,
+      "loss": 0.2042,
+      "step": 12983
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5639514382146995e-05,
+      "loss": 0.2474,
+      "step": 12984
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.563917854684063e-05,
+      "loss": 0.2488,
+      "step": 12985
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5638842711534264e-05,
+      "loss": 0.2075,
+      "step": 12986
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.56385068762279e-05,
+      "loss": 0.3065,
+      "step": 12987
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5638171040921533e-05,
+      "loss": 0.311,
+      "step": 12988
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5637835205615168e-05,
+      "loss": 0.2794,
+      "step": 12989
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5637499370308803e-05,
+      "loss": 0.2436,
+      "step": 12990
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5637163535002434e-05,
+      "loss": 0.2554,
+      "step": 12991
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5636827699696072e-05,
+      "loss": 0.4763,
+      "step": 12992
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5636491864389703e-05,
+      "loss": 0.2274,
+      "step": 12993
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5636156029083338e-05,
+      "loss": 0.3399,
+      "step": 12994
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5635820193776972e-05,
+      "loss": 0.2848,
+      "step": 12995
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5635484358470607e-05,
+      "loss": 0.2308,
+      "step": 12996
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5635148523164242e-05,
+      "loss": 0.2502,
+      "step": 12997
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5634812687857876e-05,
+      "loss": 0.4423,
+      "step": 12998
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.563447685255151e-05,
+      "loss": 0.317,
+      "step": 12999
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.5634141017245142e-05,
+      "loss": 0.2879,
+      "step": 13000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.563380518193878e-05,
+      "loss": 0.3364,
+      "step": 13001
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.563346934663241e-05,
+      "loss": 0.3158,
+      "step": 13002
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5633133511326046e-05,
+      "loss": 0.4454,
+      "step": 13003
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.563279767601968e-05,
+      "loss": 0.288,
+      "step": 13004
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5632461840713316e-05,
+      "loss": 0.2906,
+      "step": 13005
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.563212600540695e-05,
+      "loss": 0.2632,
+      "step": 13006
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5631790170100585e-05,
+      "loss": 0.356,
+      "step": 13007
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.563145433479422e-05,
+      "loss": 0.389,
+      "step": 13008
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.563111849948785e-05,
+      "loss": 0.2231,
+      "step": 13009
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.563078266418149e-05,
+      "loss": 0.26,
+      "step": 13010
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.563044682887512e-05,
+      "loss": 0.3834,
+      "step": 13011
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5630110993568755e-05,
+      "loss": 0.4192,
+      "step": 13012
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.562977515826239e-05,
+      "loss": 0.3362,
+      "step": 13013
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5629439322956024e-05,
+      "loss": 0.3654,
+      "step": 13014
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.562910348764966e-05,
+      "loss": 0.2202,
+      "step": 13015
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5628767652343293e-05,
+      "loss": 0.3857,
+      "step": 13016
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5628431817036928e-05,
+      "loss": 0.2896,
+      "step": 13017
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.562809598173056e-05,
+      "loss": 0.3423,
+      "step": 13018
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5627760146424197e-05,
+      "loss": 0.276,
+      "step": 13019
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.562742431111783e-05,
+      "loss": 0.3263,
+      "step": 13020
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5627088475811463e-05,
+      "loss": 0.2071,
+      "step": 13021
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5626752640505098e-05,
+      "loss": 0.3943,
+      "step": 13022
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5626416805198732e-05,
+      "loss": 0.2754,
+      "step": 13023
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5626080969892367e-05,
+      "loss": 0.2878,
+      "step": 13024
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5625745134586002e-05,
+      "loss": 0.5004,
+      "step": 13025
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5625409299279633e-05,
+      "loss": 0.208,
+      "step": 13026
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5625073463973268e-05,
+      "loss": 0.3829,
+      "step": 13027
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5624737628666902e-05,
+      "loss": 0.2562,
+      "step": 13028
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5624401793360537e-05,
+      "loss": 0.1699,
+      "step": 13029
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.562406595805417e-05,
+      "loss": 0.4283,
+      "step": 13030
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5623730122747806e-05,
+      "loss": 0.4233,
+      "step": 13031
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.562339428744144e-05,
+      "loss": 0.4238,
+      "step": 13032
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5623058452135072e-05,
+      "loss": 0.2486,
+      "step": 13033
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.562272261682871e-05,
+      "loss": 0.2552,
+      "step": 13034
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.562238678152234e-05,
+      "loss": 0.3059,
+      "step": 13035
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5622050946215976e-05,
+      "loss": 0.2151,
+      "step": 13036
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.562171511090961e-05,
+      "loss": 0.3188,
+      "step": 13037
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5621379275603245e-05,
+      "loss": 0.3562,
+      "step": 13038
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.562104344029688e-05,
+      "loss": 0.3305,
+      "step": 13039
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5620707604990515e-05,
+      "loss": 0.3974,
+      "step": 13040
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.562037176968415e-05,
+      "loss": 0.5293,
+      "step": 13041
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.562003593437778e-05,
+      "loss": 0.298,
+      "step": 13042
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.561970009907142e-05,
+      "loss": 0.4376,
+      "step": 13043
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.561936426376505e-05,
+      "loss": 0.2866,
+      "step": 13044
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5619028428458685e-05,
+      "loss": 0.4191,
+      "step": 13045
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.561869259315232e-05,
+      "loss": 0.3456,
+      "step": 13046
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5618356757845954e-05,
+      "loss": 0.2539,
+      "step": 13047
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.561802092253959e-05,
+      "loss": 0.3513,
+      "step": 13048
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5617685087233223e-05,
+      "loss": 0.3125,
+      "step": 13049
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5617349251926858e-05,
+      "loss": 0.1954,
+      "step": 13050
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.561701341662049e-05,
+      "loss": 0.2964,
+      "step": 13051
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5616677581314127e-05,
+      "loss": 0.3081,
+      "step": 13052
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.561634174600776e-05,
+      "loss": 0.3576,
+      "step": 13053
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5616005910701393e-05,
+      "loss": 0.2952,
+      "step": 13054
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5615670075395028e-05,
+      "loss": 0.3973,
+      "step": 13055
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5615334240088662e-05,
+      "loss": 0.2887,
+      "step": 13056
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5614998404782297e-05,
+      "loss": 0.2977,
+      "step": 13057
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.561466256947593e-05,
+      "loss": 0.4761,
+      "step": 13058
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5614326734169566e-05,
+      "loss": 0.4652,
+      "step": 13059
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5613990898863197e-05,
+      "loss": 0.1646,
+      "step": 13060
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5613655063556835e-05,
+      "loss": 0.4798,
+      "step": 13061
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5613319228250467e-05,
+      "loss": 0.1607,
+      "step": 13062
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.56129833929441e-05,
+      "loss": 0.3881,
+      "step": 13063
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5612647557637736e-05,
+      "loss": 0.4938,
+      "step": 13064
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.561231172233137e-05,
+      "loss": 0.3456,
+      "step": 13065
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5611975887025005e-05,
+      "loss": 0.1898,
+      "step": 13066
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.561164005171864e-05,
+      "loss": 0.3354,
+      "step": 13067
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5611304216412275e-05,
+      "loss": 0.4145,
+      "step": 13068
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5610968381105906e-05,
+      "loss": 0.341,
+      "step": 13069
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5610632545799544e-05,
+      "loss": 0.1825,
+      "step": 13070
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5610296710493175e-05,
+      "loss": 0.2477,
+      "step": 13071
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.560996087518681e-05,
+      "loss": 0.2697,
+      "step": 13072
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5609625039880444e-05,
+      "loss": 0.3537,
+      "step": 13073
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.560928920457408e-05,
+      "loss": 0.289,
+      "step": 13074
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.560895336926771e-05,
+      "loss": 0.2861,
+      "step": 13075
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.560861753396135e-05,
+      "loss": 0.3309,
+      "step": 13076
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.560828169865498e-05,
+      "loss": 0.2653,
+      "step": 13077
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5607945863348614e-05,
+      "loss": 0.2137,
+      "step": 13078
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.560761002804225e-05,
+      "loss": 0.2816,
+      "step": 13079
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5607274192735884e-05,
+      "loss": 0.3088,
+      "step": 13080
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5606938357429518e-05,
+      "loss": 0.3224,
+      "step": 13081
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5606602522123153e-05,
+      "loss": 0.4191,
+      "step": 13082
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5606266686816788e-05,
+      "loss": 0.2685,
+      "step": 13083
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.560593085151042e-05,
+      "loss": 0.5251,
+      "step": 13084
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5605595016204053e-05,
+      "loss": 0.2449,
+      "step": 13085
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5605259180897688e-05,
+      "loss": 0.4253,
+      "step": 13086
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5604923345591323e-05,
+      "loss": 0.4378,
+      "step": 13087
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5604587510284957e-05,
+      "loss": 0.3043,
+      "step": 13088
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5604251674978592e-05,
+      "loss": 0.2513,
+      "step": 13089
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5603915839672227e-05,
+      "loss": 0.3635,
+      "step": 13090
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5603580004365858e-05,
+      "loss": 0.1853,
+      "step": 13091
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5603244169059496e-05,
+      "loss": 0.2283,
+      "step": 13092
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5602908333753127e-05,
+      "loss": 0.3341,
+      "step": 13093
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5602572498446762e-05,
+      "loss": 0.4822,
+      "step": 13094
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5602236663140397e-05,
+      "loss": 0.5493,
+      "step": 13095
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.560190082783403e-05,
+      "loss": 0.2858,
+      "step": 13096
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5601564992527666e-05,
+      "loss": 0.3274,
+      "step": 13097
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.56012291572213e-05,
+      "loss": 0.3348,
+      "step": 13098
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5600893321914935e-05,
+      "loss": 0.2057,
+      "step": 13099
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5600557486608566e-05,
+      "loss": 0.261,
+      "step": 13100
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5600221651302204e-05,
+      "loss": 0.3337,
+      "step": 13101
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5599885815995836e-05,
+      "loss": 0.433,
+      "step": 13102
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.559954998068947e-05,
+      "loss": 0.2994,
+      "step": 13103
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5599214145383105e-05,
+      "loss": 0.3809,
+      "step": 13104
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.559887831007674e-05,
+      "loss": 0.408,
+      "step": 13105
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5598542474770374e-05,
+      "loss": 0.3969,
+      "step": 13106
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.559820663946401e-05,
+      "loss": 0.4307,
+      "step": 13107
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5597870804157644e-05,
+      "loss": 0.3496,
+      "step": 13108
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5597534968851275e-05,
+      "loss": 0.4302,
+      "step": 13109
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5597199133544913e-05,
+      "loss": 0.19,
+      "step": 13110
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5596863298238544e-05,
+      "loss": 0.374,
+      "step": 13111
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.559652746293218e-05,
+      "loss": 0.2288,
+      "step": 13112
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5596191627625813e-05,
+      "loss": 0.246,
+      "step": 13113
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5595855792319448e-05,
+      "loss": 0.3706,
+      "step": 13114
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5595519957013083e-05,
+      "loss": 0.3169,
+      "step": 13115
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5595184121706717e-05,
+      "loss": 0.3388,
+      "step": 13116
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.559484828640035e-05,
+      "loss": 0.3415,
+      "step": 13117
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5594512451093983e-05,
+      "loss": 0.3231,
+      "step": 13118
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5594176615787618e-05,
+      "loss": 0.4827,
+      "step": 13119
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5593840780481253e-05,
+      "loss": 0.2851,
+      "step": 13120
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5593504945174887e-05,
+      "loss": 0.3018,
+      "step": 13121
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5593169109868522e-05,
+      "loss": 0.3603,
+      "step": 13122
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5592833274562157e-05,
+      "loss": 0.2053,
+      "step": 13123
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5592497439255788e-05,
+      "loss": 0.1865,
+      "step": 13124
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5592161603949426e-05,
+      "loss": 0.3248,
+      "step": 13125
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5591825768643057e-05,
+      "loss": 0.3677,
+      "step": 13126
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5591489933336692e-05,
+      "loss": 0.3166,
+      "step": 13127
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5591154098030326e-05,
+      "loss": 0.2756,
+      "step": 13128
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.559081826272396e-05,
+      "loss": 0.6123,
+      "step": 13129
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5590482427417596e-05,
+      "loss": 0.4494,
+      "step": 13130
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.559014659211123e-05,
+      "loss": 0.2423,
+      "step": 13131
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5589810756804865e-05,
+      "loss": 0.3647,
+      "step": 13132
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5589474921498496e-05,
+      "loss": 0.3129,
+      "step": 13133
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5589139086192134e-05,
+      "loss": 0.4754,
+      "step": 13134
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5588803250885766e-05,
+      "loss": 0.2838,
+      "step": 13135
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.55884674155794e-05,
+      "loss": 0.2745,
+      "step": 13136
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5588131580273035e-05,
+      "loss": 0.2786,
+      "step": 13137
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.558779574496667e-05,
+      "loss": 0.3568,
+      "step": 13138
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5587459909660304e-05,
+      "loss": 0.5467,
+      "step": 13139
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.558712407435394e-05,
+      "loss": 0.3519,
+      "step": 13140
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5586788239047573e-05,
+      "loss": 0.3613,
+      "step": 13141
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5586452403741205e-05,
+      "loss": 0.4075,
+      "step": 13142
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5586116568434843e-05,
+      "loss": 0.4118,
+      "step": 13143
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5585780733128474e-05,
+      "loss": 0.4279,
+      "step": 13144
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.558544489782211e-05,
+      "loss": 0.3808,
+      "step": 13145
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5585109062515743e-05,
+      "loss": 0.3648,
+      "step": 13146
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5584773227209378e-05,
+      "loss": 0.4187,
+      "step": 13147
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5584437391903013e-05,
+      "loss": 0.3612,
+      "step": 13148
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5584101556596647e-05,
+      "loss": 0.3696,
+      "step": 13149
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5583765721290282e-05,
+      "loss": 0.3251,
+      "step": 13150
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5583429885983913e-05,
+      "loss": 0.2822,
+      "step": 13151
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.558309405067755e-05,
+      "loss": 0.2106,
+      "step": 13152
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5582758215371182e-05,
+      "loss": 0.3619,
+      "step": 13153
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5582422380064817e-05,
+      "loss": 0.2721,
+      "step": 13154
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.558208654475845e-05,
+      "loss": 0.4682,
+      "step": 13155
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5581750709452086e-05,
+      "loss": 0.3923,
+      "step": 13156
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.558141487414572e-05,
+      "loss": 0.2316,
+      "step": 13157
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5581079038839356e-05,
+      "loss": 0.2507,
+      "step": 13158
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.558074320353299e-05,
+      "loss": 0.3645,
+      "step": 13159
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.558040736822662e-05,
+      "loss": 0.3861,
+      "step": 13160
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5580071532920256e-05,
+      "loss": 0.4235,
+      "step": 13161
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.557973569761389e-05,
+      "loss": 0.1658,
+      "step": 13162
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5579399862307525e-05,
+      "loss": 0.2971,
+      "step": 13163
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.557906402700116e-05,
+      "loss": 0.3867,
+      "step": 13164
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5578728191694795e-05,
+      "loss": 0.2909,
+      "step": 13165
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5578392356388426e-05,
+      "loss": 0.5257,
+      "step": 13166
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5578056521082064e-05,
+      "loss": 0.2337,
+      "step": 13167
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5577720685775695e-05,
+      "loss": 0.3914,
+      "step": 13168
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.557738485046933e-05,
+      "loss": 0.4306,
+      "step": 13169
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5577049015162965e-05,
+      "loss": 0.3303,
+      "step": 13170
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.55767131798566e-05,
+      "loss": 0.2371,
+      "step": 13171
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5576377344550234e-05,
+      "loss": 0.4759,
+      "step": 13172
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.557604150924387e-05,
+      "loss": 0.3991,
+      "step": 13173
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5575705673937503e-05,
+      "loss": 0.4537,
+      "step": 13174
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5575369838631134e-05,
+      "loss": 0.3468,
+      "step": 13175
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5575034003324773e-05,
+      "loss": 0.5011,
+      "step": 13176
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5574698168018404e-05,
+      "loss": 0.2737,
+      "step": 13177
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.557436233271204e-05,
+      "loss": 0.3569,
+      "step": 13178
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5574026497405673e-05,
+      "loss": 0.268,
+      "step": 13179
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5573690662099308e-05,
+      "loss": 0.2134,
+      "step": 13180
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5573354826792942e-05,
+      "loss": 0.5034,
+      "step": 13181
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5573018991486577e-05,
+      "loss": 0.3446,
+      "step": 13182
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.557268315618021e-05,
+      "loss": 0.3031,
+      "step": 13183
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5572347320873843e-05,
+      "loss": 0.3624,
+      "step": 13184
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.557201148556748e-05,
+      "loss": 0.2495,
+      "step": 13185
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5571675650261112e-05,
+      "loss": 0.3569,
+      "step": 13186
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5571339814954747e-05,
+      "loss": 0.3313,
+      "step": 13187
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.557100397964838e-05,
+      "loss": 0.4426,
+      "step": 13188
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5570668144342016e-05,
+      "loss": 0.2775,
+      "step": 13189
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.557033230903565e-05,
+      "loss": 0.3494,
+      "step": 13190
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5569996473729285e-05,
+      "loss": 0.2335,
+      "step": 13191
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.556966063842292e-05,
+      "loss": 0.3527,
+      "step": 13192
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.556932480311655e-05,
+      "loss": 0.2279,
+      "step": 13193
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.556898896781019e-05,
+      "loss": 0.3621,
+      "step": 13194
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.556865313250382e-05,
+      "loss": 0.4284,
+      "step": 13195
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5568317297197455e-05,
+      "loss": 0.467,
+      "step": 13196
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.556798146189109e-05,
+      "loss": 0.2703,
+      "step": 13197
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5567645626584725e-05,
+      "loss": 0.4032,
+      "step": 13198
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.556730979127836e-05,
+      "loss": 0.2656,
+      "step": 13199
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.5566973955971994e-05,
+      "loss": 0.3562,
+      "step": 13200
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.556663812066563e-05,
+      "loss": 0.4431,
+      "step": 13201
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.556630228535926e-05,
+      "loss": 0.3009,
+      "step": 13202
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5565966450052898e-05,
+      "loss": 0.2134,
+      "step": 13203
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.556563061474653e-05,
+      "loss": 0.3141,
+      "step": 13204
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5565294779440164e-05,
+      "loss": 0.3098,
+      "step": 13205
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.55649589441338e-05,
+      "loss": 0.2471,
+      "step": 13206
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5564623108827433e-05,
+      "loss": 0.2742,
+      "step": 13207
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5564287273521064e-05,
+      "loss": 0.3335,
+      "step": 13208
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5563951438214702e-05,
+      "loss": 0.2957,
+      "step": 13209
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5563615602908334e-05,
+      "loss": 0.3659,
+      "step": 13210
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5563279767601968e-05,
+      "loss": 0.4906,
+      "step": 13211
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5562943932295603e-05,
+      "loss": 0.2922,
+      "step": 13212
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5562608096989238e-05,
+      "loss": 0.3046,
+      "step": 13213
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5562272261682872e-05,
+      "loss": 0.2868,
+      "step": 13214
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5561936426376507e-05,
+      "loss": 0.4988,
+      "step": 13215
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.556160059107014e-05,
+      "loss": 0.376,
+      "step": 13216
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5561264755763773e-05,
+      "loss": 0.4714,
+      "step": 13217
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.556092892045741e-05,
+      "loss": 0.2996,
+      "step": 13218
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5560593085151042e-05,
+      "loss": 0.2307,
+      "step": 13219
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5560257249844677e-05,
+      "loss": 0.3957,
+      "step": 13220
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.555992141453831e-05,
+      "loss": 0.282,
+      "step": 13221
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5559585579231946e-05,
+      "loss": 0.2824,
+      "step": 13222
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.555924974392558e-05,
+      "loss": 0.3077,
+      "step": 13223
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5558913908619215e-05,
+      "loss": 0.3254,
+      "step": 13224
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.555857807331285e-05,
+      "loss": 0.2963,
+      "step": 13225
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.555824223800648e-05,
+      "loss": 0.2169,
+      "step": 13226
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.555790640270012e-05,
+      "loss": 0.396,
+      "step": 13227
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.555757056739375e-05,
+      "loss": 0.4339,
+      "step": 13228
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5557234732087385e-05,
+      "loss": 0.4286,
+      "step": 13229
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.555689889678102e-05,
+      "loss": 0.2877,
+      "step": 13230
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5556563061474654e-05,
+      "loss": 0.1746,
+      "step": 13231
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.555622722616829e-05,
+      "loss": 0.3281,
+      "step": 13232
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5555891390861924e-05,
+      "loss": 0.3776,
+      "step": 13233
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.555555555555556e-05,
+      "loss": 0.224,
+      "step": 13234
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.555521972024919e-05,
+      "loss": 0.3755,
+      "step": 13235
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5554883884942828e-05,
+      "loss": 0.3172,
+      "step": 13236
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.555454804963646e-05,
+      "loss": 0.4057,
+      "step": 13237
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5554212214330094e-05,
+      "loss": 0.318,
+      "step": 13238
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5553876379023728e-05,
+      "loss": 0.157,
+      "step": 13239
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5553540543717363e-05,
+      "loss": 0.382,
+      "step": 13240
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5553204708410997e-05,
+      "loss": 0.4264,
+      "step": 13241
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5552868873104632e-05,
+      "loss": 0.2129,
+      "step": 13242
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5552533037798267e-05,
+      "loss": 0.5463,
+      "step": 13243
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5552197202491898e-05,
+      "loss": 0.3583,
+      "step": 13244
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5551861367185536e-05,
+      "loss": 0.2756,
+      "step": 13245
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5551525531879167e-05,
+      "loss": 0.3745,
+      "step": 13246
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5551189696572802e-05,
+      "loss": 0.2558,
+      "step": 13247
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5550853861266437e-05,
+      "loss": 0.3319,
+      "step": 13248
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.555051802596007e-05,
+      "loss": 0.3005,
+      "step": 13249
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5550182190653706e-05,
+      "loss": 0.4036,
+      "step": 13250
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.554984635534734e-05,
+      "loss": 0.4074,
+      "step": 13251
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5549510520040972e-05,
+      "loss": 0.2934,
+      "step": 13252
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5549174684734606e-05,
+      "loss": 0.2464,
+      "step": 13253
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.554883884942824e-05,
+      "loss": 0.204,
+      "step": 13254
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5548503014121876e-05,
+      "loss": 0.4397,
+      "step": 13255
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.554816717881551e-05,
+      "loss": 0.1657,
+      "step": 13256
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5547831343509145e-05,
+      "loss": 0.1872,
+      "step": 13257
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.554749550820278e-05,
+      "loss": 0.3738,
+      "step": 13258
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.554715967289641e-05,
+      "loss": 0.3267,
+      "step": 13259
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.554682383759005e-05,
+      "loss": 0.4567,
+      "step": 13260
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.554648800228368e-05,
+      "loss": 0.2316,
+      "step": 13261
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5546152166977315e-05,
+      "loss": 0.3261,
+      "step": 13262
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.554581633167095e-05,
+      "loss": 0.2864,
+      "step": 13263
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5545480496364584e-05,
+      "loss": 0.2656,
+      "step": 13264
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.554514466105822e-05,
+      "loss": 0.3241,
+      "step": 13265
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.554480882575185e-05,
+      "loss": 0.1844,
+      "step": 13266
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5544472990445488e-05,
+      "loss": 0.235,
+      "step": 13267
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.554413715513912e-05,
+      "loss": 0.3512,
+      "step": 13268
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5543801319832754e-05,
+      "loss": 0.243,
+      "step": 13269
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.554346548452639e-05,
+      "loss": 0.2728,
+      "step": 13270
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5543129649220023e-05,
+      "loss": 0.2418,
+      "step": 13271
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5542793813913658e-05,
+      "loss": 0.3402,
+      "step": 13272
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5542457978607293e-05,
+      "loss": 0.2938,
+      "step": 13273
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5542122143300927e-05,
+      "loss": 0.1678,
+      "step": 13274
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.554178630799456e-05,
+      "loss": 0.2038,
+      "step": 13275
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5541450472688197e-05,
+      "loss": 0.2365,
+      "step": 13276
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5541114637381828e-05,
+      "loss": 0.4259,
+      "step": 13277
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5540778802075462e-05,
+      "loss": 0.3608,
+      "step": 13278
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5540442966769097e-05,
+      "loss": 0.1972,
+      "step": 13279
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5540107131462732e-05,
+      "loss": 0.2409,
+      "step": 13280
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5539771296156366e-05,
+      "loss": 0.3467,
+      "step": 13281
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.553943546085e-05,
+      "loss": 0.3616,
+      "step": 13282
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5539099625543636e-05,
+      "loss": 0.27,
+      "step": 13283
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5538763790237267e-05,
+      "loss": 0.3023,
+      "step": 13284
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5538427954930905e-05,
+      "loss": 0.378,
+      "step": 13285
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5538092119624536e-05,
+      "loss": 0.1621,
+      "step": 13286
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.553775628431817e-05,
+      "loss": 0.3687,
+      "step": 13287
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5537420449011806e-05,
+      "loss": 0.2602,
+      "step": 13288
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.553708461370544e-05,
+      "loss": 0.2061,
+      "step": 13289
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5536748778399075e-05,
+      "loss": 0.1372,
+      "step": 13290
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.553641294309271e-05,
+      "loss": 0.5229,
+      "step": 13291
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5536077107786344e-05,
+      "loss": 0.3859,
+      "step": 13292
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5535741272479975e-05,
+      "loss": 0.3692,
+      "step": 13293
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5535405437173613e-05,
+      "loss": 0.4022,
+      "step": 13294
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5535069601867245e-05,
+      "loss": 0.1897,
+      "step": 13295
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.553473376656088e-05,
+      "loss": 0.2569,
+      "step": 13296
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5534397931254514e-05,
+      "loss": 0.1968,
+      "step": 13297
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.553406209594815e-05,
+      "loss": 0.4714,
+      "step": 13298
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.553372626064178e-05,
+      "loss": 0.485,
+      "step": 13299
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5533390425335418e-05,
+      "loss": 0.2855,
+      "step": 13300
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.553305459002905e-05,
+      "loss": 0.3527,
+      "step": 13301
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5532718754722684e-05,
+      "loss": 0.51,
+      "step": 13302
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.553238291941632e-05,
+      "loss": 0.2619,
+      "step": 13303
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5532047084109953e-05,
+      "loss": 0.3891,
+      "step": 13304
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5531711248803588e-05,
+      "loss": 0.3751,
+      "step": 13305
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5531375413497222e-05,
+      "loss": 0.5847,
+      "step": 13306
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5531039578190857e-05,
+      "loss": 0.3431,
+      "step": 13307
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.553070374288449e-05,
+      "loss": 0.2067,
+      "step": 13308
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5530367907578126e-05,
+      "loss": 0.2651,
+      "step": 13309
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5530032072271758e-05,
+      "loss": 0.2724,
+      "step": 13310
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5529696236965392e-05,
+      "loss": 0.4346,
+      "step": 13311
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5529360401659027e-05,
+      "loss": 0.3465,
+      "step": 13312
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.552902456635266e-05,
+      "loss": 0.2093,
+      "step": 13313
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5528688731046296e-05,
+      "loss": 0.3414,
+      "step": 13314
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.552835289573993e-05,
+      "loss": 0.2278,
+      "step": 13315
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5528017060433566e-05,
+      "loss": 0.3601,
+      "step": 13316
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5527681225127197e-05,
+      "loss": 0.3697,
+      "step": 13317
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5527345389820835e-05,
+      "loss": 0.3212,
+      "step": 13318
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5527009554514466e-05,
+      "loss": 0.3216,
+      "step": 13319
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.55266737192081e-05,
+      "loss": 0.4085,
+      "step": 13320
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5526337883901735e-05,
+      "loss": 0.4838,
+      "step": 13321
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.552600204859537e-05,
+      "loss": 0.3785,
+      "step": 13322
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5525666213289005e-05,
+      "loss": 0.2453,
+      "step": 13323
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.552533037798264e-05,
+      "loss": 0.1917,
+      "step": 13324
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5524994542676274e-05,
+      "loss": 0.2367,
+      "step": 13325
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5524658707369905e-05,
+      "loss": 0.397,
+      "step": 13326
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5524322872063543e-05,
+      "loss": 0.3188,
+      "step": 13327
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5523987036757175e-05,
+      "loss": 0.3643,
+      "step": 13328
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.552365120145081e-05,
+      "loss": 0.3315,
+      "step": 13329
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5523315366144444e-05,
+      "loss": 0.3097,
+      "step": 13330
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.552297953083808e-05,
+      "loss": 0.2263,
+      "step": 13331
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5522643695531713e-05,
+      "loss": 0.3176,
+      "step": 13332
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5522307860225348e-05,
+      "loss": 0.2956,
+      "step": 13333
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5521972024918982e-05,
+      "loss": 0.3775,
+      "step": 13334
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5521636189612614e-05,
+      "loss": 0.3578,
+      "step": 13335
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5521300354306252e-05,
+      "loss": 0.3133,
+      "step": 13336
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5520964518999883e-05,
+      "loss": 0.3345,
+      "step": 13337
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5520628683693518e-05,
+      "loss": 0.2846,
+      "step": 13338
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5520292848387152e-05,
+      "loss": 0.3158,
+      "step": 13339
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5519957013080787e-05,
+      "loss": 0.4413,
+      "step": 13340
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.551962117777442e-05,
+      "loss": 0.2504,
+      "step": 13341
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5519285342468056e-05,
+      "loss": 0.28,
+      "step": 13342
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5518949507161687e-05,
+      "loss": 0.2484,
+      "step": 13343
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5518613671855322e-05,
+      "loss": 0.2105,
+      "step": 13344
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5518277836548957e-05,
+      "loss": 0.2577,
+      "step": 13345
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.551794200124259e-05,
+      "loss": 0.2669,
+      "step": 13346
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5517606165936226e-05,
+      "loss": 0.3899,
+      "step": 13347
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.551727033062986e-05,
+      "loss": 0.3983,
+      "step": 13348
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5516934495323495e-05,
+      "loss": 0.4202,
+      "step": 13349
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5516598660017127e-05,
+      "loss": 0.3256,
+      "step": 13350
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5516262824710765e-05,
+      "loss": 0.2335,
+      "step": 13351
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5515926989404396e-05,
+      "loss": 0.4343,
+      "step": 13352
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.551559115409803e-05,
+      "loss": 0.2024,
+      "step": 13353
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5515255318791665e-05,
+      "loss": 0.2212,
+      "step": 13354
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.55149194834853e-05,
+      "loss": 0.2477,
+      "step": 13355
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5514583648178934e-05,
+      "loss": 0.3139,
+      "step": 13356
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.551424781287257e-05,
+      "loss": 0.303,
+      "step": 13357
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5513911977566204e-05,
+      "loss": 0.3317,
+      "step": 13358
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5513576142259835e-05,
+      "loss": 0.4783,
+      "step": 13359
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5513240306953473e-05,
+      "loss": 0.355,
+      "step": 13360
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5512904471647104e-05,
+      "loss": 0.4472,
+      "step": 13361
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.551256863634074e-05,
+      "loss": 0.3712,
+      "step": 13362
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5512232801034374e-05,
+      "loss": 0.2799,
+      "step": 13363
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5511896965728008e-05,
+      "loss": 0.2806,
+      "step": 13364
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5511561130421643e-05,
+      "loss": 0.4174,
+      "step": 13365
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5511225295115278e-05,
+      "loss": 0.3355,
+      "step": 13366
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5510889459808912e-05,
+      "loss": 0.288,
+      "step": 13367
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5510553624502543e-05,
+      "loss": 0.2976,
+      "step": 13368
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.551021778919618e-05,
+      "loss": 0.2979,
+      "step": 13369
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5509881953889813e-05,
+      "loss": 0.2521,
+      "step": 13370
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5509546118583447e-05,
+      "loss": 0.3003,
+      "step": 13371
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5509210283277082e-05,
+      "loss": 0.3785,
+      "step": 13372
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5508874447970717e-05,
+      "loss": 0.2735,
+      "step": 13373
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.550853861266435e-05,
+      "loss": 0.2501,
+      "step": 13374
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5508202777357986e-05,
+      "loss": 0.3049,
+      "step": 13375
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.550786694205162e-05,
+      "loss": 0.3429,
+      "step": 13376
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5507531106745252e-05,
+      "loss": 0.4341,
+      "step": 13377
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.550719527143889e-05,
+      "loss": 0.4299,
+      "step": 13378
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.550685943613252e-05,
+      "loss": 0.2085,
+      "step": 13379
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5506523600826156e-05,
+      "loss": 0.2805,
+      "step": 13380
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.550618776551979e-05,
+      "loss": 0.39,
+      "step": 13381
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5505851930213425e-05,
+      "loss": 0.3745,
+      "step": 13382
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.550551609490706e-05,
+      "loss": 0.3325,
+      "step": 13383
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5505180259600694e-05,
+      "loss": 0.3458,
+      "step": 13384
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.550484442429433e-05,
+      "loss": 0.2299,
+      "step": 13385
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.550450858898796e-05,
+      "loss": 0.2246,
+      "step": 13386
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5504172753681595e-05,
+      "loss": 0.4601,
+      "step": 13387
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.550383691837523e-05,
+      "loss": 0.2756,
+      "step": 13388
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5503501083068864e-05,
+      "loss": 0.3266,
+      "step": 13389
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.55031652477625e-05,
+      "loss": 0.1849,
+      "step": 13390
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5502829412456134e-05,
+      "loss": 0.3269,
+      "step": 13391
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5502493577149765e-05,
+      "loss": 0.4265,
+      "step": 13392
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5502157741843403e-05,
+      "loss": 0.3275,
+      "step": 13393
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5501821906537034e-05,
+      "loss": 0.3232,
+      "step": 13394
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.550148607123067e-05,
+      "loss": 0.1845,
+      "step": 13395
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5501150235924303e-05,
+      "loss": 0.255,
+      "step": 13396
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5500814400617938e-05,
+      "loss": 0.2793,
+      "step": 13397
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5500478565311573e-05,
+      "loss": 0.2224,
+      "step": 13398
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.5500142730005207e-05,
+      "loss": 0.4601,
+      "step": 13399
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5499806894698842e-05,
+      "loss": 0.3122,
+      "step": 13400
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5499471059392473e-05,
+      "loss": 0.2725,
+      "step": 13401
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.549913522408611e-05,
+      "loss": 0.2846,
+      "step": 13402
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5498799388779743e-05,
+      "loss": 0.267,
+      "step": 13403
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5498463553473377e-05,
+      "loss": 0.2766,
+      "step": 13404
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5498127718167012e-05,
+      "loss": 0.3435,
+      "step": 13405
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5497791882860647e-05,
+      "loss": 0.1888,
+      "step": 13406
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.549745604755428e-05,
+      "loss": 0.2238,
+      "step": 13407
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5497120212247916e-05,
+      "loss": 0.3762,
+      "step": 13408
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.549678437694155e-05,
+      "loss": 0.3447,
+      "step": 13409
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5496448541635182e-05,
+      "loss": 0.3035,
+      "step": 13410
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.549611270632882e-05,
+      "loss": 0.3592,
+      "step": 13411
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.549577687102245e-05,
+      "loss": 0.2272,
+      "step": 13412
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5495441035716086e-05,
+      "loss": 0.3036,
+      "step": 13413
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.549510520040972e-05,
+      "loss": 0.2385,
+      "step": 13414
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5494769365103355e-05,
+      "loss": 0.5188,
+      "step": 13415
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.549443352979699e-05,
+      "loss": 0.288,
+      "step": 13416
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5494097694490624e-05,
+      "loss": 0.2742,
+      "step": 13417
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.549376185918426e-05,
+      "loss": 0.3785,
+      "step": 13418
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.549342602387789e-05,
+      "loss": 0.4034,
+      "step": 13419
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5493090188571528e-05,
+      "loss": 0.2619,
+      "step": 13420
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.549275435326516e-05,
+      "loss": 0.2077,
+      "step": 13421
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5492418517958794e-05,
+      "loss": 0.2229,
+      "step": 13422
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.549208268265243e-05,
+      "loss": 0.4265,
+      "step": 13423
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5491746847346063e-05,
+      "loss": 0.4199,
+      "step": 13424
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5491411012039698e-05,
+      "loss": 0.4174,
+      "step": 13425
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5491075176733333e-05,
+      "loss": 0.3731,
+      "step": 13426
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5490739341426967e-05,
+      "loss": 0.3768,
+      "step": 13427
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.54904035061206e-05,
+      "loss": 0.4664,
+      "step": 13428
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5490067670814237e-05,
+      "loss": 0.2718,
+      "step": 13429
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5489731835507868e-05,
+      "loss": 0.1981,
+      "step": 13430
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5489396000201503e-05,
+      "loss": 0.2901,
+      "step": 13431
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5489060164895137e-05,
+      "loss": 0.1868,
+      "step": 13432
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5488724329588772e-05,
+      "loss": 0.3531,
+      "step": 13433
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5488388494282403e-05,
+      "loss": 0.231,
+      "step": 13434
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.548805265897604e-05,
+      "loss": 0.2936,
+      "step": 13435
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5487716823669672e-05,
+      "loss": 0.2859,
+      "step": 13436
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5487380988363307e-05,
+      "loss": 0.2858,
+      "step": 13437
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5487045153056942e-05,
+      "loss": 0.467,
+      "step": 13438
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5486709317750576e-05,
+      "loss": 0.48,
+      "step": 13439
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.548637348244421e-05,
+      "loss": 0.2816,
+      "step": 13440
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5486037647137846e-05,
+      "loss": 0.1934,
+      "step": 13441
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.548570181183148e-05,
+      "loss": 0.3749,
+      "step": 13442
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.548536597652511e-05,
+      "loss": 0.207,
+      "step": 13443
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.548503014121875e-05,
+      "loss": 0.3156,
+      "step": 13444
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.548469430591238e-05,
+      "loss": 0.3531,
+      "step": 13445
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5484358470606015e-05,
+      "loss": 0.4416,
+      "step": 13446
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.548402263529965e-05,
+      "loss": 0.3572,
+      "step": 13447
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5483686799993285e-05,
+      "loss": 0.3908,
+      "step": 13448
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.548335096468692e-05,
+      "loss": 0.3361,
+      "step": 13449
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.548301512938055e-05,
+      "loss": 0.4036,
+      "step": 13450
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.548267929407419e-05,
+      "loss": 0.2769,
+      "step": 13451
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.548234345876782e-05,
+      "loss": 0.4611,
+      "step": 13452
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5482007623461455e-05,
+      "loss": 0.4064,
+      "step": 13453
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.548167178815509e-05,
+      "loss": 0.2256,
+      "step": 13454
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5481335952848724e-05,
+      "loss": 0.3366,
+      "step": 13455
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.548100011754236e-05,
+      "loss": 0.3949,
+      "step": 13456
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5480664282235993e-05,
+      "loss": 0.2626,
+      "step": 13457
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5480328446929628e-05,
+      "loss": 0.3025,
+      "step": 13458
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.547999261162326e-05,
+      "loss": 0.3651,
+      "step": 13459
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5479656776316897e-05,
+      "loss": 0.2496,
+      "step": 13460
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.547932094101053e-05,
+      "loss": 0.3066,
+      "step": 13461
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5478985105704163e-05,
+      "loss": 0.4142,
+      "step": 13462
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5478649270397798e-05,
+      "loss": 0.4352,
+      "step": 13463
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5478313435091432e-05,
+      "loss": 0.2781,
+      "step": 13464
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5477977599785067e-05,
+      "loss": 0.2878,
+      "step": 13465
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.54776417644787e-05,
+      "loss": 0.3633,
+      "step": 13466
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5477305929172336e-05,
+      "loss": 0.4487,
+      "step": 13467
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5476970093865968e-05,
+      "loss": 0.3268,
+      "step": 13468
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5476634258559606e-05,
+      "loss": 0.3553,
+      "step": 13469
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5476298423253237e-05,
+      "loss": 0.2649,
+      "step": 13470
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.547596258794687e-05,
+      "loss": 0.2798,
+      "step": 13471
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5475626752640506e-05,
+      "loss": 0.2348,
+      "step": 13472
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.547529091733414e-05,
+      "loss": 0.3938,
+      "step": 13473
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5474955082027775e-05,
+      "loss": 0.3716,
+      "step": 13474
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.547461924672141e-05,
+      "loss": 0.4151,
+      "step": 13475
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5474283411415045e-05,
+      "loss": 0.4088,
+      "step": 13476
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5473947576108676e-05,
+      "loss": 0.3587,
+      "step": 13477
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.547361174080231e-05,
+      "loss": 0.4883,
+      "step": 13478
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5473275905495945e-05,
+      "loss": 0.2949,
+      "step": 13479
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.547294007018958e-05,
+      "loss": 0.2573,
+      "step": 13480
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5472604234883215e-05,
+      "loss": 0.2564,
+      "step": 13481
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.547226839957685e-05,
+      "loss": 0.2594,
+      "step": 13482
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.547193256427048e-05,
+      "loss": 0.4314,
+      "step": 13483
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.547159672896412e-05,
+      "loss": 0.2336,
+      "step": 13484
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.547126089365775e-05,
+      "loss": 0.357,
+      "step": 13485
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5470925058351384e-05,
+      "loss": 0.3068,
+      "step": 13486
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.547058922304502e-05,
+      "loss": 0.3244,
+      "step": 13487
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5470253387738654e-05,
+      "loss": 0.2921,
+      "step": 13488
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.546991755243229e-05,
+      "loss": 0.3454,
+      "step": 13489
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5469581717125923e-05,
+      "loss": 0.4238,
+      "step": 13490
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5469245881819558e-05,
+      "loss": 0.2519,
+      "step": 13491
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.546891004651319e-05,
+      "loss": 0.2373,
+      "step": 13492
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5468574211206827e-05,
+      "loss": 0.3441,
+      "step": 13493
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5468238375900458e-05,
+      "loss": 0.2567,
+      "step": 13494
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5467902540594093e-05,
+      "loss": 0.3404,
+      "step": 13495
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5467566705287728e-05,
+      "loss": 0.2899,
+      "step": 13496
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5467230869981362e-05,
+      "loss": 0.4199,
+      "step": 13497
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5466895034674997e-05,
+      "loss": 0.218,
+      "step": 13498
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.546655919936863e-05,
+      "loss": 0.2664,
+      "step": 13499
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5466223364062266e-05,
+      "loss": 0.3083,
+      "step": 13500
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5465887528755897e-05,
+      "loss": 0.3383,
+      "step": 13501
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5465551693449535e-05,
+      "loss": 0.4589,
+      "step": 13502
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5465215858143167e-05,
+      "loss": 0.2514,
+      "step": 13503
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.54648800228368e-05,
+      "loss": 0.2461,
+      "step": 13504
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5464544187530436e-05,
+      "loss": 0.3304,
+      "step": 13505
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.546420835222407e-05,
+      "loss": 0.374,
+      "step": 13506
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5463872516917705e-05,
+      "loss": 0.363,
+      "step": 13507
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.546353668161134e-05,
+      "loss": 0.3412,
+      "step": 13508
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5463200846304975e-05,
+      "loss": 0.194,
+      "step": 13509
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5462865010998606e-05,
+      "loss": 0.4674,
+      "step": 13510
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5462529175692244e-05,
+      "loss": 0.3678,
+      "step": 13511
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5462193340385875e-05,
+      "loss": 0.5067,
+      "step": 13512
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.546185750507951e-05,
+      "loss": 0.287,
+      "step": 13513
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5461521669773144e-05,
+      "loss": 0.4377,
+      "step": 13514
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.546118583446678e-05,
+      "loss": 0.458,
+      "step": 13515
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5460849999160414e-05,
+      "loss": 0.4055,
+      "step": 13516
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.546051416385405e-05,
+      "loss": 0.3496,
+      "step": 13517
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5460178328547683e-05,
+      "loss": 0.3226,
+      "step": 13518
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5459842493241314e-05,
+      "loss": 0.3596,
+      "step": 13519
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5459506657934952e-05,
+      "loss": 0.5413,
+      "step": 13520
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5459170822628584e-05,
+      "loss": 0.2402,
+      "step": 13521
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5458834987322218e-05,
+      "loss": 0.2967,
+      "step": 13522
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5458499152015853e-05,
+      "loss": 0.3389,
+      "step": 13523
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5458163316709487e-05,
+      "loss": 0.2238,
+      "step": 13524
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.545782748140312e-05,
+      "loss": 0.3791,
+      "step": 13525
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5457491646096757e-05,
+      "loss": 0.2566,
+      "step": 13526
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5457155810790388e-05,
+      "loss": 0.3265,
+      "step": 13527
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5456819975484023e-05,
+      "loss": 0.3939,
+      "step": 13528
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5456484140177657e-05,
+      "loss": 0.3914,
+      "step": 13529
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5456148304871292e-05,
+      "loss": 0.2699,
+      "step": 13530
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5455812469564927e-05,
+      "loss": 0.4034,
+      "step": 13531
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.545547663425856e-05,
+      "loss": 0.3483,
+      "step": 13532
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5455140798952196e-05,
+      "loss": 0.431,
+      "step": 13533
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5454804963645827e-05,
+      "loss": 0.303,
+      "step": 13534
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5454469128339465e-05,
+      "loss": 0.3269,
+      "step": 13535
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5454133293033096e-05,
+      "loss": 0.4169,
+      "step": 13536
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.545379745772673e-05,
+      "loss": 0.301,
+      "step": 13537
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5453461622420366e-05,
+      "loss": 0.281,
+      "step": 13538
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5453125787114e-05,
+      "loss": 0.2539,
+      "step": 13539
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5452789951807635e-05,
+      "loss": 0.2957,
+      "step": 13540
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.545245411650127e-05,
+      "loss": 0.2416,
+      "step": 13541
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5452118281194904e-05,
+      "loss": 0.4168,
+      "step": 13542
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5451782445888536e-05,
+      "loss": 0.2767,
+      "step": 13543
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5451446610582174e-05,
+      "loss": 0.2755,
+      "step": 13544
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5451110775275805e-05,
+      "loss": 0.3068,
+      "step": 13545
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.545077493996944e-05,
+      "loss": 0.2072,
+      "step": 13546
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5450439104663074e-05,
+      "loss": 0.2681,
+      "step": 13547
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.545010326935671e-05,
+      "loss": 0.2888,
+      "step": 13548
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5449767434050344e-05,
+      "loss": 0.5049,
+      "step": 13549
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5449431598743978e-05,
+      "loss": 0.4636,
+      "step": 13550
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5449095763437613e-05,
+      "loss": 0.3797,
+      "step": 13551
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5448759928131244e-05,
+      "loss": 0.3699,
+      "step": 13552
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5448424092824882e-05,
+      "loss": 0.2877,
+      "step": 13553
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5448088257518513e-05,
+      "loss": 0.3724,
+      "step": 13554
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5447752422212148e-05,
+      "loss": 0.2777,
+      "step": 13555
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5447416586905783e-05,
+      "loss": 0.3855,
+      "step": 13556
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5447080751599417e-05,
+      "loss": 0.2006,
+      "step": 13557
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5446744916293052e-05,
+      "loss": 0.2572,
+      "step": 13558
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5446409080986687e-05,
+      "loss": 0.3532,
+      "step": 13559
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.544607324568032e-05,
+      "loss": 0.3293,
+      "step": 13560
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5445737410373953e-05,
+      "loss": 0.4467,
+      "step": 13561
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.544540157506759e-05,
+      "loss": 0.2934,
+      "step": 13562
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5445065739761222e-05,
+      "loss": 0.2412,
+      "step": 13563
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5444729904454856e-05,
+      "loss": 0.4585,
+      "step": 13564
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.544439406914849e-05,
+      "loss": 0.2571,
+      "step": 13565
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5444058233842126e-05,
+      "loss": 0.3569,
+      "step": 13566
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.544372239853576e-05,
+      "loss": 0.2551,
+      "step": 13567
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5443386563229395e-05,
+      "loss": 0.386,
+      "step": 13568
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5443050727923026e-05,
+      "loss": 0.3917,
+      "step": 13569
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.544271489261666e-05,
+      "loss": 0.439,
+      "step": 13570
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5442379057310296e-05,
+      "loss": 0.234,
+      "step": 13571
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.544204322200393e-05,
+      "loss": 0.2471,
+      "step": 13572
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5441707386697565e-05,
+      "loss": 0.3513,
+      "step": 13573
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.54413715513912e-05,
+      "loss": 0.4472,
+      "step": 13574
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5441035716084834e-05,
+      "loss": 0.3393,
+      "step": 13575
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5440699880778465e-05,
+      "loss": 0.2321,
+      "step": 13576
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5440364045472103e-05,
+      "loss": 0.3828,
+      "step": 13577
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5440028210165735e-05,
+      "loss": 0.2986,
+      "step": 13578
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.543969237485937e-05,
+      "loss": 0.3153,
+      "step": 13579
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5439356539553004e-05,
+      "loss": 0.189,
+      "step": 13580
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.543902070424664e-05,
+      "loss": 0.364,
+      "step": 13581
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5438684868940273e-05,
+      "loss": 0.405,
+      "step": 13582
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5438349033633908e-05,
+      "loss": 0.3156,
+      "step": 13583
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5438013198327543e-05,
+      "loss": 0.2997,
+      "step": 13584
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5437677363021174e-05,
+      "loss": 0.295,
+      "step": 13585
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5437341527714812e-05,
+      "loss": 0.291,
+      "step": 13586
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5437005692408443e-05,
+      "loss": 0.2361,
+      "step": 13587
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5436669857102078e-05,
+      "loss": 0.2362,
+      "step": 13588
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5436334021795712e-05,
+      "loss": 0.2964,
+      "step": 13589
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5435998186489347e-05,
+      "loss": 0.5058,
+      "step": 13590
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5435662351182982e-05,
+      "loss": 0.3771,
+      "step": 13591
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5435326515876616e-05,
+      "loss": 0.2211,
+      "step": 13592
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.543499068057025e-05,
+      "loss": 0.3607,
+      "step": 13593
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5434654845263882e-05,
+      "loss": 0.2456,
+      "step": 13594
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.543431900995752e-05,
+      "loss": 0.3974,
+      "step": 13595
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.543398317465115e-05,
+      "loss": 0.292,
+      "step": 13596
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.5433647339344786e-05,
+      "loss": 0.3953,
+      "step": 13597
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.543331150403842e-05,
+      "loss": 0.2674,
+      "step": 13598
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5432975668732056e-05,
+      "loss": 0.3822,
+      "step": 13599
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.543263983342569e-05,
+      "loss": 0.1397,
+      "step": 13600
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5432303998119325e-05,
+      "loss": 0.2518,
+      "step": 13601
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.543196816281296e-05,
+      "loss": 0.361,
+      "step": 13602
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.543163232750659e-05,
+      "loss": 0.3909,
+      "step": 13603
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.543129649220023e-05,
+      "loss": 0.2529,
+      "step": 13604
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.543096065689386e-05,
+      "loss": 0.3747,
+      "step": 13605
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5430624821587495e-05,
+      "loss": 0.3452,
+      "step": 13606
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.543028898628113e-05,
+      "loss": 0.3206,
+      "step": 13607
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5429953150974764e-05,
+      "loss": 0.5021,
+      "step": 13608
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.54296173156684e-05,
+      "loss": 0.2975,
+      "step": 13609
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5429281480362033e-05,
+      "loss": 0.3333,
+      "step": 13610
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5428945645055668e-05,
+      "loss": 0.2766,
+      "step": 13611
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.54286098097493e-05,
+      "loss": 0.1594,
+      "step": 13612
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5428273974442934e-05,
+      "loss": 0.2705,
+      "step": 13613
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.542793813913657e-05,
+      "loss": 0.3232,
+      "step": 13614
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5427602303830203e-05,
+      "loss": 0.345,
+      "step": 13615
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5427266468523838e-05,
+      "loss": 0.5076,
+      "step": 13616
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5426930633217472e-05,
+      "loss": 0.433,
+      "step": 13617
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5426594797911104e-05,
+      "loss": 0.3578,
+      "step": 13618
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5426258962604742e-05,
+      "loss": 0.308,
+      "step": 13619
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5425923127298373e-05,
+      "loss": 0.2142,
+      "step": 13620
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5425587291992008e-05,
+      "loss": 0.3193,
+      "step": 13621
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5425251456685642e-05,
+      "loss": 0.3613,
+      "step": 13622
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5424915621379277e-05,
+      "loss": 0.2165,
+      "step": 13623
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.542457978607291e-05,
+      "loss": 0.2869,
+      "step": 13624
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5424243950766546e-05,
+      "loss": 0.3838,
+      "step": 13625
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.542390811546018e-05,
+      "loss": 0.547,
+      "step": 13626
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5423572280153812e-05,
+      "loss": 0.4748,
+      "step": 13627
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5423236444847447e-05,
+      "loss": 0.3216,
+      "step": 13628
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.542290060954108e-05,
+      "loss": 0.4197,
+      "step": 13629
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5422564774234716e-05,
+      "loss": 0.3616,
+      "step": 13630
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.542222893892835e-05,
+      "loss": 0.3428,
+      "step": 13631
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5421893103621985e-05,
+      "loss": 0.2077,
+      "step": 13632
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.542155726831562e-05,
+      "loss": 0.2533,
+      "step": 13633
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.542122143300925e-05,
+      "loss": 0.2932,
+      "step": 13634
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.542088559770289e-05,
+      "loss": 0.2208,
+      "step": 13635
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.542054976239652e-05,
+      "loss": 0.3019,
+      "step": 13636
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5420213927090155e-05,
+      "loss": 0.3413,
+      "step": 13637
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.541987809178379e-05,
+      "loss": 0.2743,
+      "step": 13638
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5419542256477425e-05,
+      "loss": 0.1879,
+      "step": 13639
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.541920642117106e-05,
+      "loss": 0.3006,
+      "step": 13640
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5418870585864694e-05,
+      "loss": 0.378,
+      "step": 13641
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.541853475055833e-05,
+      "loss": 0.3929,
+      "step": 13642
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.541819891525196e-05,
+      "loss": 0.2783,
+      "step": 13643
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5417863079945598e-05,
+      "loss": 0.3973,
+      "step": 13644
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.541752724463923e-05,
+      "loss": 0.3455,
+      "step": 13645
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5417191409332864e-05,
+      "loss": 0.5584,
+      "step": 13646
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5416855574026498e-05,
+      "loss": 0.407,
+      "step": 13647
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5416519738720133e-05,
+      "loss": 0.3983,
+      "step": 13648
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5416183903413768e-05,
+      "loss": 0.1904,
+      "step": 13649
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5415848068107402e-05,
+      "loss": 0.3765,
+      "step": 13650
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5415512232801037e-05,
+      "loss": 0.3796,
+      "step": 13651
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5415176397494668e-05,
+      "loss": 0.3033,
+      "step": 13652
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5414840562188306e-05,
+      "loss": 0.3096,
+      "step": 13653
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5414504726881937e-05,
+      "loss": 0.3927,
+      "step": 13654
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5414168891575572e-05,
+      "loss": 0.3566,
+      "step": 13655
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5413833056269207e-05,
+      "loss": 0.4399,
+      "step": 13656
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.541349722096284e-05,
+      "loss": 0.196,
+      "step": 13657
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5413161385656476e-05,
+      "loss": 0.2197,
+      "step": 13658
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.541282555035011e-05,
+      "loss": 0.4112,
+      "step": 13659
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5412489715043742e-05,
+      "loss": 0.2304,
+      "step": 13660
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5412153879737377e-05,
+      "loss": 0.2707,
+      "step": 13661
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.541181804443101e-05,
+      "loss": 0.3317,
+      "step": 13662
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5411482209124646e-05,
+      "loss": 0.4102,
+      "step": 13663
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.541114637381828e-05,
+      "loss": 0.4214,
+      "step": 13664
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5410810538511915e-05,
+      "loss": 0.2686,
+      "step": 13665
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.541047470320555e-05,
+      "loss": 0.2343,
+      "step": 13666
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.541013886789918e-05,
+      "loss": 0.301,
+      "step": 13667
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.540980303259282e-05,
+      "loss": 0.4015,
+      "step": 13668
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.540946719728645e-05,
+      "loss": 0.4171,
+      "step": 13669
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5409131361980085e-05,
+      "loss": 0.3137,
+      "step": 13670
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.540879552667372e-05,
+      "loss": 0.3961,
+      "step": 13671
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5408459691367354e-05,
+      "loss": 0.3553,
+      "step": 13672
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.540812385606099e-05,
+      "loss": 0.3445,
+      "step": 13673
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5407788020754624e-05,
+      "loss": 0.1511,
+      "step": 13674
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5407452185448258e-05,
+      "loss": 0.417,
+      "step": 13675
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.540711635014189e-05,
+      "loss": 0.2529,
+      "step": 13676
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5406780514835528e-05,
+      "loss": 0.3776,
+      "step": 13677
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.540644467952916e-05,
+      "loss": 0.2614,
+      "step": 13678
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5406108844222793e-05,
+      "loss": 0.2671,
+      "step": 13679
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5405773008916428e-05,
+      "loss": 0.2904,
+      "step": 13680
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5405437173610063e-05,
+      "loss": 0.3207,
+      "step": 13681
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5405101338303697e-05,
+      "loss": 0.3204,
+      "step": 13682
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5404765502997332e-05,
+      "loss": 0.3185,
+      "step": 13683
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5404429667690967e-05,
+      "loss": 0.1658,
+      "step": 13684
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5404093832384598e-05,
+      "loss": 0.3318,
+      "step": 13685
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5403757997078236e-05,
+      "loss": 0.4487,
+      "step": 13686
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5403422161771867e-05,
+      "loss": 0.3989,
+      "step": 13687
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5403086326465502e-05,
+      "loss": 0.2178,
+      "step": 13688
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5402750491159137e-05,
+      "loss": 0.4402,
+      "step": 13689
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.540241465585277e-05,
+      "loss": 0.3221,
+      "step": 13690
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5402078820546406e-05,
+      "loss": 0.3294,
+      "step": 13691
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.540174298524004e-05,
+      "loss": 0.2238,
+      "step": 13692
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5401407149933675e-05,
+      "loss": 0.3532,
+      "step": 13693
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5401071314627306e-05,
+      "loss": 0.2563,
+      "step": 13694
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5400735479320944e-05,
+      "loss": 0.2648,
+      "step": 13695
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5400399644014576e-05,
+      "loss": 0.3989,
+      "step": 13696
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.540006380870821e-05,
+      "loss": 0.2905,
+      "step": 13697
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5399727973401845e-05,
+      "loss": 0.2091,
+      "step": 13698
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.539939213809548e-05,
+      "loss": 0.4745,
+      "step": 13699
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5399056302789114e-05,
+      "loss": 0.2196,
+      "step": 13700
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.539872046748275e-05,
+      "loss": 0.5669,
+      "step": 13701
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5398384632176384e-05,
+      "loss": 0.561,
+      "step": 13702
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5398048796870015e-05,
+      "loss": 0.2905,
+      "step": 13703
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.539771296156365e-05,
+      "loss": 0.3953,
+      "step": 13704
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5397377126257284e-05,
+      "loss": 0.4949,
+      "step": 13705
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.539704129095092e-05,
+      "loss": 0.3139,
+      "step": 13706
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5396705455644553e-05,
+      "loss": 0.2051,
+      "step": 13707
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5396369620338188e-05,
+      "loss": 0.3495,
+      "step": 13708
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.539603378503182e-05,
+      "loss": 0.3267,
+      "step": 13709
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5395697949725457e-05,
+      "loss": 0.2172,
+      "step": 13710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.539536211441909e-05,
+      "loss": 0.2722,
+      "step": 13711
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5395026279112723e-05,
+      "loss": 0.2673,
+      "step": 13712
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5394690443806358e-05,
+      "loss": 0.3319,
+      "step": 13713
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5394354608499993e-05,
+      "loss": 0.4922,
+      "step": 13714
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5394018773193627e-05,
+      "loss": 0.3442,
+      "step": 13715
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5393682937887262e-05,
+      "loss": 0.2568,
+      "step": 13716
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5393347102580897e-05,
+      "loss": 0.2736,
+      "step": 13717
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5393011267274528e-05,
+      "loss": 0.4595,
+      "step": 13718
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5392675431968166e-05,
+      "loss": 0.2922,
+      "step": 13719
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5392339596661797e-05,
+      "loss": 0.3548,
+      "step": 13720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5392003761355432e-05,
+      "loss": 0.3046,
+      "step": 13721
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5391667926049066e-05,
+      "loss": 0.4649,
+      "step": 13722
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.53913320907427e-05,
+      "loss": 0.1567,
+      "step": 13723
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5390996255436336e-05,
+      "loss": 0.2398,
+      "step": 13724
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.539066042012997e-05,
+      "loss": 0.3552,
+      "step": 13725
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5390324584823605e-05,
+      "loss": 0.274,
+      "step": 13726
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5389988749517236e-05,
+      "loss": 0.2954,
+      "step": 13727
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5389652914210874e-05,
+      "loss": 0.3342,
+      "step": 13728
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5389317078904506e-05,
+      "loss": 0.4641,
+      "step": 13729
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.538898124359814e-05,
+      "loss": 0.2831,
+      "step": 13730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5388645408291775e-05,
+      "loss": 0.2781,
+      "step": 13731
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.538830957298541e-05,
+      "loss": 0.2544,
+      "step": 13732
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5387973737679044e-05,
+      "loss": 0.3506,
+      "step": 13733
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.538763790237268e-05,
+      "loss": 0.2327,
+      "step": 13734
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5387302067066313e-05,
+      "loss": 0.3206,
+      "step": 13735
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5386966231759945e-05,
+      "loss": 0.2616,
+      "step": 13736
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5386630396453583e-05,
+      "loss": 0.2791,
+      "step": 13737
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5386294561147214e-05,
+      "loss": 0.1981,
+      "step": 13738
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.538595872584085e-05,
+      "loss": 0.2599,
+      "step": 13739
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5385622890534483e-05,
+      "loss": 0.149,
+      "step": 13740
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5385287055228118e-05,
+      "loss": 0.3962,
+      "step": 13741
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5384951219921753e-05,
+      "loss": 0.3862,
+      "step": 13742
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5384615384615387e-05,
+      "loss": 0.4596,
+      "step": 13743
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5384279549309022e-05,
+      "loss": 0.4208,
+      "step": 13744
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5383943714002653e-05,
+      "loss": 0.3739,
+      "step": 13745
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.538360787869629e-05,
+      "loss": 0.4742,
+      "step": 13746
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5383272043389922e-05,
+      "loss": 0.3461,
+      "step": 13747
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5382936208083557e-05,
+      "loss": 0.2538,
+      "step": 13748
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.538260037277719e-05,
+      "loss": 0.1615,
+      "step": 13749
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5382264537470826e-05,
+      "loss": 0.2132,
+      "step": 13750
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5381928702164458e-05,
+      "loss": 0.2465,
+      "step": 13751
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5381592866858096e-05,
+      "loss": 0.3445,
+      "step": 13752
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5381257031551727e-05,
+      "loss": 0.2198,
+      "step": 13753
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.538092119624536e-05,
+      "loss": 0.374,
+      "step": 13754
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5380585360938996e-05,
+      "loss": 0.1464,
+      "step": 13755
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.538024952563263e-05,
+      "loss": 0.2686,
+      "step": 13756
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5379913690326265e-05,
+      "loss": 0.3732,
+      "step": 13757
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.53795778550199e-05,
+      "loss": 0.2288,
+      "step": 13758
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5379242019713535e-05,
+      "loss": 0.3917,
+      "step": 13759
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5378906184407166e-05,
+      "loss": 0.3061,
+      "step": 13760
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5378570349100804e-05,
+      "loss": 0.3558,
+      "step": 13761
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5378234513794435e-05,
+      "loss": 0.2099,
+      "step": 13762
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.537789867848807e-05,
+      "loss": 0.2089,
+      "step": 13763
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5377562843181705e-05,
+      "loss": 0.3489,
+      "step": 13764
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.537722700787534e-05,
+      "loss": 0.3433,
+      "step": 13765
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5376891172568974e-05,
+      "loss": 0.2516,
+      "step": 13766
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.537655533726261e-05,
+      "loss": 0.4545,
+      "step": 13767
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5376219501956243e-05,
+      "loss": 0.2555,
+      "step": 13768
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5375883666649874e-05,
+      "loss": 0.2506,
+      "step": 13769
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5375547831343512e-05,
+      "loss": 0.2662,
+      "step": 13770
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5375211996037144e-05,
+      "loss": 0.2427,
+      "step": 13771
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.537487616073078e-05,
+      "loss": 0.4147,
+      "step": 13772
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5374540325424413e-05,
+      "loss": 0.2439,
+      "step": 13773
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5374204490118048e-05,
+      "loss": 0.2239,
+      "step": 13774
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5373868654811682e-05,
+      "loss": 0.2897,
+      "step": 13775
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5373532819505317e-05,
+      "loss": 0.3004,
+      "step": 13776
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.537319698419895e-05,
+      "loss": 0.3347,
+      "step": 13777
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5372861148892583e-05,
+      "loss": 0.1874,
+      "step": 13778
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.537252531358622e-05,
+      "loss": 0.375,
+      "step": 13779
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5372189478279852e-05,
+      "loss": 0.4398,
+      "step": 13780
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5371853642973487e-05,
+      "loss": 0.301,
+      "step": 13781
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.537151780766712e-05,
+      "loss": 0.3349,
+      "step": 13782
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5371181972360756e-05,
+      "loss": 0.176,
+      "step": 13783
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.537084613705439e-05,
+      "loss": 0.5382,
+      "step": 13784
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5370510301748025e-05,
+      "loss": 0.2542,
+      "step": 13785
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.537017446644166e-05,
+      "loss": 0.2393,
+      "step": 13786
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.536983863113529e-05,
+      "loss": 0.4162,
+      "step": 13787
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.536950279582893e-05,
+      "loss": 0.4057,
+      "step": 13788
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.536916696052256e-05,
+      "loss": 0.3278,
+      "step": 13789
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5368831125216195e-05,
+      "loss": 0.3273,
+      "step": 13790
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.536849528990983e-05,
+      "loss": 0.4002,
+      "step": 13791
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5368159454603465e-05,
+      "loss": 0.3586,
+      "step": 13792
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.53678236192971e-05,
+      "loss": 0.3083,
+      "step": 13793
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5367487783990734e-05,
+      "loss": 0.2991,
+      "step": 13794
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5367151948684365e-05,
+      "loss": 0.2501,
+      "step": 13795
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.5366816113378e-05,
+      "loss": 0.4115,
+      "step": 13796
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5366480278071634e-05,
+      "loss": 0.3875,
+      "step": 13797
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.536614444276527e-05,
+      "loss": 0.3534,
+      "step": 13798
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5365808607458904e-05,
+      "loss": 0.3233,
+      "step": 13799
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.536547277215254e-05,
+      "loss": 0.2199,
+      "step": 13800
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5365136936846173e-05,
+      "loss": 0.2544,
+      "step": 13801
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5364801101539804e-05,
+      "loss": 0.3561,
+      "step": 13802
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5364465266233442e-05,
+      "loss": 0.5416,
+      "step": 13803
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5364129430927074e-05,
+      "loss": 0.2626,
+      "step": 13804
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5363793595620708e-05,
+      "loss": 0.2902,
+      "step": 13805
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5363457760314343e-05,
+      "loss": 0.5491,
+      "step": 13806
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5363121925007978e-05,
+      "loss": 0.312,
+      "step": 13807
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5362786089701612e-05,
+      "loss": 0.4108,
+      "step": 13808
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5362450254395243e-05,
+      "loss": 0.3017,
+      "step": 13809
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.536211441908888e-05,
+      "loss": 0.4817,
+      "step": 13810
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5361778583782513e-05,
+      "loss": 0.3044,
+      "step": 13811
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5361442748476147e-05,
+      "loss": 0.3591,
+      "step": 13812
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5361106913169782e-05,
+      "loss": 0.3204,
+      "step": 13813
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5360771077863417e-05,
+      "loss": 0.2664,
+      "step": 13814
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.536043524255705e-05,
+      "loss": 0.3175,
+      "step": 13815
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5360099407250686e-05,
+      "loss": 0.3464,
+      "step": 13816
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.535976357194432e-05,
+      "loss": 0.2905,
+      "step": 13817
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5359427736637952e-05,
+      "loss": 0.3208,
+      "step": 13818
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.535909190133159e-05,
+      "loss": 0.4219,
+      "step": 13819
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.535875606602522e-05,
+      "loss": 0.2883,
+      "step": 13820
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5358420230718856e-05,
+      "loss": 0.2893,
+      "step": 13821
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.535808439541249e-05,
+      "loss": 0.2748,
+      "step": 13822
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5357748560106125e-05,
+      "loss": 0.3342,
+      "step": 13823
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.535741272479976e-05,
+      "loss": 0.2368,
+      "step": 13824
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5357076889493394e-05,
+      "loss": 0.2135,
+      "step": 13825
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.535674105418703e-05,
+      "loss": 0.2152,
+      "step": 13826
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.535640521888066e-05,
+      "loss": 0.3154,
+      "step": 13827
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.53560693835743e-05,
+      "loss": 0.2607,
+      "step": 13828
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.535573354826793e-05,
+      "loss": 0.3693,
+      "step": 13829
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5355397712961564e-05,
+      "loss": 0.3304,
+      "step": 13830
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.53550618776552e-05,
+      "loss": 0.4862,
+      "step": 13831
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5354726042348834e-05,
+      "loss": 0.3053,
+      "step": 13832
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5354390207042468e-05,
+      "loss": 0.3486,
+      "step": 13833
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5354054371736103e-05,
+      "loss": 0.3442,
+      "step": 13834
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5353718536429737e-05,
+      "loss": 0.4384,
+      "step": 13835
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.535338270112337e-05,
+      "loss": 0.3872,
+      "step": 13836
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5353046865817007e-05,
+      "loss": 0.2102,
+      "step": 13837
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5352711030510638e-05,
+      "loss": 0.2045,
+      "step": 13838
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5352375195204273e-05,
+      "loss": 0.301,
+      "step": 13839
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5352039359897907e-05,
+      "loss": 0.1787,
+      "step": 13840
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5351703524591542e-05,
+      "loss": 0.2794,
+      "step": 13841
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5351367689285173e-05,
+      "loss": 0.3938,
+      "step": 13842
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.535103185397881e-05,
+      "loss": 0.4597,
+      "step": 13843
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5350696018672443e-05,
+      "loss": 0.4438,
+      "step": 13844
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5350360183366077e-05,
+      "loss": 0.386,
+      "step": 13845
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5350024348059712e-05,
+      "loss": 0.3275,
+      "step": 13846
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5349688512753346e-05,
+      "loss": 0.2351,
+      "step": 13847
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.534935267744698e-05,
+      "loss": 0.2328,
+      "step": 13848
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5349016842140616e-05,
+      "loss": 0.172,
+      "step": 13849
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.534868100683425e-05,
+      "loss": 0.3473,
+      "step": 13850
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.534834517152788e-05,
+      "loss": 0.3429,
+      "step": 13851
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.534800933622152e-05,
+      "loss": 0.3829,
+      "step": 13852
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.534767350091515e-05,
+      "loss": 0.567,
+      "step": 13853
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5347337665608786e-05,
+      "loss": 0.4161,
+      "step": 13854
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.534700183030242e-05,
+      "loss": 0.2387,
+      "step": 13855
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5346665994996055e-05,
+      "loss": 0.3079,
+      "step": 13856
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.534633015968969e-05,
+      "loss": 0.4116,
+      "step": 13857
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5345994324383324e-05,
+      "loss": 0.4801,
+      "step": 13858
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.534565848907696e-05,
+      "loss": 0.2299,
+      "step": 13859
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.534532265377059e-05,
+      "loss": 0.246,
+      "step": 13860
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5344986818464228e-05,
+      "loss": 0.2445,
+      "step": 13861
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.534465098315786e-05,
+      "loss": 0.2989,
+      "step": 13862
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5344315147851494e-05,
+      "loss": 0.2126,
+      "step": 13863
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.534397931254513e-05,
+      "loss": 0.369,
+      "step": 13864
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5343643477238763e-05,
+      "loss": 0.2781,
+      "step": 13865
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5343307641932398e-05,
+      "loss": 0.243,
+      "step": 13866
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5342971806626033e-05,
+      "loss": 0.3606,
+      "step": 13867
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5342635971319667e-05,
+      "loss": 0.5064,
+      "step": 13868
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.53423001360133e-05,
+      "loss": 0.2846,
+      "step": 13869
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5341964300706937e-05,
+      "loss": 0.1929,
+      "step": 13870
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5341628465400568e-05,
+      "loss": 0.4115,
+      "step": 13871
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5341292630094202e-05,
+      "loss": 0.5092,
+      "step": 13872
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5340956794787837e-05,
+      "loss": 0.307,
+      "step": 13873
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5340620959481472e-05,
+      "loss": 0.247,
+      "step": 13874
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5340285124175106e-05,
+      "loss": 0.236,
+      "step": 13875
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.533994928886874e-05,
+      "loss": 0.5472,
+      "step": 13876
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5339613453562376e-05,
+      "loss": 0.2974,
+      "step": 13877
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5339277618256007e-05,
+      "loss": 0.4,
+      "step": 13878
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5338941782949645e-05,
+      "loss": 0.2661,
+      "step": 13879
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5338605947643276e-05,
+      "loss": 0.3969,
+      "step": 13880
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.533827011233691e-05,
+      "loss": 0.2769,
+      "step": 13881
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5337934277030546e-05,
+      "loss": 0.2285,
+      "step": 13882
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.533759844172418e-05,
+      "loss": 0.2799,
+      "step": 13883
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5337262606417815e-05,
+      "loss": 0.2571,
+      "step": 13884
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.533692677111145e-05,
+      "loss": 0.35,
+      "step": 13885
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.533659093580508e-05,
+      "loss": 0.4508,
+      "step": 13886
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5336255100498715e-05,
+      "loss": 0.4368,
+      "step": 13887
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.533591926519235e-05,
+      "loss": 0.3827,
+      "step": 13888
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5335583429885985e-05,
+      "loss": 0.1321,
+      "step": 13889
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.533524759457962e-05,
+      "loss": 0.297,
+      "step": 13890
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5334911759273254e-05,
+      "loss": 0.2148,
+      "step": 13891
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.533457592396689e-05,
+      "loss": 0.3609,
+      "step": 13892
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.533424008866052e-05,
+      "loss": 0.2399,
+      "step": 13893
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5333904253354158e-05,
+      "loss": 0.3165,
+      "step": 13894
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.533356841804779e-05,
+      "loss": 0.243,
+      "step": 13895
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5333232582741424e-05,
+      "loss": 0.3846,
+      "step": 13896
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.533289674743506e-05,
+      "loss": 0.3059,
+      "step": 13897
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5332560912128693e-05,
+      "loss": 0.2445,
+      "step": 13898
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5332225076822328e-05,
+      "loss": 0.4798,
+      "step": 13899
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5331889241515962e-05,
+      "loss": 0.2463,
+      "step": 13900
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5331553406209597e-05,
+      "loss": 0.2359,
+      "step": 13901
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.533121757090323e-05,
+      "loss": 0.416,
+      "step": 13902
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5330881735596866e-05,
+      "loss": 0.3069,
+      "step": 13903
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5330545900290498e-05,
+      "loss": 0.2828,
+      "step": 13904
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5330210064984132e-05,
+      "loss": 0.418,
+      "step": 13905
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5329874229677767e-05,
+      "loss": 0.2929,
+      "step": 13906
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.53295383943714e-05,
+      "loss": 0.2506,
+      "step": 13907
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5329202559065036e-05,
+      "loss": 0.3474,
+      "step": 13908
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.532886672375867e-05,
+      "loss": 0.5183,
+      "step": 13909
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5328530888452306e-05,
+      "loss": 0.2994,
+      "step": 13910
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5328195053145937e-05,
+      "loss": 0.2958,
+      "step": 13911
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5327859217839575e-05,
+      "loss": 0.3642,
+      "step": 13912
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5327523382533206e-05,
+      "loss": 0.4183,
+      "step": 13913
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.532718754722684e-05,
+      "loss": 0.2313,
+      "step": 13914
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5326851711920475e-05,
+      "loss": 0.3386,
+      "step": 13915
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.532651587661411e-05,
+      "loss": 0.3332,
+      "step": 13916
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5326180041307745e-05,
+      "loss": 0.2621,
+      "step": 13917
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.532584420600138e-05,
+      "loss": 0.3531,
+      "step": 13918
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5325508370695014e-05,
+      "loss": 0.21,
+      "step": 13919
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5325172535388645e-05,
+      "loss": 0.4316,
+      "step": 13920
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5324836700082283e-05,
+      "loss": 0.322,
+      "step": 13921
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5324500864775915e-05,
+      "loss": 0.7215,
+      "step": 13922
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.532416502946955e-05,
+      "loss": 0.309,
+      "step": 13923
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5323829194163184e-05,
+      "loss": 0.2622,
+      "step": 13924
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.532349335885682e-05,
+      "loss": 0.3302,
+      "step": 13925
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5323157523550453e-05,
+      "loss": 0.2989,
+      "step": 13926
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5322821688244088e-05,
+      "loss": 0.2699,
+      "step": 13927
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5322485852937722e-05,
+      "loss": 0.2936,
+      "step": 13928
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5322150017631354e-05,
+      "loss": 0.2855,
+      "step": 13929
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.532181418232499e-05,
+      "loss": 0.3086,
+      "step": 13930
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5321478347018623e-05,
+      "loss": 0.3047,
+      "step": 13931
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5321142511712258e-05,
+      "loss": 0.2651,
+      "step": 13932
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5320806676405892e-05,
+      "loss": 0.1291,
+      "step": 13933
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5320470841099527e-05,
+      "loss": 0.2771,
+      "step": 13934
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5320135005793158e-05,
+      "loss": 0.2155,
+      "step": 13935
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5319799170486796e-05,
+      "loss": 0.3406,
+      "step": 13936
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5319463335180427e-05,
+      "loss": 0.3413,
+      "step": 13937
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5319127499874062e-05,
+      "loss": 0.227,
+      "step": 13938
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5318791664567697e-05,
+      "loss": 0.3603,
+      "step": 13939
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.531845582926133e-05,
+      "loss": 0.2184,
+      "step": 13940
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5318119993954966e-05,
+      "loss": 0.2839,
+      "step": 13941
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.53177841586486e-05,
+      "loss": 0.5157,
+      "step": 13942
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5317448323342235e-05,
+      "loss": 0.3823,
+      "step": 13943
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5317112488035867e-05,
+      "loss": 0.1659,
+      "step": 13944
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5316776652729505e-05,
+      "loss": 0.2837,
+      "step": 13945
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5316440817423136e-05,
+      "loss": 0.2594,
+      "step": 13946
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.531610498211677e-05,
+      "loss": 0.4293,
+      "step": 13947
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5315769146810405e-05,
+      "loss": 0.4225,
+      "step": 13948
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.531543331150404e-05,
+      "loss": 0.3969,
+      "step": 13949
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5315097476197674e-05,
+      "loss": 0.2857,
+      "step": 13950
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.531476164089131e-05,
+      "loss": 0.396,
+      "step": 13951
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5314425805584944e-05,
+      "loss": 0.2599,
+      "step": 13952
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5314089970278575e-05,
+      "loss": 0.1851,
+      "step": 13953
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5313754134972213e-05,
+      "loss": 0.2935,
+      "step": 13954
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5313418299665844e-05,
+      "loss": 0.3678,
+      "step": 13955
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.531308246435948e-05,
+      "loss": 0.2759,
+      "step": 13956
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5312746629053114e-05,
+      "loss": 0.2778,
+      "step": 13957
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5312410793746748e-05,
+      "loss": 0.3262,
+      "step": 13958
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5312074958440383e-05,
+      "loss": 0.3068,
+      "step": 13959
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5311739123134018e-05,
+      "loss": 0.1163,
+      "step": 13960
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5311403287827652e-05,
+      "loss": 0.3383,
+      "step": 13961
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5311067452521283e-05,
+      "loss": 0.2682,
+      "step": 13962
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.531073161721492e-05,
+      "loss": 0.3553,
+      "step": 13963
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5310395781908553e-05,
+      "loss": 0.22,
+      "step": 13964
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5310059946602187e-05,
+      "loss": 0.2173,
+      "step": 13965
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5309724111295822e-05,
+      "loss": 0.3449,
+      "step": 13966
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5309388275989457e-05,
+      "loss": 0.319,
+      "step": 13967
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.530905244068309e-05,
+      "loss": 0.3988,
+      "step": 13968
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5308716605376726e-05,
+      "loss": 0.3052,
+      "step": 13969
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.530838077007036e-05,
+      "loss": 0.2994,
+      "step": 13970
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5308044934763992e-05,
+      "loss": 0.3642,
+      "step": 13971
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.530770909945763e-05,
+      "loss": 0.4024,
+      "step": 13972
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.530737326415126e-05,
+      "loss": 0.2408,
+      "step": 13973
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5307037428844896e-05,
+      "loss": 0.4259,
+      "step": 13974
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.530670159353853e-05,
+      "loss": 0.432,
+      "step": 13975
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5306365758232165e-05,
+      "loss": 0.441,
+      "step": 13976
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5306029922925796e-05,
+      "loss": 0.2923,
+      "step": 13977
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5305694087619434e-05,
+      "loss": 0.4211,
+      "step": 13978
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5305358252313066e-05,
+      "loss": 0.2722,
+      "step": 13979
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.53050224170067e-05,
+      "loss": 0.2287,
+      "step": 13980
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5304686581700335e-05,
+      "loss": 0.3232,
+      "step": 13981
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.530435074639397e-05,
+      "loss": 0.2373,
+      "step": 13982
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5304014911087604e-05,
+      "loss": 0.3462,
+      "step": 13983
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.530367907578124e-05,
+      "loss": 0.3372,
+      "step": 13984
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5303343240474874e-05,
+      "loss": 0.21,
+      "step": 13985
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5303007405168505e-05,
+      "loss": 0.4374,
+      "step": 13986
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.530267156986214e-05,
+      "loss": 0.1836,
+      "step": 13987
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5302335734555774e-05,
+      "loss": 0.4648,
+      "step": 13988
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.530199989924941e-05,
+      "loss": 0.3329,
+      "step": 13989
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5301664063943043e-05,
+      "loss": 0.3501,
+      "step": 13990
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5301328228636678e-05,
+      "loss": 0.3233,
+      "step": 13991
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5300992393330313e-05,
+      "loss": 0.4047,
+      "step": 13992
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5300656558023944e-05,
+      "loss": 0.3008,
+      "step": 13993
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.5300320722717582e-05,
+      "loss": 0.4111,
+      "step": 13994
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5299984887411213e-05,
+      "loss": 0.3438,
+      "step": 13995
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5299649052104848e-05,
+      "loss": 0.3126,
+      "step": 13996
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5299313216798483e-05,
+      "loss": 0.3985,
+      "step": 13997
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5298977381492117e-05,
+      "loss": 0.3323,
+      "step": 13998
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5298641546185752e-05,
+      "loss": 0.5397,
+      "step": 13999
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5298305710879387e-05,
+      "loss": 0.2904,
+      "step": 14000
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.529796987557302e-05,
+      "loss": 0.2528,
+      "step": 14001
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5297634040266652e-05,
+      "loss": 0.2224,
+      "step": 14002
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.529729820496029e-05,
+      "loss": 0.4918,
+      "step": 14003
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5296962369653922e-05,
+      "loss": 0.4519,
+      "step": 14004
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5296626534347556e-05,
+      "loss": 0.3585,
+      "step": 14005
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.529629069904119e-05,
+      "loss": 0.3545,
+      "step": 14006
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5295954863734826e-05,
+      "loss": 0.1865,
+      "step": 14007
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.529561902842846e-05,
+      "loss": 0.2968,
+      "step": 14008
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5295283193122095e-05,
+      "loss": 0.2682,
+      "step": 14009
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.529494735781573e-05,
+      "loss": 0.3739,
+      "step": 14010
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.529461152250936e-05,
+      "loss": 0.356,
+      "step": 14011
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5294275687203e-05,
+      "loss": 0.2871,
+      "step": 14012
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.529393985189663e-05,
+      "loss": 0.3728,
+      "step": 14013
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5293604016590265e-05,
+      "loss": 0.4004,
+      "step": 14014
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.52932681812839e-05,
+      "loss": 0.3795,
+      "step": 14015
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5292932345977534e-05,
+      "loss": 0.2824,
+      "step": 14016
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.529259651067117e-05,
+      "loss": 0.488,
+      "step": 14017
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5292260675364803e-05,
+      "loss": 0.4179,
+      "step": 14018
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5291924840058438e-05,
+      "loss": 0.321,
+      "step": 14019
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.529158900475207e-05,
+      "loss": 0.4837,
+      "step": 14020
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5291253169445704e-05,
+      "loss": 0.3634,
+      "step": 14021
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.529091733413934e-05,
+      "loss": 0.3783,
+      "step": 14022
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5290581498832973e-05,
+      "loss": 0.2598,
+      "step": 14023
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5290245663526608e-05,
+      "loss": 0.4501,
+      "step": 14024
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5289909828220243e-05,
+      "loss": 0.235,
+      "step": 14025
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5289573992913874e-05,
+      "loss": 0.3661,
+      "step": 14026
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5289238157607512e-05,
+      "loss": 0.2232,
+      "step": 14027
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5288902322301143e-05,
+      "loss": 0.2814,
+      "step": 14028
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5288566486994778e-05,
+      "loss": 0.4442,
+      "step": 14029
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5288230651688412e-05,
+      "loss": 0.3349,
+      "step": 14030
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5287894816382047e-05,
+      "loss": 0.2394,
+      "step": 14031
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.528755898107568e-05,
+      "loss": 0.2777,
+      "step": 14032
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5287223145769316e-05,
+      "loss": 0.4822,
+      "step": 14033
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.528688731046295e-05,
+      "loss": 0.4186,
+      "step": 14034
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5286551475156582e-05,
+      "loss": 0.298,
+      "step": 14035
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.528621563985022e-05,
+      "loss": 0.2488,
+      "step": 14036
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.528587980454385e-05,
+      "loss": 0.3362,
+      "step": 14037
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5285543969237486e-05,
+      "loss": 0.2575,
+      "step": 14038
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.528520813393112e-05,
+      "loss": 0.4546,
+      "step": 14039
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5284872298624755e-05,
+      "loss": 0.4603,
+      "step": 14040
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.528453646331839e-05,
+      "loss": 0.3631,
+      "step": 14041
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5284200628012025e-05,
+      "loss": 0.2974,
+      "step": 14042
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.528386479270566e-05,
+      "loss": 0.2311,
+      "step": 14043
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.528352895739929e-05,
+      "loss": 0.27,
+      "step": 14044
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.528319312209293e-05,
+      "loss": 0.3426,
+      "step": 14045
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.528285728678656e-05,
+      "loss": 0.2068,
+      "step": 14046
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5282521451480195e-05,
+      "loss": 0.2094,
+      "step": 14047
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.528218561617383e-05,
+      "loss": 0.2716,
+      "step": 14048
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5281849780867464e-05,
+      "loss": 0.3584,
+      "step": 14049
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.52815139455611e-05,
+      "loss": 0.1786,
+      "step": 14050
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5281178110254733e-05,
+      "loss": 0.1431,
+      "step": 14051
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5280842274948368e-05,
+      "loss": 0.4916,
+      "step": 14052
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5280506439642e-05,
+      "loss": 0.1945,
+      "step": 14053
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5280170604335637e-05,
+      "loss": 0.4851,
+      "step": 14054
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527983476902927e-05,
+      "loss": 0.2436,
+      "step": 14055
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5279498933722903e-05,
+      "loss": 0.3216,
+      "step": 14056
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5279163098416538e-05,
+      "loss": 0.5674,
+      "step": 14057
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5278827263110172e-05,
+      "loss": 0.3432,
+      "step": 14058
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5278491427803807e-05,
+      "loss": 0.1829,
+      "step": 14059
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527815559249744e-05,
+      "loss": 0.547,
+      "step": 14060
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5277819757191076e-05,
+      "loss": 0.3975,
+      "step": 14061
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5277483921884708e-05,
+      "loss": 0.2197,
+      "step": 14062
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5277148086578346e-05,
+      "loss": 0.3217,
+      "step": 14063
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5276812251271977e-05,
+      "loss": 0.2798,
+      "step": 14064
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527647641596561e-05,
+      "loss": 0.3879,
+      "step": 14065
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5276140580659246e-05,
+      "loss": 0.2753,
+      "step": 14066
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527580474535288e-05,
+      "loss": 0.3131,
+      "step": 14067
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5275468910046512e-05,
+      "loss": 0.4858,
+      "step": 14068
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527513307474015e-05,
+      "loss": 0.3613,
+      "step": 14069
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527479723943378e-05,
+      "loss": 0.3366,
+      "step": 14070
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5274461404127416e-05,
+      "loss": 0.3529,
+      "step": 14071
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527412556882105e-05,
+      "loss": 0.2756,
+      "step": 14072
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5273789733514685e-05,
+      "loss": 0.3529,
+      "step": 14073
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527345389820832e-05,
+      "loss": 0.3334,
+      "step": 14074
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5273118062901955e-05,
+      "loss": 0.2717,
+      "step": 14075
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527278222759559e-05,
+      "loss": 0.2666,
+      "step": 14076
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527244639228922e-05,
+      "loss": 0.2878,
+      "step": 14077
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527211055698286e-05,
+      "loss": 0.2815,
+      "step": 14078
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527177472167649e-05,
+      "loss": 0.3189,
+      "step": 14079
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5271438886370124e-05,
+      "loss": 0.2933,
+      "step": 14080
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527110305106376e-05,
+      "loss": 0.258,
+      "step": 14081
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5270767215757394e-05,
+      "loss": 0.4614,
+      "step": 14082
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.527043138045103e-05,
+      "loss": 0.2959,
+      "step": 14083
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5270095545144663e-05,
+      "loss": 0.2235,
+      "step": 14084
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5269759709838298e-05,
+      "loss": 0.3702,
+      "step": 14085
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.526942387453193e-05,
+      "loss": 0.4112,
+      "step": 14086
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5269088039225567e-05,
+      "loss": 0.2536,
+      "step": 14087
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5268752203919198e-05,
+      "loss": 0.4721,
+      "step": 14088
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5268416368612833e-05,
+      "loss": 0.3203,
+      "step": 14089
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5268080533306468e-05,
+      "loss": 0.3297,
+      "step": 14090
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5267744698000102e-05,
+      "loss": 0.5581,
+      "step": 14091
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5267408862693737e-05,
+      "loss": 0.3332,
+      "step": 14092
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.526707302738737e-05,
+      "loss": 0.1977,
+      "step": 14093
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5266737192081006e-05,
+      "loss": 0.3523,
+      "step": 14094
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5266401356774637e-05,
+      "loss": 0.3774,
+      "step": 14095
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5266065521468275e-05,
+      "loss": 0.3469,
+      "step": 14096
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5265729686161907e-05,
+      "loss": 0.2348,
+      "step": 14097
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.526539385085554e-05,
+      "loss": 0.4736,
+      "step": 14098
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5265058015549176e-05,
+      "loss": 0.2269,
+      "step": 14099
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.526472218024281e-05,
+      "loss": 0.3306,
+      "step": 14100
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5264386344936445e-05,
+      "loss": 0.4412,
+      "step": 14101
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.526405050963008e-05,
+      "loss": 0.4234,
+      "step": 14102
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5263714674323715e-05,
+      "loss": 0.3867,
+      "step": 14103
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5263378839017346e-05,
+      "loss": 0.2966,
+      "step": 14104
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5263043003710984e-05,
+      "loss": 0.5196,
+      "step": 14105
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5262707168404615e-05,
+      "loss": 0.217,
+      "step": 14106
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.526237133309825e-05,
+      "loss": 0.2153,
+      "step": 14107
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5262035497791884e-05,
+      "loss": 0.3267,
+      "step": 14108
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.526169966248552e-05,
+      "loss": 0.3636,
+      "step": 14109
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.526136382717915e-05,
+      "loss": 0.4291,
+      "step": 14110
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.526102799187279e-05,
+      "loss": 0.1842,
+      "step": 14111
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.526069215656642e-05,
+      "loss": 0.3755,
+      "step": 14112
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5260356321260054e-05,
+      "loss": 0.278,
+      "step": 14113
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.526002048595369e-05,
+      "loss": 0.3525,
+      "step": 14114
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5259684650647324e-05,
+      "loss": 0.4159,
+      "step": 14115
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5259348815340958e-05,
+      "loss": 0.2938,
+      "step": 14116
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5259012980034593e-05,
+      "loss": 0.1602,
+      "step": 14117
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5258677144728227e-05,
+      "loss": 0.2484,
+      "step": 14118
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.525834130942186e-05,
+      "loss": 0.2473,
+      "step": 14119
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5258005474115497e-05,
+      "loss": 0.4115,
+      "step": 14120
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.525766963880913e-05,
+      "loss": 0.3287,
+      "step": 14121
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5257333803502763e-05,
+      "loss": 0.3197,
+      "step": 14122
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5256997968196399e-05,
+      "loss": 0.239,
+      "step": 14123
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5256662132890032e-05,
+      "loss": 0.2358,
+      "step": 14124
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5256326297583665e-05,
+      "loss": 0.2717,
+      "step": 14125
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5255990462277301e-05,
+      "loss": 0.3069,
+      "step": 14126
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5255654626970934e-05,
+      "loss": 0.2772,
+      "step": 14127
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5255318791664569e-05,
+      "loss": 0.3628,
+      "step": 14128
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5254982956358204e-05,
+      "loss": 0.4544,
+      "step": 14129
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5254647121051838e-05,
+      "loss": 0.2385,
+      "step": 14130
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5254311285745471e-05,
+      "loss": 0.2702,
+      "step": 14131
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5253975450439107e-05,
+      "loss": 0.2552,
+      "step": 14132
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.525363961513274e-05,
+      "loss": 0.4258,
+      "step": 14133
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5253303779826373e-05,
+      "loss": 0.2329,
+      "step": 14134
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.525296794452001e-05,
+      "loss": 0.1973,
+      "step": 14135
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5252632109213643e-05,
+      "loss": 0.4232,
+      "step": 14136
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5252296273907277e-05,
+      "loss": 0.2053,
+      "step": 14137
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5251960438600912e-05,
+      "loss": 0.2743,
+      "step": 14138
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5251624603294547e-05,
+      "loss": 0.5245,
+      "step": 14139
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.525128876798818e-05,
+      "loss": 0.4509,
+      "step": 14140
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5250952932681816e-05,
+      "loss": 0.284,
+      "step": 14141
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5250617097375449e-05,
+      "loss": 0.5057,
+      "step": 14142
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5250281262069082e-05,
+      "loss": 0.4667,
+      "step": 14143
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5249945426762718e-05,
+      "loss": 0.2702,
+      "step": 14144
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5249609591456351e-05,
+      "loss": 0.2783,
+      "step": 14145
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5249273756149984e-05,
+      "loss": 0.3319,
+      "step": 14146
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.524893792084362e-05,
+      "loss": 0.3231,
+      "step": 14147
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5248602085537253e-05,
+      "loss": 0.4479,
+      "step": 14148
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5248266250230888e-05,
+      "loss": 0.2188,
+      "step": 14149
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5247930414924523e-05,
+      "loss": 0.2907,
+      "step": 14150
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5247594579618157e-05,
+      "loss": 0.5141,
+      "step": 14151
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.524725874431179e-05,
+      "loss": 0.3443,
+      "step": 14152
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5246922909005427e-05,
+      "loss": 0.4129,
+      "step": 14153
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.524658707369906e-05,
+      "loss": 0.3932,
+      "step": 14154
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5246251238392692e-05,
+      "loss": 0.347,
+      "step": 14155
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5245915403086329e-05,
+      "loss": 0.2916,
+      "step": 14156
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5245579567779962e-05,
+      "loss": 0.316,
+      "step": 14157
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5245243732473596e-05,
+      "loss": 0.2557,
+      "step": 14158
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5244907897167231e-05,
+      "loss": 0.2777,
+      "step": 14159
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5244572061860866e-05,
+      "loss": 0.4048,
+      "step": 14160
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5244236226554499e-05,
+      "loss": 0.4012,
+      "step": 14161
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5243900391248135e-05,
+      "loss": 0.3377,
+      "step": 14162
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5243564555941768e-05,
+      "loss": 0.3628,
+      "step": 14163
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5243228720635401e-05,
+      "loss": 0.3721,
+      "step": 14164
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5242892885329037e-05,
+      "loss": 0.3152,
+      "step": 14165
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.524255705002267e-05,
+      "loss": 0.1129,
+      "step": 14166
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5242221214716303e-05,
+      "loss": 0.3182,
+      "step": 14167
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5241885379409938e-05,
+      "loss": 0.4522,
+      "step": 14168
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5241549544103572e-05,
+      "loss": 0.1349,
+      "step": 14169
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5241213708797207e-05,
+      "loss": 0.2949,
+      "step": 14170
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.524087787349084e-05,
+      "loss": 0.2779,
+      "step": 14171
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5240542038184476e-05,
+      "loss": 0.223,
+      "step": 14172
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.524020620287811e-05,
+      "loss": 0.325,
+      "step": 14173
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5239870367571742e-05,
+      "loss": 0.2558,
+      "step": 14174
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5239534532265379e-05,
+      "loss": 0.4146,
+      "step": 14175
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5239198696959012e-05,
+      "loss": 0.3698,
+      "step": 14176
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5238862861652646e-05,
+      "loss": 0.448,
+      "step": 14177
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5238527026346281e-05,
+      "loss": 0.2765,
+      "step": 14178
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5238191191039916e-05,
+      "loss": 0.3258,
+      "step": 14179
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5237855355733549e-05,
+      "loss": 0.2322,
+      "step": 14180
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5237519520427185e-05,
+      "loss": 0.305,
+      "step": 14181
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5237183685120818e-05,
+      "loss": 0.2921,
+      "step": 14182
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.523684784981445e-05,
+      "loss": 0.3826,
+      "step": 14183
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5236512014508087e-05,
+      "loss": 0.3101,
+      "step": 14184
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.523617617920172e-05,
+      "loss": 0.2747,
+      "step": 14185
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5235840343895355e-05,
+      "loss": 0.4647,
+      "step": 14186
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.523550450858899e-05,
+      "loss": 0.4238,
+      "step": 14187
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5235168673282624e-05,
+      "loss": 0.4269,
+      "step": 14188
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5234832837976257e-05,
+      "loss": 0.2731,
+      "step": 14189
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5234497002669892e-05,
+      "loss": 0.3661,
+      "step": 14190
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5234161167363526e-05,
+      "loss": 0.3749,
+      "step": 14191
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.523382533205716e-05,
+      "loss": 0.2667,
+      "step": 14192
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.5233489496750796e-05,
+      "loss": 0.3729,
+      "step": 14193
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5233153661444428e-05,
+      "loss": 0.24,
+      "step": 14194
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5232817826138061e-05,
+      "loss": 0.1531,
+      "step": 14195
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5232481990831698e-05,
+      "loss": 0.4825,
+      "step": 14196
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.523214615552533e-05,
+      "loss": 0.2684,
+      "step": 14197
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5231810320218965e-05,
+      "loss": 0.2442,
+      "step": 14198
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.52314744849126e-05,
+      "loss": 0.3094,
+      "step": 14199
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5231138649606235e-05,
+      "loss": 0.2756,
+      "step": 14200
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5230802814299868e-05,
+      "loss": 0.36,
+      "step": 14201
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5230466978993504e-05,
+      "loss": 0.2739,
+      "step": 14202
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5230131143687137e-05,
+      "loss": 0.2193,
+      "step": 14203
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.522979530838077e-05,
+      "loss": 0.3796,
+      "step": 14204
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5229459473074406e-05,
+      "loss": 0.4015,
+      "step": 14205
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.522912363776804e-05,
+      "loss": 0.4574,
+      "step": 14206
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5228787802461674e-05,
+      "loss": 0.4407,
+      "step": 14207
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5228451967155308e-05,
+      "loss": 0.4813,
+      "step": 14208
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5228116131848943e-05,
+      "loss": 0.1933,
+      "step": 14209
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5227780296542576e-05,
+      "loss": 0.3359,
+      "step": 14210
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.522744446123621e-05,
+      "loss": 0.2562,
+      "step": 14211
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5227108625929845e-05,
+      "loss": 0.391,
+      "step": 14212
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5226772790623478e-05,
+      "loss": 0.2977,
+      "step": 14213
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5226436955317115e-05,
+      "loss": 0.2544,
+      "step": 14214
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5226101120010748e-05,
+      "loss": 0.4416,
+      "step": 14215
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.522576528470438e-05,
+      "loss": 0.5483,
+      "step": 14216
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5225429449398017e-05,
+      "loss": 0.3152,
+      "step": 14217
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.522509361409165e-05,
+      "loss": 0.2864,
+      "step": 14218
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5224757778785285e-05,
+      "loss": 0.2576,
+      "step": 14219
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.522442194347892e-05,
+      "loss": 0.2175,
+      "step": 14220
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5224086108172554e-05,
+      "loss": 0.2311,
+      "step": 14221
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5223750272866187e-05,
+      "loss": 0.4446,
+      "step": 14222
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5223414437559823e-05,
+      "loss": 0.4213,
+      "step": 14223
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5223078602253456e-05,
+      "loss": 0.3171,
+      "step": 14224
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5222742766947089e-05,
+      "loss": 0.3661,
+      "step": 14225
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5222406931640725e-05,
+      "loss": 0.4346,
+      "step": 14226
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5222071096334358e-05,
+      "loss": 0.421,
+      "step": 14227
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5221735261027993e-05,
+      "loss": 0.2005,
+      "step": 14228
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5221399425721628e-05,
+      "loss": 0.3288,
+      "step": 14229
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5221063590415262e-05,
+      "loss": 0.3847,
+      "step": 14230
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5220727755108895e-05,
+      "loss": 0.21,
+      "step": 14231
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5220391919802532e-05,
+      "loss": 0.3377,
+      "step": 14232
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5220056084496164e-05,
+      "loss": 0.2296,
+      "step": 14233
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5219720249189797e-05,
+      "loss": 0.2221,
+      "step": 14234
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5219384413883434e-05,
+      "loss": 0.3173,
+      "step": 14235
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5219048578577067e-05,
+      "loss": 0.282,
+      "step": 14236
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.52187127432707e-05,
+      "loss": 0.3109,
+      "step": 14237
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5218376907964336e-05,
+      "loss": 0.3174,
+      "step": 14238
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5218041072657969e-05,
+      "loss": 0.2937,
+      "step": 14239
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5217705237351604e-05,
+      "loss": 0.2744,
+      "step": 14240
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5217369402045238e-05,
+      "loss": 0.3997,
+      "step": 14241
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5217033566738873e-05,
+      "loss": 0.2123,
+      "step": 14242
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5216697731432506e-05,
+      "loss": 0.4141,
+      "step": 14243
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5216361896126142e-05,
+      "loss": 0.2608,
+      "step": 14244
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5216026060819775e-05,
+      "loss": 0.257,
+      "step": 14245
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5215690225513408e-05,
+      "loss": 0.2759,
+      "step": 14246
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5215354390207044e-05,
+      "loss": 0.3514,
+      "step": 14247
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5215018554900677e-05,
+      "loss": 0.4359,
+      "step": 14248
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5214682719594312e-05,
+      "loss": 0.2734,
+      "step": 14249
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5214346884287947e-05,
+      "loss": 0.4378,
+      "step": 14250
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5214011048981581e-05,
+      "loss": 0.3535,
+      "step": 14251
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5213675213675214e-05,
+      "loss": 0.3309,
+      "step": 14252
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.521333937836885e-05,
+      "loss": 0.3562,
+      "step": 14253
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5213003543062484e-05,
+      "loss": 0.27,
+      "step": 14254
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5212667707756117e-05,
+      "loss": 0.1798,
+      "step": 14255
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5212331872449753e-05,
+      "loss": 0.292,
+      "step": 14256
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5211996037143386e-05,
+      "loss": 0.2485,
+      "step": 14257
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5211660201837019e-05,
+      "loss": 0.3376,
+      "step": 14258
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5211324366530655e-05,
+      "loss": 0.1555,
+      "step": 14259
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5210988531224288e-05,
+      "loss": 0.4784,
+      "step": 14260
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5210652695917923e-05,
+      "loss": 0.3664,
+      "step": 14261
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5210316860611557e-05,
+      "loss": 0.2581,
+      "step": 14262
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5209981025305192e-05,
+      "loss": 0.2234,
+      "step": 14263
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5209645189998825e-05,
+      "loss": 0.1914,
+      "step": 14264
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5209309354692461e-05,
+      "loss": 0.3273,
+      "step": 14265
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5208973519386094e-05,
+      "loss": 0.3274,
+      "step": 14266
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5208637684079727e-05,
+      "loss": 0.3857,
+      "step": 14267
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5208301848773364e-05,
+      "loss": 0.196,
+      "step": 14268
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5207966013466997e-05,
+      "loss": 0.2287,
+      "step": 14269
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5207630178160631e-05,
+      "loss": 0.3045,
+      "step": 14270
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5207294342854266e-05,
+      "loss": 0.379,
+      "step": 14271
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.52069585075479e-05,
+      "loss": 0.2695,
+      "step": 14272
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5206622672241533e-05,
+      "loss": 0.2657,
+      "step": 14273
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.520628683693517e-05,
+      "loss": 0.2333,
+      "step": 14274
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5205951001628803e-05,
+      "loss": 0.2202,
+      "step": 14275
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5205615166322436e-05,
+      "loss": 0.3243,
+      "step": 14276
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5205279331016072e-05,
+      "loss": 0.2774,
+      "step": 14277
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5204943495709705e-05,
+      "loss": 0.2919,
+      "step": 14278
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.520460766040334e-05,
+      "loss": 0.3637,
+      "step": 14279
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5204271825096974e-05,
+      "loss": 0.3504,
+      "step": 14280
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5203935989790607e-05,
+      "loss": 0.3658,
+      "step": 14281
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5203600154484242e-05,
+      "loss": 0.3541,
+      "step": 14282
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5203264319177877e-05,
+      "loss": 0.2882,
+      "step": 14283
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5202928483871511e-05,
+      "loss": 0.2631,
+      "step": 14284
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5202592648565144e-05,
+      "loss": 0.354,
+      "step": 14285
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.520225681325878e-05,
+      "loss": 0.2882,
+      "step": 14286
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5201920977952413e-05,
+      "loss": 0.3713,
+      "step": 14287
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5201585142646046e-05,
+      "loss": 0.3858,
+      "step": 14288
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5201249307339683e-05,
+      "loss": 0.418,
+      "step": 14289
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5200913472033316e-05,
+      "loss": 0.306,
+      "step": 14290
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.520057763672695e-05,
+      "loss": 0.2321,
+      "step": 14291
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5200241801420585e-05,
+      "loss": 0.4544,
+      "step": 14292
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.519990596611422e-05,
+      "loss": 0.3576,
+      "step": 14293
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5199570130807853e-05,
+      "loss": 0.3824,
+      "step": 14294
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5199234295501489e-05,
+      "loss": 0.1803,
+      "step": 14295
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5198898460195122e-05,
+      "loss": 0.4539,
+      "step": 14296
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5198562624888755e-05,
+      "loss": 0.2164,
+      "step": 14297
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5198226789582391e-05,
+      "loss": 0.3426,
+      "step": 14298
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5197890954276024e-05,
+      "loss": 0.2586,
+      "step": 14299
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5197555118969659e-05,
+      "loss": 0.4405,
+      "step": 14300
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5197219283663293e-05,
+      "loss": 0.364,
+      "step": 14301
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5196883448356926e-05,
+      "loss": 0.213,
+      "step": 14302
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5196547613050561e-05,
+      "loss": 0.3116,
+      "step": 14303
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5196211777744196e-05,
+      "loss": 0.2814,
+      "step": 14304
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.519587594243783e-05,
+      "loss": 0.4728,
+      "step": 14305
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5195540107131463e-05,
+      "loss": 0.3174,
+      "step": 14306
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.51952042718251e-05,
+      "loss": 0.2878,
+      "step": 14307
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5194868436518733e-05,
+      "loss": 0.3087,
+      "step": 14308
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5194532601212366e-05,
+      "loss": 0.1882,
+      "step": 14309
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5194196765906002e-05,
+      "loss": 0.286,
+      "step": 14310
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5193860930599635e-05,
+      "loss": 0.3707,
+      "step": 14311
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.519352509529327e-05,
+      "loss": 0.3479,
+      "step": 14312
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5193189259986904e-05,
+      "loss": 0.3266,
+      "step": 14313
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5192853424680539e-05,
+      "loss": 0.2325,
+      "step": 14314
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5192517589374172e-05,
+      "loss": 0.4128,
+      "step": 14315
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5192181754067808e-05,
+      "loss": 0.2863,
+      "step": 14316
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5191845918761441e-05,
+      "loss": 0.3298,
+      "step": 14317
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5191510083455074e-05,
+      "loss": 0.3678,
+      "step": 14318
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.519117424814871e-05,
+      "loss": 0.2351,
+      "step": 14319
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5190838412842343e-05,
+      "loss": 0.2851,
+      "step": 14320
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5190502577535978e-05,
+      "loss": 0.2036,
+      "step": 14321
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5190166742229613e-05,
+      "loss": 0.4146,
+      "step": 14322
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5189830906923247e-05,
+      "loss": 0.3675,
+      "step": 14323
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.518949507161688e-05,
+      "loss": 0.3202,
+      "step": 14324
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5189159236310515e-05,
+      "loss": 0.2349,
+      "step": 14325
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.518882340100415e-05,
+      "loss": 0.3494,
+      "step": 14326
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5188487565697782e-05,
+      "loss": 0.4039,
+      "step": 14327
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5188151730391419e-05,
+      "loss": 0.6521,
+      "step": 14328
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5187815895085052e-05,
+      "loss": 0.279,
+      "step": 14329
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5187480059778685e-05,
+      "loss": 0.2303,
+      "step": 14330
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5187144224472321e-05,
+      "loss": 0.388,
+      "step": 14331
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5186808389165954e-05,
+      "loss": 0.2152,
+      "step": 14332
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5186472553859589e-05,
+      "loss": 0.4244,
+      "step": 14333
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5186136718553223e-05,
+      "loss": 0.3452,
+      "step": 14334
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5185800883246858e-05,
+      "loss": 0.1871,
+      "step": 14335
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.518546504794049e-05,
+      "loss": 0.3708,
+      "step": 14336
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5185129212634127e-05,
+      "loss": 0.3042,
+      "step": 14337
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.518479337732776e-05,
+      "loss": 0.5325,
+      "step": 14338
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5184457542021393e-05,
+      "loss": 0.3066,
+      "step": 14339
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.518412170671503e-05,
+      "loss": 0.2193,
+      "step": 14340
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5183785871408662e-05,
+      "loss": 0.1787,
+      "step": 14341
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5183450036102297e-05,
+      "loss": 0.1611,
+      "step": 14342
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5183114200795932e-05,
+      "loss": 0.3628,
+      "step": 14343
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5182778365489566e-05,
+      "loss": 0.4492,
+      "step": 14344
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.51824425301832e-05,
+      "loss": 0.3349,
+      "step": 14345
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5182106694876834e-05,
+      "loss": 0.1976,
+      "step": 14346
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5181770859570469e-05,
+      "loss": 0.4223,
+      "step": 14347
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5181435024264102e-05,
+      "loss": 0.284,
+      "step": 14348
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5181099188957734e-05,
+      "loss": 0.3236,
+      "step": 14349
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.518076335365137e-05,
+      "loss": 0.3908,
+      "step": 14350
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5180427518345004e-05,
+      "loss": 0.4207,
+      "step": 14351
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5180091683038638e-05,
+      "loss": 0.2739,
+      "step": 14352
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5179755847732273e-05,
+      "loss": 0.4938,
+      "step": 14353
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5179420012425908e-05,
+      "loss": 0.3082,
+      "step": 14354
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.517908417711954e-05,
+      "loss": 0.3519,
+      "step": 14355
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5178748341813177e-05,
+      "loss": 0.2815,
+      "step": 14356
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.517841250650681e-05,
+      "loss": 0.2736,
+      "step": 14357
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5178076671200443e-05,
+      "loss": 0.2799,
+      "step": 14358
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.517774083589408e-05,
+      "loss": 0.2928,
+      "step": 14359
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5177405000587712e-05,
+      "loss": 0.2735,
+      "step": 14360
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5177069165281347e-05,
+      "loss": 0.4087,
+      "step": 14361
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5176733329974981e-05,
+      "loss": 0.4545,
+      "step": 14362
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5176397494668616e-05,
+      "loss": 0.2623,
+      "step": 14363
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5176061659362249e-05,
+      "loss": 0.3892,
+      "step": 14364
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5175725824055885e-05,
+      "loss": 0.2839,
+      "step": 14365
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5175389988749518e-05,
+      "loss": 0.2747,
+      "step": 14366
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5175054153443151e-05,
+      "loss": 0.3086,
+      "step": 14367
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5174718318136788e-05,
+      "loss": 0.4677,
+      "step": 14368
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.517438248283042e-05,
+      "loss": 0.3514,
+      "step": 14369
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5174046647524054e-05,
+      "loss": 0.2922,
+      "step": 14370
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.517371081221769e-05,
+      "loss": 0.2332,
+      "step": 14371
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5173374976911323e-05,
+      "loss": 0.5791,
+      "step": 14372
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5173039141604958e-05,
+      "loss": 0.3508,
+      "step": 14373
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5172703306298592e-05,
+      "loss": 0.1921,
+      "step": 14374
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5172367470992227e-05,
+      "loss": 0.3524,
+      "step": 14375
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.517203163568586e-05,
+      "loss": 0.1458,
+      "step": 14376
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5171695800379496e-05,
+      "loss": 0.3902,
+      "step": 14377
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5171359965073129e-05,
+      "loss": 0.4767,
+      "step": 14378
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5171024129766762e-05,
+      "loss": 0.3431,
+      "step": 14379
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5170688294460398e-05,
+      "loss": 0.2601,
+      "step": 14380
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5170352459154031e-05,
+      "loss": 0.3668,
+      "step": 14381
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5170016623847666e-05,
+      "loss": 0.2895,
+      "step": 14382
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.51696807885413e-05,
+      "loss": 0.2718,
+      "step": 14383
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5169344953234935e-05,
+      "loss": 0.2991,
+      "step": 14384
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5169009117928568e-05,
+      "loss": 0.3611,
+      "step": 14385
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5168673282622205e-05,
+      "loss": 0.319,
+      "step": 14386
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5168337447315838e-05,
+      "loss": 0.2905,
+      "step": 14387
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.516800161200947e-05,
+      "loss": 0.3224,
+      "step": 14388
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5167665776703107e-05,
+      "loss": 0.4727,
+      "step": 14389
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.516732994139674e-05,
+      "loss": 0.2446,
+      "step": 14390
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.5166994106090374e-05,
+      "loss": 0.2399,
+      "step": 14391
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5166658270784009e-05,
+      "loss": 0.282,
+      "step": 14392
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5166322435477642e-05,
+      "loss": 0.2187,
+      "step": 14393
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5165986600171277e-05,
+      "loss": 0.2863,
+      "step": 14394
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5165650764864911e-05,
+      "loss": 0.4254,
+      "step": 14395
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5165314929558546e-05,
+      "loss": 0.3087,
+      "step": 14396
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5164979094252179e-05,
+      "loss": 0.3545,
+      "step": 14397
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5164643258945815e-05,
+      "loss": 0.3611,
+      "step": 14398
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5164307423639448e-05,
+      "loss": 0.2384,
+      "step": 14399
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5163971588333081e-05,
+      "loss": 0.3248,
+      "step": 14400
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5163635753026717e-05,
+      "loss": 0.4503,
+      "step": 14401
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.516329991772035e-05,
+      "loss": 0.4623,
+      "step": 14402
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5162964082413985e-05,
+      "loss": 0.3428,
+      "step": 14403
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.516262824710762e-05,
+      "loss": 0.3946,
+      "step": 14404
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5162292411801254e-05,
+      "loss": 0.2913,
+      "step": 14405
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5161956576494887e-05,
+      "loss": 0.4083,
+      "step": 14406
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5161620741188524e-05,
+      "loss": 0.351,
+      "step": 14407
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5161284905882157e-05,
+      "loss": 0.4208,
+      "step": 14408
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.516094907057579e-05,
+      "loss": 0.3689,
+      "step": 14409
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5160613235269426e-05,
+      "loss": 0.2309,
+      "step": 14410
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5160277399963059e-05,
+      "loss": 0.2316,
+      "step": 14411
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5159941564656694e-05,
+      "loss": 0.2796,
+      "step": 14412
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5159605729350328e-05,
+      "loss": 0.3879,
+      "step": 14413
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5159269894043963e-05,
+      "loss": 0.359,
+      "step": 14414
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5158934058737596e-05,
+      "loss": 0.3636,
+      "step": 14415
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.515859822343123e-05,
+      "loss": 0.3164,
+      "step": 14416
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5158262388124865e-05,
+      "loss": 0.1579,
+      "step": 14417
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5157926552818498e-05,
+      "loss": 0.2815,
+      "step": 14418
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5157590717512134e-05,
+      "loss": 0.3009,
+      "step": 14419
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5157254882205767e-05,
+      "loss": 0.3843,
+      "step": 14420
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.51569190468994e-05,
+      "loss": 0.2127,
+      "step": 14421
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5156583211593037e-05,
+      "loss": 0.2884,
+      "step": 14422
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.515624737628667e-05,
+      "loss": 0.2761,
+      "step": 14423
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5155911540980304e-05,
+      "loss": 0.3197,
+      "step": 14424
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5155575705673939e-05,
+      "loss": 0.3133,
+      "step": 14425
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5155239870367574e-05,
+      "loss": 0.2791,
+      "step": 14426
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5154904035061206e-05,
+      "loss": 0.3009,
+      "step": 14427
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5154568199754843e-05,
+      "loss": 0.316,
+      "step": 14428
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5154232364448476e-05,
+      "loss": 0.4224,
+      "step": 14429
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5153896529142109e-05,
+      "loss": 0.2677,
+      "step": 14430
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5153560693835745e-05,
+      "loss": 0.2355,
+      "step": 14431
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5153224858529378e-05,
+      "loss": 0.2056,
+      "step": 14432
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5152889023223013e-05,
+      "loss": 0.2689,
+      "step": 14433
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5152553187916647e-05,
+      "loss": 0.3053,
+      "step": 14434
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5152217352610282e-05,
+      "loss": 0.2045,
+      "step": 14435
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5151881517303915e-05,
+      "loss": 0.2567,
+      "step": 14436
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.515154568199755e-05,
+      "loss": 0.3662,
+      "step": 14437
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5151209846691184e-05,
+      "loss": 0.2767,
+      "step": 14438
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5150874011384817e-05,
+      "loss": 0.2326,
+      "step": 14439
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5150538176078453e-05,
+      "loss": 0.4249,
+      "step": 14440
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5150202340772086e-05,
+      "loss": 0.3398,
+      "step": 14441
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.514986650546572e-05,
+      "loss": 0.2613,
+      "step": 14442
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5149530670159356e-05,
+      "loss": 0.3657,
+      "step": 14443
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5149194834852989e-05,
+      "loss": 0.2589,
+      "step": 14444
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5148858999546623e-05,
+      "loss": 0.2142,
+      "step": 14445
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5148523164240258e-05,
+      "loss": 0.2743,
+      "step": 14446
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5148187328933893e-05,
+      "loss": 0.2013,
+      "step": 14447
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5147851493627526e-05,
+      "loss": 0.3799,
+      "step": 14448
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5147515658321162e-05,
+      "loss": 0.375,
+      "step": 14449
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5147179823014795e-05,
+      "loss": 0.3039,
+      "step": 14450
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5146843987708428e-05,
+      "loss": 0.2401,
+      "step": 14451
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5146508152402064e-05,
+      "loss": 0.3428,
+      "step": 14452
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5146172317095697e-05,
+      "loss": 0.2849,
+      "step": 14453
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5145836481789332e-05,
+      "loss": 0.2871,
+      "step": 14454
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5145500646482966e-05,
+      "loss": 0.2849,
+      "step": 14455
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5145164811176601e-05,
+      "loss": 0.3539,
+      "step": 14456
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5144828975870234e-05,
+      "loss": 0.2612,
+      "step": 14457
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.514449314056387e-05,
+      "loss": 0.3758,
+      "step": 14458
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5144157305257503e-05,
+      "loss": 0.442,
+      "step": 14459
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5143821469951136e-05,
+      "loss": 0.3048,
+      "step": 14460
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5143485634644773e-05,
+      "loss": 0.1385,
+      "step": 14461
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5143149799338406e-05,
+      "loss": 0.1557,
+      "step": 14462
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5142813964032039e-05,
+      "loss": 0.2151,
+      "step": 14463
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5142478128725675e-05,
+      "loss": 0.2979,
+      "step": 14464
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5142142293419308e-05,
+      "loss": 0.2191,
+      "step": 14465
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5141806458112942e-05,
+      "loss": 0.3278,
+      "step": 14466
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5141470622806577e-05,
+      "loss": 0.197,
+      "step": 14467
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5141134787500212e-05,
+      "loss": 0.4029,
+      "step": 14468
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5140798952193845e-05,
+      "loss": 0.2225,
+      "step": 14469
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5140463116887481e-05,
+      "loss": 0.3839,
+      "step": 14470
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5140127281581114e-05,
+      "loss": 0.3119,
+      "step": 14471
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5139791446274747e-05,
+      "loss": 0.268,
+      "step": 14472
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5139455610968383e-05,
+      "loss": 0.1794,
+      "step": 14473
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5139119775662016e-05,
+      "loss": 0.2675,
+      "step": 14474
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5138783940355651e-05,
+      "loss": 0.4038,
+      "step": 14475
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5138448105049286e-05,
+      "loss": 0.2862,
+      "step": 14476
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.513811226974292e-05,
+      "loss": 0.238,
+      "step": 14477
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5137776434436553e-05,
+      "loss": 0.2115,
+      "step": 14478
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.513744059913019e-05,
+      "loss": 0.4185,
+      "step": 14479
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5137104763823822e-05,
+      "loss": 0.4048,
+      "step": 14480
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5136768928517455e-05,
+      "loss": 0.2258,
+      "step": 14481
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5136433093211092e-05,
+      "loss": 0.2907,
+      "step": 14482
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5136097257904725e-05,
+      "loss": 0.2288,
+      "step": 14483
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5135761422598358e-05,
+      "loss": 0.3694,
+      "step": 14484
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5135425587291994e-05,
+      "loss": 0.2557,
+      "step": 14485
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5135089751985627e-05,
+      "loss": 0.3598,
+      "step": 14486
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5134753916679262e-05,
+      "loss": 0.3091,
+      "step": 14487
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5134418081372896e-05,
+      "loss": 0.4016,
+      "step": 14488
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5134082246066531e-05,
+      "loss": 0.4457,
+      "step": 14489
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5133746410760164e-05,
+      "loss": 0.312,
+      "step": 14490
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.51334105754538e-05,
+      "loss": 0.3936,
+      "step": 14491
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5133074740147433e-05,
+      "loss": 0.3808,
+      "step": 14492
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5132738904841066e-05,
+      "loss": 0.3201,
+      "step": 14493
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5132403069534702e-05,
+      "loss": 0.3098,
+      "step": 14494
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5132067234228335e-05,
+      "loss": 0.3194,
+      "step": 14495
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.513173139892197e-05,
+      "loss": 0.417,
+      "step": 14496
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5131395563615605e-05,
+      "loss": 0.4843,
+      "step": 14497
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.513105972830924e-05,
+      "loss": 0.3274,
+      "step": 14498
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5130723893002872e-05,
+      "loss": 0.32,
+      "step": 14499
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5130388057696509e-05,
+      "loss": 0.3256,
+      "step": 14500
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5130052222390142e-05,
+      "loss": 0.3694,
+      "step": 14501
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5129716387083775e-05,
+      "loss": 0.3813,
+      "step": 14502
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5129380551777411e-05,
+      "loss": 0.3427,
+      "step": 14503
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5129044716471044e-05,
+      "loss": 0.3127,
+      "step": 14504
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5128708881164677e-05,
+      "loss": 0.2781,
+      "step": 14505
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5128373045858313e-05,
+      "loss": 0.3009,
+      "step": 14506
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5128037210551946e-05,
+      "loss": 0.1735,
+      "step": 14507
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.512770137524558e-05,
+      "loss": 0.2278,
+      "step": 14508
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5127365539939215e-05,
+      "loss": 0.3616,
+      "step": 14509
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.512702970463285e-05,
+      "loss": 0.3177,
+      "step": 14510
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5126693869326483e-05,
+      "loss": 0.3547,
+      "step": 14511
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.512635803402012e-05,
+      "loss": 0.3238,
+      "step": 14512
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5126022198713752e-05,
+      "loss": 0.3397,
+      "step": 14513
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5125686363407385e-05,
+      "loss": 0.3076,
+      "step": 14514
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5125350528101022e-05,
+      "loss": 0.3318,
+      "step": 14515
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5125014692794655e-05,
+      "loss": 0.4141,
+      "step": 14516
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.512467885748829e-05,
+      "loss": 0.2881,
+      "step": 14517
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5124343022181924e-05,
+      "loss": 0.3982,
+      "step": 14518
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5124007186875558e-05,
+      "loss": 0.38,
+      "step": 14519
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5123671351569191e-05,
+      "loss": 0.319,
+      "step": 14520
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5123335516262828e-05,
+      "loss": 0.3541,
+      "step": 14521
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.512299968095646e-05,
+      "loss": 0.5594,
+      "step": 14522
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5122663845650094e-05,
+      "loss": 0.3435,
+      "step": 14523
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.512232801034373e-05,
+      "loss": 0.5562,
+      "step": 14524
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5121992175037363e-05,
+      "loss": 0.57,
+      "step": 14525
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5121656339730998e-05,
+      "loss": 0.3676,
+      "step": 14526
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5121320504424632e-05,
+      "loss": 0.3889,
+      "step": 14527
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5120984669118265e-05,
+      "loss": 0.2937,
+      "step": 14528
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.51206488338119e-05,
+      "loss": 0.3322,
+      "step": 14529
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5120312998505533e-05,
+      "loss": 0.4818,
+      "step": 14530
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5119977163199169e-05,
+      "loss": 0.4117,
+      "step": 14531
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5119641327892802e-05,
+      "loss": 0.2282,
+      "step": 14532
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5119305492586435e-05,
+      "loss": 0.4111,
+      "step": 14533
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5118969657280071e-05,
+      "loss": 0.3171,
+      "step": 14534
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5118633821973704e-05,
+      "loss": 0.215,
+      "step": 14535
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5118297986667339e-05,
+      "loss": 0.2238,
+      "step": 14536
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5117962151360974e-05,
+      "loss": 0.3939,
+      "step": 14537
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5117626316054608e-05,
+      "loss": 0.2407,
+      "step": 14538
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5117290480748241e-05,
+      "loss": 0.215,
+      "step": 14539
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5116954645441878e-05,
+      "loss": 0.2993,
+      "step": 14540
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.511661881013551e-05,
+      "loss": 0.2226,
+      "step": 14541
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5116282974829143e-05,
+      "loss": 0.1553,
+      "step": 14542
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.511594713952278e-05,
+      "loss": 0.4087,
+      "step": 14543
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5115611304216413e-05,
+      "loss": 0.2857,
+      "step": 14544
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5115275468910047e-05,
+      "loss": 0.1862,
+      "step": 14545
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5114939633603682e-05,
+      "loss": 0.3382,
+      "step": 14546
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5114603798297317e-05,
+      "loss": 0.3544,
+      "step": 14547
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.511426796299095e-05,
+      "loss": 0.5488,
+      "step": 14548
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5113932127684586e-05,
+      "loss": 0.2083,
+      "step": 14549
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5113596292378219e-05,
+      "loss": 0.5202,
+      "step": 14550
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5113260457071852e-05,
+      "loss": 0.2265,
+      "step": 14551
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5112924621765488e-05,
+      "loss": 0.3284,
+      "step": 14552
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5112588786459121e-05,
+      "loss": 0.3811,
+      "step": 14553
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5112252951152754e-05,
+      "loss": 0.3671,
+      "step": 14554
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.511191711584639e-05,
+      "loss": 0.2409,
+      "step": 14555
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5111581280540023e-05,
+      "loss": 0.4161,
+      "step": 14556
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5111245445233658e-05,
+      "loss": 0.2798,
+      "step": 14557
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5110909609927293e-05,
+      "loss": 0.4175,
+      "step": 14558
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5110573774620927e-05,
+      "loss": 0.3735,
+      "step": 14559
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.511023793931456e-05,
+      "loss": 0.3198,
+      "step": 14560
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5109902104008197e-05,
+      "loss": 0.3465,
+      "step": 14561
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.510956626870183e-05,
+      "loss": 0.4416,
+      "step": 14562
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5109230433395463e-05,
+      "loss": 0.2336,
+      "step": 14563
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5108894598089099e-05,
+      "loss": 0.4741,
+      "step": 14564
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5108558762782732e-05,
+      "loss": 0.3728,
+      "step": 14565
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5108222927476367e-05,
+      "loss": 0.2492,
+      "step": 14566
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5107887092170001e-05,
+      "loss": 0.3135,
+      "step": 14567
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5107551256863636e-05,
+      "loss": 0.4667,
+      "step": 14568
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5107215421557269e-05,
+      "loss": 0.2728,
+      "step": 14569
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5106879586250905e-05,
+      "loss": 0.4172,
+      "step": 14570
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5106543750944538e-05,
+      "loss": 0.2946,
+      "step": 14571
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5106207915638171e-05,
+      "loss": 0.3295,
+      "step": 14572
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5105872080331807e-05,
+      "loss": 0.3113,
+      "step": 14573
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.510553624502544e-05,
+      "loss": 0.1813,
+      "step": 14574
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5105200409719073e-05,
+      "loss": 0.3794,
+      "step": 14575
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.510486457441271e-05,
+      "loss": 0.2439,
+      "step": 14576
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5104528739106343e-05,
+      "loss": 0.5021,
+      "step": 14577
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5104192903799977e-05,
+      "loss": 0.2762,
+      "step": 14578
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5103857068493612e-05,
+      "loss": 0.2764,
+      "step": 14579
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5103521233187247e-05,
+      "loss": 0.3229,
+      "step": 14580
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.510318539788088e-05,
+      "loss": 0.3007,
+      "step": 14581
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5102849562574516e-05,
+      "loss": 0.3928,
+      "step": 14582
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5102513727268149e-05,
+      "loss": 0.4825,
+      "step": 14583
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5102177891961782e-05,
+      "loss": 0.1969,
+      "step": 14584
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5101842056655418e-05,
+      "loss": 0.2803,
+      "step": 14585
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5101506221349051e-05,
+      "loss": 0.2709,
+      "step": 14586
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5101170386042686e-05,
+      "loss": 0.3299,
+      "step": 14587
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.510083455073632e-05,
+      "loss": 0.2073,
+      "step": 14588
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5100498715429955e-05,
+      "loss": 0.2647,
+      "step": 14589
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.5100162880123588e-05,
+      "loss": 0.2987,
+      "step": 14590
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5099827044817224e-05,
+      "loss": 0.278,
+      "step": 14591
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5099491209510857e-05,
+      "loss": 0.3819,
+      "step": 14592
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.509915537420449e-05,
+      "loss": 0.3601,
+      "step": 14593
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5098819538898127e-05,
+      "loss": 0.2761,
+      "step": 14594
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.509848370359176e-05,
+      "loss": 0.34,
+      "step": 14595
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5098147868285392e-05,
+      "loss": 0.351,
+      "step": 14596
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5097812032979029e-05,
+      "loss": 0.4299,
+      "step": 14597
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5097476197672662e-05,
+      "loss": 0.1381,
+      "step": 14598
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5097140362366296e-05,
+      "loss": 0.2921,
+      "step": 14599
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5096804527059931e-05,
+      "loss": 0.2468,
+      "step": 14600
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5096468691753566e-05,
+      "loss": 0.3249,
+      "step": 14601
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5096132856447199e-05,
+      "loss": 0.3396,
+      "step": 14602
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5095797021140835e-05,
+      "loss": 0.3759,
+      "step": 14603
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5095461185834468e-05,
+      "loss": 0.2802,
+      "step": 14604
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5095125350528101e-05,
+      "loss": 0.3223,
+      "step": 14605
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5094789515221737e-05,
+      "loss": 0.3017,
+      "step": 14606
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.509445367991537e-05,
+      "loss": 0.2411,
+      "step": 14607
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5094117844609005e-05,
+      "loss": 0.2165,
+      "step": 14608
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.509378200930264e-05,
+      "loss": 0.3425,
+      "step": 14609
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5093446173996274e-05,
+      "loss": 0.4024,
+      "step": 14610
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5093110338689907e-05,
+      "loss": 0.4487,
+      "step": 14611
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5092774503383543e-05,
+      "loss": 0.4691,
+      "step": 14612
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5092438668077176e-05,
+      "loss": 0.5237,
+      "step": 14613
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.509210283277081e-05,
+      "loss": 0.2872,
+      "step": 14614
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5091766997464446e-05,
+      "loss": 0.1165,
+      "step": 14615
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5091431162158079e-05,
+      "loss": 0.2602,
+      "step": 14616
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5091095326851713e-05,
+      "loss": 0.3217,
+      "step": 14617
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5090759491545348e-05,
+      "loss": 0.2814,
+      "step": 14618
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.509042365623898e-05,
+      "loss": 0.333,
+      "step": 14619
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5090087820932615e-05,
+      "loss": 0.3047,
+      "step": 14620
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.508975198562625e-05,
+      "loss": 0.4002,
+      "step": 14621
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5089416150319885e-05,
+      "loss": 0.3714,
+      "step": 14622
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5089080315013518e-05,
+      "loss": 0.3226,
+      "step": 14623
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5088744479707154e-05,
+      "loss": 0.5507,
+      "step": 14624
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5088408644400787e-05,
+      "loss": 0.2858,
+      "step": 14625
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.508807280909442e-05,
+      "loss": 0.3688,
+      "step": 14626
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5087736973788056e-05,
+      "loss": 0.1469,
+      "step": 14627
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.508740113848169e-05,
+      "loss": 0.3131,
+      "step": 14628
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5087065303175324e-05,
+      "loss": 0.3671,
+      "step": 14629
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5086729467868959e-05,
+      "loss": 0.2887,
+      "step": 14630
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5086393632562593e-05,
+      "loss": 0.2031,
+      "step": 14631
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5086057797256226e-05,
+      "loss": 0.2826,
+      "step": 14632
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5085721961949863e-05,
+      "loss": 0.3811,
+      "step": 14633
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5085386126643495e-05,
+      "loss": 0.3019,
+      "step": 14634
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5085050291337128e-05,
+      "loss": 0.3884,
+      "step": 14635
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5084714456030765e-05,
+      "loss": 0.3479,
+      "step": 14636
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5084378620724398e-05,
+      "loss": 0.4275,
+      "step": 14637
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5084042785418032e-05,
+      "loss": 0.2197,
+      "step": 14638
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5083706950111667e-05,
+      "loss": 0.2927,
+      "step": 14639
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.50833711148053e-05,
+      "loss": 0.3829,
+      "step": 14640
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5083035279498935e-05,
+      "loss": 0.1594,
+      "step": 14641
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.508269944419257e-05,
+      "loss": 0.2007,
+      "step": 14642
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5082363608886204e-05,
+      "loss": 0.346,
+      "step": 14643
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5082027773579837e-05,
+      "loss": 0.324,
+      "step": 14644
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5081691938273473e-05,
+      "loss": 0.4417,
+      "step": 14645
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5081356102967106e-05,
+      "loss": 0.4023,
+      "step": 14646
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5081020267660739e-05,
+      "loss": 0.3099,
+      "step": 14647
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5080684432354375e-05,
+      "loss": 0.3422,
+      "step": 14648
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5080348597048008e-05,
+      "loss": 0.3975,
+      "step": 14649
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5080012761741643e-05,
+      "loss": 0.2896,
+      "step": 14650
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5079676926435278e-05,
+      "loss": 0.2105,
+      "step": 14651
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5079341091128912e-05,
+      "loss": 0.2196,
+      "step": 14652
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5079005255822545e-05,
+      "loss": 0.4218,
+      "step": 14653
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5078669420516182e-05,
+      "loss": 0.2932,
+      "step": 14654
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5078333585209815e-05,
+      "loss": 0.267,
+      "step": 14655
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5077997749903448e-05,
+      "loss": 0.2338,
+      "step": 14656
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5077661914597084e-05,
+      "loss": 0.3137,
+      "step": 14657
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5077326079290717e-05,
+      "loss": 0.2105,
+      "step": 14658
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5076990243984351e-05,
+      "loss": 0.3402,
+      "step": 14659
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5076654408677986e-05,
+      "loss": 0.2907,
+      "step": 14660
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.507631857337162e-05,
+      "loss": 0.4499,
+      "step": 14661
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5075982738065254e-05,
+      "loss": 0.3212,
+      "step": 14662
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5075646902758888e-05,
+      "loss": 0.4376,
+      "step": 14663
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5075311067452523e-05,
+      "loss": 0.3369,
+      "step": 14664
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5074975232146156e-05,
+      "loss": 0.2765,
+      "step": 14665
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5074639396839792e-05,
+      "loss": 0.1918,
+      "step": 14666
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5074303561533425e-05,
+      "loss": 0.4174,
+      "step": 14667
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5073967726227058e-05,
+      "loss": 0.3032,
+      "step": 14668
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5073631890920695e-05,
+      "loss": 0.2137,
+      "step": 14669
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5073296055614328e-05,
+      "loss": 0.2648,
+      "step": 14670
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5072960220307962e-05,
+      "loss": 0.3264,
+      "step": 14671
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5072624385001597e-05,
+      "loss": 0.5465,
+      "step": 14672
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5072288549695231e-05,
+      "loss": 0.3723,
+      "step": 14673
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5071952714388864e-05,
+      "loss": 0.3074,
+      "step": 14674
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.50716168790825e-05,
+      "loss": 0.2004,
+      "step": 14675
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5071281043776134e-05,
+      "loss": 0.3296,
+      "step": 14676
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5070945208469767e-05,
+      "loss": 0.3828,
+      "step": 14677
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5070609373163403e-05,
+      "loss": 0.1521,
+      "step": 14678
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5070273537857036e-05,
+      "loss": 0.2252,
+      "step": 14679
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.506993770255067e-05,
+      "loss": 0.4066,
+      "step": 14680
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5069601867244305e-05,
+      "loss": 0.2559,
+      "step": 14681
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.506926603193794e-05,
+      "loss": 0.5546,
+      "step": 14682
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5068930196631573e-05,
+      "loss": 0.2256,
+      "step": 14683
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.506859436132521e-05,
+      "loss": 0.3792,
+      "step": 14684
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5068258526018842e-05,
+      "loss": 0.3076,
+      "step": 14685
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5067922690712475e-05,
+      "loss": 0.145,
+      "step": 14686
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5067586855406111e-05,
+      "loss": 0.3623,
+      "step": 14687
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5067251020099744e-05,
+      "loss": 0.2809,
+      "step": 14688
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5066915184793377e-05,
+      "loss": 0.3125,
+      "step": 14689
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5066579349487014e-05,
+      "loss": 0.2142,
+      "step": 14690
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5066243514180647e-05,
+      "loss": 0.3425,
+      "step": 14691
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5065907678874281e-05,
+      "loss": 0.3949,
+      "step": 14692
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5065571843567916e-05,
+      "loss": 0.2595,
+      "step": 14693
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.506523600826155e-05,
+      "loss": 0.4431,
+      "step": 14694
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5064900172955184e-05,
+      "loss": 0.1972,
+      "step": 14695
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.506456433764882e-05,
+      "loss": 0.2357,
+      "step": 14696
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5064228502342453e-05,
+      "loss": 0.2203,
+      "step": 14697
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5063892667036086e-05,
+      "loss": 0.3649,
+      "step": 14698
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5063556831729722e-05,
+      "loss": 0.358,
+      "step": 14699
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5063220996423355e-05,
+      "loss": 0.4692,
+      "step": 14700
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.506288516111699e-05,
+      "loss": 0.2333,
+      "step": 14701
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5062549325810624e-05,
+      "loss": 0.3064,
+      "step": 14702
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5062213490504259e-05,
+      "loss": 0.2118,
+      "step": 14703
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5061877655197892e-05,
+      "loss": 0.4368,
+      "step": 14704
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5061541819891528e-05,
+      "loss": 0.3058,
+      "step": 14705
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5061205984585161e-05,
+      "loss": 0.3469,
+      "step": 14706
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5060870149278794e-05,
+      "loss": 0.2723,
+      "step": 14707
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5060534313972429e-05,
+      "loss": 0.3387,
+      "step": 14708
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5060198478666064e-05,
+      "loss": 0.2722,
+      "step": 14709
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5059862643359696e-05,
+      "loss": 0.5113,
+      "step": 14710
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5059526808053331e-05,
+      "loss": 0.2507,
+      "step": 14711
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5059190972746966e-05,
+      "loss": 0.2262,
+      "step": 14712
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.50588551374406e-05,
+      "loss": 0.1861,
+      "step": 14713
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5058519302134233e-05,
+      "loss": 0.4184,
+      "step": 14714
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.505818346682787e-05,
+      "loss": 0.3971,
+      "step": 14715
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5057847631521503e-05,
+      "loss": 0.3137,
+      "step": 14716
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5057511796215136e-05,
+      "loss": 0.2349,
+      "step": 14717
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5057175960908772e-05,
+      "loss": 0.4233,
+      "step": 14718
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5056840125602405e-05,
+      "loss": 0.2851,
+      "step": 14719
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.505650429029604e-05,
+      "loss": 0.3791,
+      "step": 14720
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5056168454989674e-05,
+      "loss": 0.2661,
+      "step": 14721
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5055832619683309e-05,
+      "loss": 0.2716,
+      "step": 14722
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5055496784376942e-05,
+      "loss": 0.2357,
+      "step": 14723
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5055160949070578e-05,
+      "loss": 0.3309,
+      "step": 14724
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5054825113764211e-05,
+      "loss": 0.1398,
+      "step": 14725
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5054489278457844e-05,
+      "loss": 0.2922,
+      "step": 14726
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.505415344315148e-05,
+      "loss": 0.3479,
+      "step": 14727
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5053817607845113e-05,
+      "loss": 0.2738,
+      "step": 14728
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5053481772538748e-05,
+      "loss": 0.5134,
+      "step": 14729
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5053145937232383e-05,
+      "loss": 0.2923,
+      "step": 14730
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5052810101926016e-05,
+      "loss": 0.2873,
+      "step": 14731
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.505247426661965e-05,
+      "loss": 0.3102,
+      "step": 14732
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5052138431313285e-05,
+      "loss": 0.5565,
+      "step": 14733
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.505180259600692e-05,
+      "loss": 0.3368,
+      "step": 14734
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5051466760700553e-05,
+      "loss": 0.2383,
+      "step": 14735
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5051130925394189e-05,
+      "loss": 0.2917,
+      "step": 14736
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5050795090087822e-05,
+      "loss": 0.3073,
+      "step": 14737
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5050459254781455e-05,
+      "loss": 0.2997,
+      "step": 14738
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5050123419475091e-05,
+      "loss": 0.2352,
+      "step": 14739
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5049787584168724e-05,
+      "loss": 0.3082,
+      "step": 14740
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5049451748862359e-05,
+      "loss": 0.4706,
+      "step": 14741
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5049115913555993e-05,
+      "loss": 0.3717,
+      "step": 14742
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5048780078249628e-05,
+      "loss": 0.2496,
+      "step": 14743
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5048444242943261e-05,
+      "loss": 0.263,
+      "step": 14744
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5048108407636897e-05,
+      "loss": 0.4125,
+      "step": 14745
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.504777257233053e-05,
+      "loss": 0.4641,
+      "step": 14746
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5047436737024163e-05,
+      "loss": 0.4148,
+      "step": 14747
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.50471009017178e-05,
+      "loss": 0.4923,
+      "step": 14748
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5046765066411432e-05,
+      "loss": 0.4305,
+      "step": 14749
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5046429231105067e-05,
+      "loss": 0.4879,
+      "step": 14750
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5046093395798702e-05,
+      "loss": 0.421,
+      "step": 14751
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5045757560492336e-05,
+      "loss": 0.2703,
+      "step": 14752
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.504542172518597e-05,
+      "loss": 0.2529,
+      "step": 14753
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5045085889879604e-05,
+      "loss": 0.2745,
+      "step": 14754
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5044750054573239e-05,
+      "loss": 0.2588,
+      "step": 14755
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5044414219266872e-05,
+      "loss": 0.3057,
+      "step": 14756
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5044078383960508e-05,
+      "loss": 0.3952,
+      "step": 14757
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5043742548654141e-05,
+      "loss": 0.329,
+      "step": 14758
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5043406713347774e-05,
+      "loss": 0.2504,
+      "step": 14759
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.504307087804141e-05,
+      "loss": 0.4108,
+      "step": 14760
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5042735042735043e-05,
+      "loss": 0.5297,
+      "step": 14761
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5042399207428678e-05,
+      "loss": 0.2723,
+      "step": 14762
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5042063372122312e-05,
+      "loss": 0.3989,
+      "step": 14763
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5041727536815947e-05,
+      "loss": 0.2833,
+      "step": 14764
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.504139170150958e-05,
+      "loss": 0.3213,
+      "step": 14765
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5041055866203216e-05,
+      "loss": 0.3179,
+      "step": 14766
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.504072003089685e-05,
+      "loss": 0.2415,
+      "step": 14767
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5040384195590482e-05,
+      "loss": 0.3201,
+      "step": 14768
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5040048360284119e-05,
+      "loss": 0.2453,
+      "step": 14769
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5039712524977752e-05,
+      "loss": 0.32,
+      "step": 14770
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5039376689671386e-05,
+      "loss": 0.2578,
+      "step": 14771
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5039040854365021e-05,
+      "loss": 0.2862,
+      "step": 14772
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5038705019058656e-05,
+      "loss": 0.2703,
+      "step": 14773
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5038369183752289e-05,
+      "loss": 0.258,
+      "step": 14774
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5038033348445923e-05,
+      "loss": 0.4372,
+      "step": 14775
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5037697513139558e-05,
+      "loss": 0.3449,
+      "step": 14776
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.503736167783319e-05,
+      "loss": 0.3434,
+      "step": 14777
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5037025842526827e-05,
+      "loss": 0.4198,
+      "step": 14778
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.503669000722046e-05,
+      "loss": 0.2458,
+      "step": 14779
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5036354171914093e-05,
+      "loss": 0.279,
+      "step": 14780
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.503601833660773e-05,
+      "loss": 0.4216,
+      "step": 14781
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5035682501301362e-05,
+      "loss": 0.4082,
+      "step": 14782
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5035346665994997e-05,
+      "loss": 0.4037,
+      "step": 14783
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5035010830688632e-05,
+      "loss": 0.3784,
+      "step": 14784
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5034674995382266e-05,
+      "loss": 0.3377,
+      "step": 14785
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.50343391600759e-05,
+      "loss": 0.3387,
+      "step": 14786
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5034003324769536e-05,
+      "loss": 0.2838,
+      "step": 14787
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.5033667489463168e-05,
+      "loss": 0.3269,
+      "step": 14788
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5033331654156801e-05,
+      "loss": 0.2437,
+      "step": 14789
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5032995818850438e-05,
+      "loss": 0.2467,
+      "step": 14790
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.503265998354407e-05,
+      "loss": 0.3096,
+      "step": 14791
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5032324148237705e-05,
+      "loss": 0.2935,
+      "step": 14792
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.503198831293134e-05,
+      "loss": 0.1751,
+      "step": 14793
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5031652477624975e-05,
+      "loss": 0.4373,
+      "step": 14794
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5031316642318608e-05,
+      "loss": 0.1949,
+      "step": 14795
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5030980807012244e-05,
+      "loss": 0.3377,
+      "step": 14796
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5030644971705877e-05,
+      "loss": 0.309,
+      "step": 14797
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.503030913639951e-05,
+      "loss": 0.3216,
+      "step": 14798
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5029973301093146e-05,
+      "loss": 0.247,
+      "step": 14799
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.502963746578678e-05,
+      "loss": 0.2942,
+      "step": 14800
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5029301630480412e-05,
+      "loss": 0.4586,
+      "step": 14801
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5028965795174048e-05,
+      "loss": 0.4649,
+      "step": 14802
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5028629959867681e-05,
+      "loss": 0.2518,
+      "step": 14803
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5028294124561316e-05,
+      "loss": 0.3273,
+      "step": 14804
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.502795828925495e-05,
+      "loss": 0.377,
+      "step": 14805
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5027622453948585e-05,
+      "loss": 0.4333,
+      "step": 14806
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5027286618642218e-05,
+      "loss": 0.3234,
+      "step": 14807
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5026950783335855e-05,
+      "loss": 0.3314,
+      "step": 14808
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5026614948029488e-05,
+      "loss": 0.2836,
+      "step": 14809
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.502627911272312e-05,
+      "loss": 0.3082,
+      "step": 14810
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5025943277416757e-05,
+      "loss": 0.4383,
+      "step": 14811
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.502560744211039e-05,
+      "loss": 0.3371,
+      "step": 14812
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5025271606804025e-05,
+      "loss": 0.2008,
+      "step": 14813
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5024935771497659e-05,
+      "loss": 0.34,
+      "step": 14814
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5024599936191294e-05,
+      "loss": 0.3688,
+      "step": 14815
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5024264100884927e-05,
+      "loss": 0.2042,
+      "step": 14816
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5023928265578563e-05,
+      "loss": 0.5183,
+      "step": 14817
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5023592430272196e-05,
+      "loss": 0.4227,
+      "step": 14818
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5023256594965829e-05,
+      "loss": 0.1793,
+      "step": 14819
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5022920759659465e-05,
+      "loss": 0.3416,
+      "step": 14820
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5022584924353098e-05,
+      "loss": 0.3616,
+      "step": 14821
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5022249089046731e-05,
+      "loss": 0.6377,
+      "step": 14822
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5021913253740368e-05,
+      "loss": 0.2018,
+      "step": 14823
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5021577418434e-05,
+      "loss": 0.3205,
+      "step": 14824
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5021241583127635e-05,
+      "loss": 0.2964,
+      "step": 14825
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.502090574782127e-05,
+      "loss": 0.2549,
+      "step": 14826
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5020569912514904e-05,
+      "loss": 0.1979,
+      "step": 14827
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5020234077208537e-05,
+      "loss": 0.3572,
+      "step": 14828
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5019898241902174e-05,
+      "loss": 0.3031,
+      "step": 14829
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5019562406595807e-05,
+      "loss": 0.5027,
+      "step": 14830
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.501922657128944e-05,
+      "loss": 0.3958,
+      "step": 14831
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5018890735983076e-05,
+      "loss": 0.4526,
+      "step": 14832
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5018554900676709e-05,
+      "loss": 0.1473,
+      "step": 14833
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5018219065370344e-05,
+      "loss": 0.3408,
+      "step": 14834
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5017883230063978e-05,
+      "loss": 0.2627,
+      "step": 14835
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5017547394757613e-05,
+      "loss": 0.248,
+      "step": 14836
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5017211559451246e-05,
+      "loss": 0.2126,
+      "step": 14837
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5016875724144882e-05,
+      "loss": 0.3714,
+      "step": 14838
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5016539888838515e-05,
+      "loss": 0.2179,
+      "step": 14839
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5016204053532148e-05,
+      "loss": 0.4868,
+      "step": 14840
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5015868218225784e-05,
+      "loss": 0.4267,
+      "step": 14841
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5015532382919417e-05,
+      "loss": 0.2539,
+      "step": 14842
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5015196547613052e-05,
+      "loss": 0.395,
+      "step": 14843
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5014860712306687e-05,
+      "loss": 0.2165,
+      "step": 14844
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.501452487700032e-05,
+      "loss": 0.2871,
+      "step": 14845
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5014189041693954e-05,
+      "loss": 0.2073,
+      "step": 14846
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5013853206387589e-05,
+      "loss": 0.2238,
+      "step": 14847
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5013517371081224e-05,
+      "loss": 0.345,
+      "step": 14848
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5013181535774857e-05,
+      "loss": 0.4813,
+      "step": 14849
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5012845700468493e-05,
+      "loss": 0.4833,
+      "step": 14850
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5012509865162126e-05,
+      "loss": 0.2353,
+      "step": 14851
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5012174029855759e-05,
+      "loss": 0.2427,
+      "step": 14852
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5011838194549395e-05,
+      "loss": 0.3004,
+      "step": 14853
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5011502359243028e-05,
+      "loss": 0.2729,
+      "step": 14854
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5011166523936663e-05,
+      "loss": 0.3561,
+      "step": 14855
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5010830688630297e-05,
+      "loss": 0.3915,
+      "step": 14856
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5010494853323932e-05,
+      "loss": 0.3587,
+      "step": 14857
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5010159018017565e-05,
+      "loss": 0.3515,
+      "step": 14858
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5009823182711201e-05,
+      "loss": 0.1606,
+      "step": 14859
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5009487347404834e-05,
+      "loss": 0.2921,
+      "step": 14860
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5009151512098467e-05,
+      "loss": 0.282,
+      "step": 14861
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5008815676792104e-05,
+      "loss": 0.3684,
+      "step": 14862
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5008479841485737e-05,
+      "loss": 0.4755,
+      "step": 14863
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5008144006179371e-05,
+      "loss": 0.3916,
+      "step": 14864
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5007808170873006e-05,
+      "loss": 0.4587,
+      "step": 14865
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5007472335566639e-05,
+      "loss": 0.3237,
+      "step": 14866
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5007136500260273e-05,
+      "loss": 0.4664,
+      "step": 14867
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5006800664953908e-05,
+      "loss": 0.5397,
+      "step": 14868
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5006464829647543e-05,
+      "loss": 0.3323,
+      "step": 14869
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5006128994341176e-05,
+      "loss": 0.5548,
+      "step": 14870
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5005793159034812e-05,
+      "loss": 0.2828,
+      "step": 14871
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5005457323728445e-05,
+      "loss": 0.3611,
+      "step": 14872
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5005121488422078e-05,
+      "loss": 0.5111,
+      "step": 14873
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5004785653115714e-05,
+      "loss": 0.291,
+      "step": 14874
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5004449817809347e-05,
+      "loss": 0.4967,
+      "step": 14875
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5004113982502982e-05,
+      "loss": 0.3595,
+      "step": 14876
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5003778147196617e-05,
+      "loss": 0.2277,
+      "step": 14877
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5003442311890251e-05,
+      "loss": 0.4933,
+      "step": 14878
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5003106476583884e-05,
+      "loss": 0.4039,
+      "step": 14879
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.500277064127752e-05,
+      "loss": 0.4112,
+      "step": 14880
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5002434805971153e-05,
+      "loss": 0.3952,
+      "step": 14881
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5002098970664786e-05,
+      "loss": 0.4314,
+      "step": 14882
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5001763135358423e-05,
+      "loss": 0.4125,
+      "step": 14883
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5001427300052056e-05,
+      "loss": 0.4959,
+      "step": 14884
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.500109146474569e-05,
+      "loss": 0.2767,
+      "step": 14885
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5000755629439325e-05,
+      "loss": 0.1918,
+      "step": 14886
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.500041979413296e-05,
+      "loss": 0.3773,
+      "step": 14887
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.5000083958826593e-05,
+      "loss": 0.192,
+      "step": 14888
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4999748123520226e-05,
+      "loss": 0.278,
+      "step": 14889
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4999412288213862e-05,
+      "loss": 0.3755,
+      "step": 14890
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4999076452907495e-05,
+      "loss": 0.3439,
+      "step": 14891
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4998740617601128e-05,
+      "loss": 0.3197,
+      "step": 14892
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4998404782294764e-05,
+      "loss": 0.2866,
+      "step": 14893
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4998068946988397e-05,
+      "loss": 0.2885,
+      "step": 14894
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4997733111682032e-05,
+      "loss": 0.4674,
+      "step": 14895
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4997397276375666e-05,
+      "loss": 0.3798,
+      "step": 14896
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4997061441069301e-05,
+      "loss": 0.2349,
+      "step": 14897
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4996725605762934e-05,
+      "loss": 0.39,
+      "step": 14898
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.499638977045657e-05,
+      "loss": 0.2556,
+      "step": 14899
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4996053935150203e-05,
+      "loss": 0.2988,
+      "step": 14900
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4995718099843836e-05,
+      "loss": 0.2648,
+      "step": 14901
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4995382264537473e-05,
+      "loss": 0.4115,
+      "step": 14902
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4995046429231105e-05,
+      "loss": 0.3889,
+      "step": 14903
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.499471059392474e-05,
+      "loss": 0.2945,
+      "step": 14904
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4994374758618375e-05,
+      "loss": 0.1859,
+      "step": 14905
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.499403892331201e-05,
+      "loss": 0.3218,
+      "step": 14906
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4993703088005642e-05,
+      "loss": 0.3326,
+      "step": 14907
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4993367252699279e-05,
+      "loss": 0.2333,
+      "step": 14908
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4993031417392912e-05,
+      "loss": 0.4006,
+      "step": 14909
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4992695582086545e-05,
+      "loss": 0.3178,
+      "step": 14910
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4992359746780181e-05,
+      "loss": 0.2513,
+      "step": 14911
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4992023911473814e-05,
+      "loss": 0.3113,
+      "step": 14912
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4991688076167447e-05,
+      "loss": 0.292,
+      "step": 14913
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4991352240861083e-05,
+      "loss": 0.4441,
+      "step": 14914
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4991016405554716e-05,
+      "loss": 0.2343,
+      "step": 14915
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.499068057024835e-05,
+      "loss": 0.259,
+      "step": 14916
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4990344734941985e-05,
+      "loss": 0.2931,
+      "step": 14917
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.499000889963562e-05,
+      "loss": 0.4112,
+      "step": 14918
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4989673064329253e-05,
+      "loss": 0.5088,
+      "step": 14919
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.498933722902289e-05,
+      "loss": 0.4111,
+      "step": 14920
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4989001393716522e-05,
+      "loss": 0.2326,
+      "step": 14921
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4988665558410155e-05,
+      "loss": 0.4811,
+      "step": 14922
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4988329723103792e-05,
+      "loss": 0.5256,
+      "step": 14923
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4987993887797425e-05,
+      "loss": 0.3842,
+      "step": 14924
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.498765805249106e-05,
+      "loss": 0.1623,
+      "step": 14925
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4987322217184694e-05,
+      "loss": 0.4496,
+      "step": 14926
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4986986381878329e-05,
+      "loss": 0.2941,
+      "step": 14927
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4986650546571962e-05,
+      "loss": 0.3209,
+      "step": 14928
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4986314711265598e-05,
+      "loss": 0.4428,
+      "step": 14929
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.498597887595923e-05,
+      "loss": 0.3959,
+      "step": 14930
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4985643040652864e-05,
+      "loss": 0.4467,
+      "step": 14931
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.49853072053465e-05,
+      "loss": 0.2489,
+      "step": 14932
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4984971370040133e-05,
+      "loss": 0.2919,
+      "step": 14933
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4984635534733768e-05,
+      "loss": 0.3555,
+      "step": 14934
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4984299699427402e-05,
+      "loss": 0.2299,
+      "step": 14935
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4983963864121035e-05,
+      "loss": 0.3034,
+      "step": 14936
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.498362802881467e-05,
+      "loss": 0.383,
+      "step": 14937
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4983292193508305e-05,
+      "loss": 0.3115,
+      "step": 14938
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.498295635820194e-05,
+      "loss": 0.3984,
+      "step": 14939
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4982620522895572e-05,
+      "loss": 0.3561,
+      "step": 14940
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4982284687589209e-05,
+      "loss": 0.3527,
+      "step": 14941
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4981948852282842e-05,
+      "loss": 0.2066,
+      "step": 14942
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4981613016976474e-05,
+      "loss": 0.3117,
+      "step": 14943
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.498127718167011e-05,
+      "loss": 0.4129,
+      "step": 14944
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4980941346363744e-05,
+      "loss": 0.5989,
+      "step": 14945
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4980605511057378e-05,
+      "loss": 0.2924,
+      "step": 14946
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4980269675751013e-05,
+      "loss": 0.3829,
+      "step": 14947
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4979933840444648e-05,
+      "loss": 0.2858,
+      "step": 14948
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.497959800513828e-05,
+      "loss": 0.5562,
+      "step": 14949
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4979262169831917e-05,
+      "loss": 0.5781,
+      "step": 14950
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.497892633452555e-05,
+      "loss": 0.2824,
+      "step": 14951
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4978590499219183e-05,
+      "loss": 0.2884,
+      "step": 14952
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.497825466391282e-05,
+      "loss": 0.3514,
+      "step": 14953
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4977918828606452e-05,
+      "loss": 0.4954,
+      "step": 14954
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4977582993300087e-05,
+      "loss": 0.3487,
+      "step": 14955
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4977247157993721e-05,
+      "loss": 0.3468,
+      "step": 14956
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4976911322687354e-05,
+      "loss": 0.3289,
+      "step": 14957
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4976575487380989e-05,
+      "loss": 0.2841,
+      "step": 14958
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4976239652074624e-05,
+      "loss": 0.4577,
+      "step": 14959
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4975903816768258e-05,
+      "loss": 0.4073,
+      "step": 14960
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4975567981461891e-05,
+      "loss": 0.3978,
+      "step": 14961
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4975232146155528e-05,
+      "loss": 0.3371,
+      "step": 14962
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.497489631084916e-05,
+      "loss": 0.3693,
+      "step": 14963
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4974560475542794e-05,
+      "loss": 0.464,
+      "step": 14964
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.497422464023643e-05,
+      "loss": 0.4141,
+      "step": 14965
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4973888804930063e-05,
+      "loss": 0.2638,
+      "step": 14966
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4973552969623698e-05,
+      "loss": 0.3904,
+      "step": 14967
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4973217134317332e-05,
+      "loss": 0.3705,
+      "step": 14968
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4972881299010967e-05,
+      "loss": 0.2787,
+      "step": 14969
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.49725454637046e-05,
+      "loss": 0.33,
+      "step": 14970
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4972209628398236e-05,
+      "loss": 0.2449,
+      "step": 14971
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4971873793091869e-05,
+      "loss": 0.3902,
+      "step": 14972
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4971537957785502e-05,
+      "loss": 0.3321,
+      "step": 14973
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4971202122479138e-05,
+      "loss": 0.2159,
+      "step": 14974
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4970866287172771e-05,
+      "loss": 0.256,
+      "step": 14975
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4970530451866406e-05,
+      "loss": 0.2475,
+      "step": 14976
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.497019461656004e-05,
+      "loss": 0.321,
+      "step": 14977
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4969858781253675e-05,
+      "loss": 0.3097,
+      "step": 14978
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4969522945947308e-05,
+      "loss": 0.2321,
+      "step": 14979
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4969187110640943e-05,
+      "loss": 0.2129,
+      "step": 14980
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4968851275334578e-05,
+      "loss": 0.3961,
+      "step": 14981
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.496851544002821e-05,
+      "loss": 0.288,
+      "step": 14982
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4968179604721847e-05,
+      "loss": 0.2854,
+      "step": 14983
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.496784376941548e-05,
+      "loss": 0.4681,
+      "step": 14984
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4967507934109113e-05,
+      "loss": 0.2717,
+      "step": 14985
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4967172098802749e-05,
+      "loss": 0.2115,
+      "step": 14986
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.4966836263496382e-05,
+      "loss": 0.2708,
+      "step": 14987
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4966500428190017e-05,
+      "loss": 0.2011,
+      "step": 14988
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4966164592883651e-05,
+      "loss": 0.2999,
+      "step": 14989
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4965828757577286e-05,
+      "loss": 0.246,
+      "step": 14990
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4965492922270919e-05,
+      "loss": 0.2018,
+      "step": 14991
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4965157086964555e-05,
+      "loss": 0.3521,
+      "step": 14992
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4964821251658188e-05,
+      "loss": 0.2812,
+      "step": 14993
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4964485416351821e-05,
+      "loss": 0.3094,
+      "step": 14994
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4964149581045457e-05,
+      "loss": 0.267,
+      "step": 14995
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.496381374573909e-05,
+      "loss": 0.5734,
+      "step": 14996
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4963477910432725e-05,
+      "loss": 0.423,
+      "step": 14997
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.496314207512636e-05,
+      "loss": 0.2245,
+      "step": 14998
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4962806239819994e-05,
+      "loss": 0.3224,
+      "step": 14999
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4962470404513627e-05,
+      "loss": 0.2223,
+      "step": 15000
+    },
+    {
+      "epoch": 0.76,
+      "eval_accuracy": 0.8793151580602268,
+      "eval_f1_microaverage": 0.8793151580602268,
+      "eval_loss": 0.32314229011535645,
+      "eval_precision_macroaverage": 0.7486141742760165,
+      "eval_precision_microaverage": 0.8793151580602268,
+      "eval_recall_macroaverage": 0.6447705849870043,
+      "eval_recall_microaverage": 0.8793151580602268,
+      "eval_runtime": 1173.0066,
+      "eval_samples_per_second": 232.086,
+      "eval_steps_per_second": 3.627,
+      "step": 15000
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4962134569207262e-05,
+      "loss": 0.1692,
+      "step": 15001
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4961798733900897e-05,
+      "loss": 0.2621,
+      "step": 15002
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.496146289859453e-05,
+      "loss": 0.4787,
+      "step": 15003
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4961127063288166e-05,
+      "loss": 0.3664,
+      "step": 15004
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4960791227981799e-05,
+      "loss": 0.2541,
+      "step": 15005
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4960455392675432e-05,
+      "loss": 0.2277,
+      "step": 15006
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4960119557369068e-05,
+      "loss": 0.3163,
+      "step": 15007
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4959783722062701e-05,
+      "loss": 0.33,
+      "step": 15008
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4959447886756336e-05,
+      "loss": 0.3529,
+      "step": 15009
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.495911205144997e-05,
+      "loss": 0.3441,
+      "step": 15010
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4958776216143605e-05,
+      "loss": 0.3345,
+      "step": 15011
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4958440380837238e-05,
+      "loss": 0.2552,
+      "step": 15012
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4958104545530874e-05,
+      "loss": 0.3919,
+      "step": 15013
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4957768710224507e-05,
+      "loss": 0.2109,
+      "step": 15014
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.495743287491814e-05,
+      "loss": 0.6745,
+      "step": 15015
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4957097039611777e-05,
+      "loss": 0.5166,
+      "step": 15016
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.495676120430541e-05,
+      "loss": 0.3834,
+      "step": 15017
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4956425368999044e-05,
+      "loss": 0.453,
+      "step": 15018
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4956089533692679e-05,
+      "loss": 0.3844,
+      "step": 15019
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4955753698386314e-05,
+      "loss": 0.2741,
+      "step": 15020
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4955417863079946e-05,
+      "loss": 0.3334,
+      "step": 15021
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4955082027773583e-05,
+      "loss": 0.2722,
+      "step": 15022
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4954746192467216e-05,
+      "loss": 0.3943,
+      "step": 15023
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4954410357160849e-05,
+      "loss": 0.363,
+      "step": 15024
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4954074521854485e-05,
+      "loss": 0.2365,
+      "step": 15025
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4953738686548118e-05,
+      "loss": 0.3557,
+      "step": 15026
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4953402851241751e-05,
+      "loss": 0.3205,
+      "step": 15027
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4953067015935387e-05,
+      "loss": 0.3048,
+      "step": 15028
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.495273118062902e-05,
+      "loss": 0.3565,
+      "step": 15029
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4952395345322655e-05,
+      "loss": 0.3934,
+      "step": 15030
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.495205951001629e-05,
+      "loss": 0.2779,
+      "step": 15031
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4951723674709924e-05,
+      "loss": 0.2437,
+      "step": 15032
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4951387839403557e-05,
+      "loss": 0.4639,
+      "step": 15033
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4951052004097193e-05,
+      "loss": 0.2559,
+      "step": 15034
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4950716168790826e-05,
+      "loss": 0.3063,
+      "step": 15035
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.495038033348446e-05,
+      "loss": 0.3199,
+      "step": 15036
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4950044498178096e-05,
+      "loss": 0.4517,
+      "step": 15037
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4949708662871729e-05,
+      "loss": 0.3485,
+      "step": 15038
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4949372827565363e-05,
+      "loss": 0.3063,
+      "step": 15039
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4949036992258998e-05,
+      "loss": 0.2025,
+      "step": 15040
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4948701156952633e-05,
+      "loss": 0.2423,
+      "step": 15041
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4948365321646266e-05,
+      "loss": 0.4006,
+      "step": 15042
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4948029486339902e-05,
+      "loss": 0.3794,
+      "step": 15043
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4947693651033535e-05,
+      "loss": 0.2615,
+      "step": 15044
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4947357815727168e-05,
+      "loss": 0.4426,
+      "step": 15045
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4947021980420804e-05,
+      "loss": 0.2081,
+      "step": 15046
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4946686145114437e-05,
+      "loss": 0.3319,
+      "step": 15047
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.494635030980807e-05,
+      "loss": 0.194,
+      "step": 15048
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4946014474501706e-05,
+      "loss": 0.2115,
+      "step": 15049
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.494567863919534e-05,
+      "loss": 0.2793,
+      "step": 15050
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4945342803888974e-05,
+      "loss": 0.4307,
+      "step": 15051
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4945006968582609e-05,
+      "loss": 0.3328,
+      "step": 15052
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4944671133276243e-05,
+      "loss": 0.3839,
+      "step": 15053
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4944335297969876e-05,
+      "loss": 0.3551,
+      "step": 15054
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4943999462663513e-05,
+      "loss": 0.2534,
+      "step": 15055
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4943663627357146e-05,
+      "loss": 0.2407,
+      "step": 15056
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4943327792050779e-05,
+      "loss": 0.3925,
+      "step": 15057
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4942991956744415e-05,
+      "loss": 0.3508,
+      "step": 15058
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4942656121438048e-05,
+      "loss": 0.1997,
+      "step": 15059
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4942320286131682e-05,
+      "loss": 0.2823,
+      "step": 15060
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4941984450825317e-05,
+      "loss": 0.2859,
+      "step": 15061
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4941648615518952e-05,
+      "loss": 0.4941,
+      "step": 15062
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4941312780212585e-05,
+      "loss": 0.4316,
+      "step": 15063
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4940976944906221e-05,
+      "loss": 0.2129,
+      "step": 15064
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4940641109599854e-05,
+      "loss": 0.3189,
+      "step": 15065
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4940305274293487e-05,
+      "loss": 0.4,
+      "step": 15066
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4939969438987123e-05,
+      "loss": 0.2267,
+      "step": 15067
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4939633603680756e-05,
+      "loss": 0.1555,
+      "step": 15068
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4939297768374391e-05,
+      "loss": 0.4065,
+      "step": 15069
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4938961933068024e-05,
+      "loss": 0.3245,
+      "step": 15070
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4938626097761658e-05,
+      "loss": 0.62,
+      "step": 15071
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4938290262455293e-05,
+      "loss": 0.1782,
+      "step": 15072
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4937954427148926e-05,
+      "loss": 0.2042,
+      "step": 15073
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4937618591842562e-05,
+      "loss": 0.3873,
+      "step": 15074
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4937282756536195e-05,
+      "loss": 0.182,
+      "step": 15075
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4936946921229828e-05,
+      "loss": 0.4149,
+      "step": 15076
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4936611085923465e-05,
+      "loss": 0.2479,
+      "step": 15077
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4936275250617098e-05,
+      "loss": 0.3097,
+      "step": 15078
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4935939415310732e-05,
+      "loss": 0.3811,
+      "step": 15079
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4935603580004367e-05,
+      "loss": 0.1593,
+      "step": 15080
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4935267744698002e-05,
+      "loss": 0.3378,
+      "step": 15081
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4934931909391635e-05,
+      "loss": 0.3681,
+      "step": 15082
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4934596074085271e-05,
+      "loss": 0.1031,
+      "step": 15083
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4934260238778904e-05,
+      "loss": 0.4103,
+      "step": 15084
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4933924403472537e-05,
+      "loss": 0.628,
+      "step": 15085
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4933588568166173e-05,
+      "loss": 0.2751,
+      "step": 15086
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4933252732859806e-05,
+      "loss": 0.2207,
+      "step": 15087
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.493291689755344e-05,
+      "loss": 0.3414,
+      "step": 15088
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4932581062247075e-05,
+      "loss": 0.2654,
+      "step": 15089
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.493224522694071e-05,
+      "loss": 0.2164,
+      "step": 15090
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4931909391634343e-05,
+      "loss": 0.3197,
+      "step": 15091
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4931573556327978e-05,
+      "loss": 0.4256,
+      "step": 15092
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4931237721021612e-05,
+      "loss": 0.2122,
+      "step": 15093
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4930901885715245e-05,
+      "loss": 0.3166,
+      "step": 15094
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4930566050408882e-05,
+      "loss": 0.4311,
+      "step": 15095
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4930230215102515e-05,
+      "loss": 0.1899,
+      "step": 15096
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4929894379796147e-05,
+      "loss": 0.2415,
+      "step": 15097
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4929558544489784e-05,
+      "loss": 0.3982,
+      "step": 15098
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4929222709183417e-05,
+      "loss": 0.2657,
+      "step": 15099
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4928886873877051e-05,
+      "loss": 0.5468,
+      "step": 15100
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4928551038570686e-05,
+      "loss": 0.3413,
+      "step": 15101
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.492821520326432e-05,
+      "loss": 0.2508,
+      "step": 15102
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4927879367957954e-05,
+      "loss": 0.4485,
+      "step": 15103
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.492754353265159e-05,
+      "loss": 0.3424,
+      "step": 15104
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4927207697345223e-05,
+      "loss": 0.4053,
+      "step": 15105
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4926871862038856e-05,
+      "loss": 0.2894,
+      "step": 15106
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4926536026732492e-05,
+      "loss": 0.2623,
+      "step": 15107
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4926200191426125e-05,
+      "loss": 0.217,
+      "step": 15108
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.492586435611976e-05,
+      "loss": 0.3474,
+      "step": 15109
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4925528520813394e-05,
+      "loss": 0.3623,
+      "step": 15110
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4925192685507029e-05,
+      "loss": 0.2826,
+      "step": 15111
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4924856850200662e-05,
+      "loss": 0.3566,
+      "step": 15112
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4924521014894298e-05,
+      "loss": 0.3396,
+      "step": 15113
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4924185179587931e-05,
+      "loss": 0.2931,
+      "step": 15114
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4923849344281564e-05,
+      "loss": 0.2349,
+      "step": 15115
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.49235135089752e-05,
+      "loss": 0.2961,
+      "step": 15116
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4923177673668834e-05,
+      "loss": 0.1789,
+      "step": 15117
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4922841838362467e-05,
+      "loss": 0.3789,
+      "step": 15118
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4922506003056103e-05,
+      "loss": 0.4477,
+      "step": 15119
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4922170167749736e-05,
+      "loss": 0.3474,
+      "step": 15120
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.492183433244337e-05,
+      "loss": 0.2363,
+      "step": 15121
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4921498497137005e-05,
+      "loss": 0.365,
+      "step": 15122
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.492116266183064e-05,
+      "loss": 0.254,
+      "step": 15123
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4920826826524273e-05,
+      "loss": 0.4164,
+      "step": 15124
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4920490991217909e-05,
+      "loss": 0.3863,
+      "step": 15125
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4920155155911542e-05,
+      "loss": 0.2538,
+      "step": 15126
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4919819320605175e-05,
+      "loss": 0.2846,
+      "step": 15127
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4919483485298811e-05,
+      "loss": 0.2251,
+      "step": 15128
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4919147649992444e-05,
+      "loss": 0.3184,
+      "step": 15129
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4918811814686079e-05,
+      "loss": 0.3781,
+      "step": 15130
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4918475979379714e-05,
+      "loss": 0.4563,
+      "step": 15131
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4918140144073348e-05,
+      "loss": 0.6093,
+      "step": 15132
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4917804308766981e-05,
+      "loss": 0.3875,
+      "step": 15133
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4917468473460618e-05,
+      "loss": 0.3858,
+      "step": 15134
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.491713263815425e-05,
+      "loss": 0.2392,
+      "step": 15135
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4916796802847883e-05,
+      "loss": 0.2975,
+      "step": 15136
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.491646096754152e-05,
+      "loss": 0.4467,
+      "step": 15137
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4916125132235153e-05,
+      "loss": 0.2355,
+      "step": 15138
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4915789296928786e-05,
+      "loss": 0.3199,
+      "step": 15139
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4915453461622422e-05,
+      "loss": 0.3068,
+      "step": 15140
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4915117626316055e-05,
+      "loss": 0.2976,
+      "step": 15141
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.491478179100969e-05,
+      "loss": 0.3147,
+      "step": 15142
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4914445955703324e-05,
+      "loss": 0.481,
+      "step": 15143
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4914110120396959e-05,
+      "loss": 0.2271,
+      "step": 15144
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4913774285090592e-05,
+      "loss": 0.3331,
+      "step": 15145
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4913438449784228e-05,
+      "loss": 0.4601,
+      "step": 15146
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4913102614477861e-05,
+      "loss": 0.2307,
+      "step": 15147
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4912766779171494e-05,
+      "loss": 0.2367,
+      "step": 15148
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.491243094386513e-05,
+      "loss": 0.2513,
+      "step": 15149
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4912095108558763e-05,
+      "loss": 0.2206,
+      "step": 15150
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4911759273252398e-05,
+      "loss": 0.1651,
+      "step": 15151
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4911423437946033e-05,
+      "loss": 0.4056,
+      "step": 15152
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4911087602639667e-05,
+      "loss": 0.3673,
+      "step": 15153
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.49107517673333e-05,
+      "loss": 0.2086,
+      "step": 15154
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4910415932026937e-05,
+      "loss": 0.2088,
+      "step": 15155
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.491008009672057e-05,
+      "loss": 0.2732,
+      "step": 15156
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4909744261414203e-05,
+      "loss": 0.1451,
+      "step": 15157
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4909408426107839e-05,
+      "loss": 0.2904,
+      "step": 15158
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4909072590801472e-05,
+      "loss": 0.3829,
+      "step": 15159
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4908736755495105e-05,
+      "loss": 0.388,
+      "step": 15160
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4908400920188741e-05,
+      "loss": 0.4011,
+      "step": 15161
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4908065084882374e-05,
+      "loss": 0.2815,
+      "step": 15162
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4907729249576009e-05,
+      "loss": 0.386,
+      "step": 15163
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4907393414269643e-05,
+      "loss": 0.4526,
+      "step": 15164
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4907057578963278e-05,
+      "loss": 0.4158,
+      "step": 15165
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4906721743656911e-05,
+      "loss": 0.4671,
+      "step": 15166
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4906385908350547e-05,
+      "loss": 0.4403,
+      "step": 15167
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.490605007304418e-05,
+      "loss": 0.2712,
+      "step": 15168
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4905714237737813e-05,
+      "loss": 0.3255,
+      "step": 15169
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.490537840243145e-05,
+      "loss": 0.3939,
+      "step": 15170
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4905042567125083e-05,
+      "loss": 0.3648,
+      "step": 15171
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4904706731818717e-05,
+      "loss": 0.302,
+      "step": 15172
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4904370896512352e-05,
+      "loss": 0.4464,
+      "step": 15173
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4904035061205987e-05,
+      "loss": 0.2648,
+      "step": 15174
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.490369922589962e-05,
+      "loss": 0.298,
+      "step": 15175
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4903363390593256e-05,
+      "loss": 0.2471,
+      "step": 15176
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4903027555286889e-05,
+      "loss": 0.2633,
+      "step": 15177
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4902691719980522e-05,
+      "loss": 0.3666,
+      "step": 15178
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4902355884674158e-05,
+      "loss": 0.3098,
+      "step": 15179
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4902020049367791e-05,
+      "loss": 0.3315,
+      "step": 15180
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4901684214061426e-05,
+      "loss": 0.1967,
+      "step": 15181
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.490134837875506e-05,
+      "loss": 0.4211,
+      "step": 15182
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4901012543448693e-05,
+      "loss": 0.4029,
+      "step": 15183
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4900676708142328e-05,
+      "loss": 0.2319,
+      "step": 15184
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4900340872835963e-05,
+      "loss": 0.2959,
+      "step": 15185
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.4900005037529597e-05,
+      "loss": 0.3508,
+      "step": 15186
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.489966920222323e-05,
+      "loss": 0.3734,
+      "step": 15187
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4899333366916867e-05,
+      "loss": 0.2513,
+      "step": 15188
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.48989975316105e-05,
+      "loss": 0.3876,
+      "step": 15189
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4898661696304132e-05,
+      "loss": 0.1467,
+      "step": 15190
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4898325860997769e-05,
+      "loss": 0.4071,
+      "step": 15191
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4897990025691402e-05,
+      "loss": 0.3297,
+      "step": 15192
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4897654190385036e-05,
+      "loss": 0.3824,
+      "step": 15193
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4897318355078671e-05,
+      "loss": 0.3931,
+      "step": 15194
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4896982519772306e-05,
+      "loss": 0.2222,
+      "step": 15195
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4896646684465939e-05,
+      "loss": 0.2853,
+      "step": 15196
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4896310849159575e-05,
+      "loss": 0.3854,
+      "step": 15197
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4895975013853208e-05,
+      "loss": 0.3888,
+      "step": 15198
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4895639178546841e-05,
+      "loss": 0.3859,
+      "step": 15199
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4895303343240477e-05,
+      "loss": 0.5559,
+      "step": 15200
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.489496750793411e-05,
+      "loss": 0.242,
+      "step": 15201
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4894631672627745e-05,
+      "loss": 0.2194,
+      "step": 15202
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.489429583732138e-05,
+      "loss": 0.259,
+      "step": 15203
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4893960002015014e-05,
+      "loss": 0.3077,
+      "step": 15204
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4893624166708647e-05,
+      "loss": 0.3738,
+      "step": 15205
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4893288331402282e-05,
+      "loss": 0.2797,
+      "step": 15206
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4892952496095916e-05,
+      "loss": 0.2471,
+      "step": 15207
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.489261666078955e-05,
+      "loss": 0.3788,
+      "step": 15208
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4892280825483186e-05,
+      "loss": 0.4389,
+      "step": 15209
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4891944990176819e-05,
+      "loss": 0.3206,
+      "step": 15210
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4891609154870452e-05,
+      "loss": 0.4224,
+      "step": 15211
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4891273319564088e-05,
+      "loss": 0.2822,
+      "step": 15212
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.489093748425772e-05,
+      "loss": 0.266,
+      "step": 15213
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4890601648951355e-05,
+      "loss": 0.2658,
+      "step": 15214
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.489026581364499e-05,
+      "loss": 0.3366,
+      "step": 15215
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4889929978338625e-05,
+      "loss": 0.3659,
+      "step": 15216
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4889594143032258e-05,
+      "loss": 0.2627,
+      "step": 15217
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4889258307725894e-05,
+      "loss": 0.248,
+      "step": 15218
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4888922472419527e-05,
+      "loss": 0.4445,
+      "step": 15219
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.488858663711316e-05,
+      "loss": 0.3088,
+      "step": 15220
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4888250801806796e-05,
+      "loss": 0.4479,
+      "step": 15221
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.488791496650043e-05,
+      "loss": 0.2537,
+      "step": 15222
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4887579131194064e-05,
+      "loss": 0.5565,
+      "step": 15223
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4887243295887699e-05,
+      "loss": 0.4276,
+      "step": 15224
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4886907460581333e-05,
+      "loss": 0.4269,
+      "step": 15225
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4886571625274966e-05,
+      "loss": 0.2346,
+      "step": 15226
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.48862357899686e-05,
+      "loss": 0.3666,
+      "step": 15227
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4885899954662235e-05,
+      "loss": 0.3228,
+      "step": 15228
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4885564119355868e-05,
+      "loss": 0.4349,
+      "step": 15229
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4885228284049505e-05,
+      "loss": 0.2213,
+      "step": 15230
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4884892448743138e-05,
+      "loss": 0.3896,
+      "step": 15231
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.488455661343677e-05,
+      "loss": 0.2645,
+      "step": 15232
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4884220778130407e-05,
+      "loss": 0.2202,
+      "step": 15233
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.488388494282404e-05,
+      "loss": 0.334,
+      "step": 15234
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4883549107517675e-05,
+      "loss": 0.3852,
+      "step": 15235
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.488321327221131e-05,
+      "loss": 0.2258,
+      "step": 15236
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4882877436904944e-05,
+      "loss": 0.3384,
+      "step": 15237
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4882541601598577e-05,
+      "loss": 0.4621,
+      "step": 15238
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4882205766292213e-05,
+      "loss": 0.5102,
+      "step": 15239
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4881869930985846e-05,
+      "loss": 0.2838,
+      "step": 15240
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4881534095679479e-05,
+      "loss": 0.3316,
+      "step": 15241
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4881198260373115e-05,
+      "loss": 0.4547,
+      "step": 15242
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4880862425066748e-05,
+      "loss": 0.3377,
+      "step": 15243
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4880526589760383e-05,
+      "loss": 0.333,
+      "step": 15244
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4880190754454018e-05,
+      "loss": 0.3422,
+      "step": 15245
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4879854919147652e-05,
+      "loss": 0.368,
+      "step": 15246
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4879519083841285e-05,
+      "loss": 0.3646,
+      "step": 15247
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4879183248534922e-05,
+      "loss": 0.4835,
+      "step": 15248
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4878847413228555e-05,
+      "loss": 0.3232,
+      "step": 15249
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4878511577922188e-05,
+      "loss": 0.4231,
+      "step": 15250
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.487817574261582e-05,
+      "loss": 0.4117,
+      "step": 15251
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4877839907309457e-05,
+      "loss": 0.5697,
+      "step": 15252
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.487750407200309e-05,
+      "loss": 0.1565,
+      "step": 15253
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4877168236696724e-05,
+      "loss": 0.4014,
+      "step": 15254
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4876832401390359e-05,
+      "loss": 0.3346,
+      "step": 15255
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4876496566083994e-05,
+      "loss": 0.3176,
+      "step": 15256
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4876160730777627e-05,
+      "loss": 0.3183,
+      "step": 15257
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4875824895471263e-05,
+      "loss": 0.2452,
+      "step": 15258
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4875489060164896e-05,
+      "loss": 0.2293,
+      "step": 15259
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4875153224858529e-05,
+      "loss": 0.2198,
+      "step": 15260
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4874817389552165e-05,
+      "loss": 0.3558,
+      "step": 15261
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4874481554245798e-05,
+      "loss": 0.3572,
+      "step": 15262
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4874145718939433e-05,
+      "loss": 0.2852,
+      "step": 15263
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4873809883633068e-05,
+      "loss": 0.3043,
+      "step": 15264
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4873474048326702e-05,
+      "loss": 0.3723,
+      "step": 15265
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4873138213020335e-05,
+      "loss": 0.2867,
+      "step": 15266
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4872802377713971e-05,
+      "loss": 0.391,
+      "step": 15267
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4872466542407604e-05,
+      "loss": 0.0961,
+      "step": 15268
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4872130707101237e-05,
+      "loss": 0.2392,
+      "step": 15269
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4871794871794874e-05,
+      "loss": 0.4905,
+      "step": 15270
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4871459036488507e-05,
+      "loss": 0.2595,
+      "step": 15271
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4871123201182141e-05,
+      "loss": 0.3517,
+      "step": 15272
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4870787365875776e-05,
+      "loss": 0.3556,
+      "step": 15273
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4870451530569409e-05,
+      "loss": 0.335,
+      "step": 15274
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4870115695263044e-05,
+      "loss": 0.3112,
+      "step": 15275
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4869779859956678e-05,
+      "loss": 0.3494,
+      "step": 15276
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4869444024650313e-05,
+      "loss": 0.1725,
+      "step": 15277
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4869108189343946e-05,
+      "loss": 0.1211,
+      "step": 15278
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4868772354037582e-05,
+      "loss": 0.1501,
+      "step": 15279
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4868436518731215e-05,
+      "loss": 0.2211,
+      "step": 15280
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4868100683424848e-05,
+      "loss": 0.1745,
+      "step": 15281
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4867764848118484e-05,
+      "loss": 0.4926,
+      "step": 15282
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4867429012812117e-05,
+      "loss": 0.1864,
+      "step": 15283
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4867093177505752e-05,
+      "loss": 0.3174,
+      "step": 15284
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4866757342199387e-05,
+      "loss": 0.3045,
+      "step": 15285
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4866421506893021e-05,
+      "loss": 0.2713,
+      "step": 15286
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4866085671586654e-05,
+      "loss": 0.2842,
+      "step": 15287
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.486574983628029e-05,
+      "loss": 0.4206,
+      "step": 15288
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4865414000973924e-05,
+      "loss": 0.3001,
+      "step": 15289
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4865078165667556e-05,
+      "loss": 0.2595,
+      "step": 15290
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4864742330361193e-05,
+      "loss": 0.2359,
+      "step": 15291
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4864406495054826e-05,
+      "loss": 0.3551,
+      "step": 15292
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.486407065974846e-05,
+      "loss": 0.3265,
+      "step": 15293
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4863734824442095e-05,
+      "loss": 0.3786,
+      "step": 15294
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4863398989135728e-05,
+      "loss": 0.1141,
+      "step": 15295
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4863063153829363e-05,
+      "loss": 0.2571,
+      "step": 15296
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4862727318522997e-05,
+      "loss": 0.4984,
+      "step": 15297
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4862391483216632e-05,
+      "loss": 0.3271,
+      "step": 15298
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4862055647910265e-05,
+      "loss": 0.2883,
+      "step": 15299
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4861719812603901e-05,
+      "loss": 0.3598,
+      "step": 15300
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4861383977297534e-05,
+      "loss": 0.1491,
+      "step": 15301
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4861048141991167e-05,
+      "loss": 0.3476,
+      "step": 15302
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4860712306684804e-05,
+      "loss": 0.3089,
+      "step": 15303
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4860376471378436e-05,
+      "loss": 0.2399,
+      "step": 15304
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4860040636072071e-05,
+      "loss": 0.4993,
+      "step": 15305
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4859704800765706e-05,
+      "loss": 0.2406,
+      "step": 15306
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.485936896545934e-05,
+      "loss": 0.4584,
+      "step": 15307
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4859033130152973e-05,
+      "loss": 0.2685,
+      "step": 15308
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.485869729484661e-05,
+      "loss": 0.2847,
+      "step": 15309
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4858361459540243e-05,
+      "loss": 0.2581,
+      "step": 15310
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4858025624233876e-05,
+      "loss": 0.3218,
+      "step": 15311
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4857689788927512e-05,
+      "loss": 0.2466,
+      "step": 15312
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4857353953621145e-05,
+      "loss": 0.3393,
+      "step": 15313
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.485701811831478e-05,
+      "loss": 0.4158,
+      "step": 15314
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4856682283008414e-05,
+      "loss": 0.3411,
+      "step": 15315
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4856346447702049e-05,
+      "loss": 0.3104,
+      "step": 15316
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4856010612395682e-05,
+      "loss": 0.2952,
+      "step": 15317
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4855674777089316e-05,
+      "loss": 0.4385,
+      "step": 15318
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4855338941782951e-05,
+      "loss": 0.3549,
+      "step": 15319
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4855003106476584e-05,
+      "loss": 0.3274,
+      "step": 15320
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.485466727117022e-05,
+      "loss": 0.348,
+      "step": 15321
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4854331435863853e-05,
+      "loss": 0.2957,
+      "step": 15322
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4853995600557486e-05,
+      "loss": 0.3538,
+      "step": 15323
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4853659765251123e-05,
+      "loss": 0.3158,
+      "step": 15324
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4853323929944756e-05,
+      "loss": 0.2724,
+      "step": 15325
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.485298809463839e-05,
+      "loss": 0.3261,
+      "step": 15326
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4852652259332025e-05,
+      "loss": 0.3277,
+      "step": 15327
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.485231642402566e-05,
+      "loss": 0.4839,
+      "step": 15328
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4851980588719292e-05,
+      "loss": 0.3231,
+      "step": 15329
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4851644753412929e-05,
+      "loss": 0.3815,
+      "step": 15330
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4851308918106562e-05,
+      "loss": 0.2369,
+      "step": 15331
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4850973082800195e-05,
+      "loss": 0.2139,
+      "step": 15332
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4850637247493831e-05,
+      "loss": 0.2475,
+      "step": 15333
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4850301412187464e-05,
+      "loss": 0.4442,
+      "step": 15334
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4849965576881099e-05,
+      "loss": 0.2458,
+      "step": 15335
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4849629741574733e-05,
+      "loss": 0.3284,
+      "step": 15336
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4849293906268368e-05,
+      "loss": 0.4999,
+      "step": 15337
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4848958070962001e-05,
+      "loss": 0.1654,
+      "step": 15338
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4848622235655637e-05,
+      "loss": 0.2169,
+      "step": 15339
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.484828640034927e-05,
+      "loss": 0.2319,
+      "step": 15340
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4847950565042903e-05,
+      "loss": 0.3154,
+      "step": 15341
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.484761472973654e-05,
+      "loss": 0.3949,
+      "step": 15342
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4847278894430172e-05,
+      "loss": 0.2714,
+      "step": 15343
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4846943059123805e-05,
+      "loss": 0.3153,
+      "step": 15344
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4846607223817442e-05,
+      "loss": 0.3631,
+      "step": 15345
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4846271388511075e-05,
+      "loss": 0.4554,
+      "step": 15346
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.484593555320471e-05,
+      "loss": 0.3118,
+      "step": 15347
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4845599717898344e-05,
+      "loss": 0.2436,
+      "step": 15348
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4845263882591979e-05,
+      "loss": 0.2779,
+      "step": 15349
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4844928047285612e-05,
+      "loss": 0.2622,
+      "step": 15350
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4844592211979248e-05,
+      "loss": 0.2576,
+      "step": 15351
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4844256376672881e-05,
+      "loss": 0.4112,
+      "step": 15352
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4843920541366514e-05,
+      "loss": 0.3599,
+      "step": 15353
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.484358470606015e-05,
+      "loss": 0.2502,
+      "step": 15354
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4843248870753783e-05,
+      "loss": 0.4063,
+      "step": 15355
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4842913035447418e-05,
+      "loss": 0.3153,
+      "step": 15356
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4842577200141052e-05,
+      "loss": 0.3371,
+      "step": 15357
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4842241364834687e-05,
+      "loss": 0.393,
+      "step": 15358
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.484190552952832e-05,
+      "loss": 0.3072,
+      "step": 15359
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4841569694221956e-05,
+      "loss": 0.3321,
+      "step": 15360
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.484123385891559e-05,
+      "loss": 0.4305,
+      "step": 15361
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4840898023609222e-05,
+      "loss": 0.4017,
+      "step": 15362
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4840562188302859e-05,
+      "loss": 0.2434,
+      "step": 15363
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4840226352996492e-05,
+      "loss": 0.3457,
+      "step": 15364
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4839890517690125e-05,
+      "loss": 0.4461,
+      "step": 15365
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4839554682383761e-05,
+      "loss": 0.3081,
+      "step": 15366
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4839218847077394e-05,
+      "loss": 0.2871,
+      "step": 15367
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4838883011771028e-05,
+      "loss": 0.257,
+      "step": 15368
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4838547176464663e-05,
+      "loss": 0.3561,
+      "step": 15369
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4838211341158298e-05,
+      "loss": 0.331,
+      "step": 15370
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.483787550585193e-05,
+      "loss": 0.4031,
+      "step": 15371
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4837539670545567e-05,
+      "loss": 0.487,
+      "step": 15372
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.48372038352392e-05,
+      "loss": 0.284,
+      "step": 15373
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4836867999932833e-05,
+      "loss": 0.3333,
+      "step": 15374
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.483653216462647e-05,
+      "loss": 0.5412,
+      "step": 15375
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4836196329320102e-05,
+      "loss": 0.3401,
+      "step": 15376
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4835860494013737e-05,
+      "loss": 0.3972,
+      "step": 15377
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4835524658707372e-05,
+      "loss": 0.4766,
+      "step": 15378
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4835188823401006e-05,
+      "loss": 0.3147,
+      "step": 15379
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.483485298809464e-05,
+      "loss": 0.2823,
+      "step": 15380
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4834517152788276e-05,
+      "loss": 0.3462,
+      "step": 15381
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4834181317481908e-05,
+      "loss": 0.4693,
+      "step": 15382
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4833845482175541e-05,
+      "loss": 0.3313,
+      "step": 15383
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.4833509646869178e-05,
+      "loss": 0.1864,
+      "step": 15384
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.483317381156281e-05,
+      "loss": 0.2898,
+      "step": 15385
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4832837976256444e-05,
+      "loss": 0.3342,
+      "step": 15386
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.483250214095008e-05,
+      "loss": 0.3134,
+      "step": 15387
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4832166305643713e-05,
+      "loss": 0.2223,
+      "step": 15388
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4831830470337348e-05,
+      "loss": 0.3173,
+      "step": 15389
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4831494635030982e-05,
+      "loss": 0.3558,
+      "step": 15390
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4831158799724617e-05,
+      "loss": 0.2938,
+      "step": 15391
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.483082296441825e-05,
+      "loss": 0.2941,
+      "step": 15392
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4830487129111886e-05,
+      "loss": 0.2789,
+      "step": 15393
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.483015129380552e-05,
+      "loss": 0.3961,
+      "step": 15394
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4829815458499152e-05,
+      "loss": 0.247,
+      "step": 15395
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4829479623192788e-05,
+      "loss": 0.2107,
+      "step": 15396
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4829143787886421e-05,
+      "loss": 0.2854,
+      "step": 15397
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4828807952580056e-05,
+      "loss": 0.3358,
+      "step": 15398
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.482847211727369e-05,
+      "loss": 0.3176,
+      "step": 15399
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4828136281967325e-05,
+      "loss": 0.3319,
+      "step": 15400
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4827800446660958e-05,
+      "loss": 0.3832,
+      "step": 15401
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4827464611354595e-05,
+      "loss": 0.2572,
+      "step": 15402
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4827128776048228e-05,
+      "loss": 0.2293,
+      "step": 15403
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.482679294074186e-05,
+      "loss": 0.3118,
+      "step": 15404
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4826457105435497e-05,
+      "loss": 0.3182,
+      "step": 15405
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.482612127012913e-05,
+      "loss": 0.3671,
+      "step": 15406
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4825785434822764e-05,
+      "loss": 0.2086,
+      "step": 15407
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4825449599516399e-05,
+      "loss": 0.2836,
+      "step": 15408
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4825113764210032e-05,
+      "loss": 0.3758,
+      "step": 15409
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4824777928903667e-05,
+      "loss": 0.3345,
+      "step": 15410
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4824442093597301e-05,
+      "loss": 0.2676,
+      "step": 15411
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4824106258290936e-05,
+      "loss": 0.2356,
+      "step": 15412
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4823770422984569e-05,
+      "loss": 0.2269,
+      "step": 15413
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4823434587678205e-05,
+      "loss": 0.5122,
+      "step": 15414
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4823098752371838e-05,
+      "loss": 0.2501,
+      "step": 15415
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4822762917065471e-05,
+      "loss": 0.4491,
+      "step": 15416
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4822427081759108e-05,
+      "loss": 0.2616,
+      "step": 15417
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.482209124645274e-05,
+      "loss": 0.2847,
+      "step": 15418
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4821755411146375e-05,
+      "loss": 0.3497,
+      "step": 15419
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.482141957584001e-05,
+      "loss": 0.3282,
+      "step": 15420
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4821083740533644e-05,
+      "loss": 0.4389,
+      "step": 15421
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4820747905227277e-05,
+      "loss": 0.2421,
+      "step": 15422
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4820412069920914e-05,
+      "loss": 0.1839,
+      "step": 15423
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4820076234614547e-05,
+      "loss": 0.3805,
+      "step": 15424
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.481974039930818e-05,
+      "loss": 0.2287,
+      "step": 15425
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4819404564001816e-05,
+      "loss": 0.3172,
+      "step": 15426
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4819068728695449e-05,
+      "loss": 0.2896,
+      "step": 15427
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4818732893389084e-05,
+      "loss": 0.2939,
+      "step": 15428
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4818397058082718e-05,
+      "loss": 0.2892,
+      "step": 15429
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4818061222776351e-05,
+      "loss": 0.2582,
+      "step": 15430
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4817725387469986e-05,
+      "loss": 0.3176,
+      "step": 15431
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4817389552163619e-05,
+      "loss": 0.2643,
+      "step": 15432
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4817053716857255e-05,
+      "loss": 0.3719,
+      "step": 15433
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4816717881550888e-05,
+      "loss": 0.3736,
+      "step": 15434
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4816382046244521e-05,
+      "loss": 0.2459,
+      "step": 15435
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4816046210938157e-05,
+      "loss": 0.2839,
+      "step": 15436
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.481571037563179e-05,
+      "loss": 0.3244,
+      "step": 15437
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4815374540325425e-05,
+      "loss": 0.3135,
+      "step": 15438
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.481503870501906e-05,
+      "loss": 0.264,
+      "step": 15439
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4814702869712694e-05,
+      "loss": 0.2136,
+      "step": 15440
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4814367034406327e-05,
+      "loss": 0.2796,
+      "step": 15441
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4814031199099964e-05,
+      "loss": 0.2671,
+      "step": 15442
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4813695363793597e-05,
+      "loss": 0.3535,
+      "step": 15443
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.481335952848723e-05,
+      "loss": 0.1921,
+      "step": 15444
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4813023693180866e-05,
+      "loss": 0.3426,
+      "step": 15445
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4812687857874499e-05,
+      "loss": 0.3677,
+      "step": 15446
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4812352022568133e-05,
+      "loss": 0.2381,
+      "step": 15447
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4812016187261768e-05,
+      "loss": 0.3766,
+      "step": 15448
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4811680351955403e-05,
+      "loss": 0.3174,
+      "step": 15449
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4811344516649036e-05,
+      "loss": 0.2825,
+      "step": 15450
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4811008681342672e-05,
+      "loss": 0.2982,
+      "step": 15451
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4810672846036305e-05,
+      "loss": 0.1897,
+      "step": 15452
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4810337010729938e-05,
+      "loss": 0.3369,
+      "step": 15453
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4810001175423574e-05,
+      "loss": 0.3274,
+      "step": 15454
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4809665340117207e-05,
+      "loss": 0.3132,
+      "step": 15455
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.480932950481084e-05,
+      "loss": 0.234,
+      "step": 15456
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4808993669504477e-05,
+      "loss": 0.2617,
+      "step": 15457
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.480865783419811e-05,
+      "loss": 0.2187,
+      "step": 15458
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4808321998891744e-05,
+      "loss": 0.3013,
+      "step": 15459
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4807986163585379e-05,
+      "loss": 0.2334,
+      "step": 15460
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4807650328279013e-05,
+      "loss": 0.3197,
+      "step": 15461
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4807314492972646e-05,
+      "loss": 0.3742,
+      "step": 15462
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4806978657666283e-05,
+      "loss": 0.2743,
+      "step": 15463
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4806642822359916e-05,
+      "loss": 0.3208,
+      "step": 15464
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4806306987053549e-05,
+      "loss": 0.5102,
+      "step": 15465
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4805971151747185e-05,
+      "loss": 0.1809,
+      "step": 15466
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4805635316440818e-05,
+      "loss": 0.3213,
+      "step": 15467
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4805299481134453e-05,
+      "loss": 0.2839,
+      "step": 15468
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4804963645828087e-05,
+      "loss": 0.2454,
+      "step": 15469
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4804627810521722e-05,
+      "loss": 0.2894,
+      "step": 15470
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4804291975215355e-05,
+      "loss": 0.2159,
+      "step": 15471
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4803956139908991e-05,
+      "loss": 0.2661,
+      "step": 15472
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4803620304602624e-05,
+      "loss": 0.3009,
+      "step": 15473
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4803284469296257e-05,
+      "loss": 0.2612,
+      "step": 15474
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4802948633989893e-05,
+      "loss": 0.2214,
+      "step": 15475
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4802612798683526e-05,
+      "loss": 0.2774,
+      "step": 15476
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.480227696337716e-05,
+      "loss": 0.1878,
+      "step": 15477
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4801941128070796e-05,
+      "loss": 0.2632,
+      "step": 15478
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4801605292764429e-05,
+      "loss": 0.3411,
+      "step": 15479
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4801269457458063e-05,
+      "loss": 0.2951,
+      "step": 15480
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4800933622151698e-05,
+      "loss": 0.2984,
+      "step": 15481
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4800597786845333e-05,
+      "loss": 0.3778,
+      "step": 15482
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4800261951538966e-05,
+      "loss": 0.3269,
+      "step": 15483
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4799926116232602e-05,
+      "loss": 0.2879,
+      "step": 15484
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4799590280926235e-05,
+      "loss": 0.6542,
+      "step": 15485
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4799254445619868e-05,
+      "loss": 0.4008,
+      "step": 15486
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4798918610313504e-05,
+      "loss": 0.2844,
+      "step": 15487
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4798582775007137e-05,
+      "loss": 0.3633,
+      "step": 15488
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4798246939700772e-05,
+      "loss": 0.3117,
+      "step": 15489
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4797911104394406e-05,
+      "loss": 0.5447,
+      "step": 15490
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4797575269088041e-05,
+      "loss": 0.2285,
+      "step": 15491
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4797239433781674e-05,
+      "loss": 0.2963,
+      "step": 15492
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.479690359847531e-05,
+      "loss": 0.432,
+      "step": 15493
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4796567763168943e-05,
+      "loss": 0.3448,
+      "step": 15494
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4796231927862576e-05,
+      "loss": 0.4798,
+      "step": 15495
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4795896092556213e-05,
+      "loss": 0.2635,
+      "step": 15496
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4795560257249845e-05,
+      "loss": 0.3308,
+      "step": 15497
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.479522442194348e-05,
+      "loss": 0.2296,
+      "step": 15498
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4794888586637115e-05,
+      "loss": 0.2002,
+      "step": 15499
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4794552751330748e-05,
+      "loss": 0.3672,
+      "step": 15500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4794216916024382e-05,
+      "loss": 0.3258,
+      "step": 15501
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4793881080718017e-05,
+      "loss": 0.2913,
+      "step": 15502
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4793545245411652e-05,
+      "loss": 0.3289,
+      "step": 15503
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4793209410105285e-05,
+      "loss": 0.4116,
+      "step": 15504
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4792873574798921e-05,
+      "loss": 0.3019,
+      "step": 15505
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4792537739492554e-05,
+      "loss": 0.3204,
+      "step": 15506
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4792201904186187e-05,
+      "loss": 0.347,
+      "step": 15507
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4791866068879823e-05,
+      "loss": 0.22,
+      "step": 15508
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4791530233573456e-05,
+      "loss": 0.3618,
+      "step": 15509
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.479119439826709e-05,
+      "loss": 0.2163,
+      "step": 15510
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4790858562960725e-05,
+      "loss": 0.2943,
+      "step": 15511
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.479052272765436e-05,
+      "loss": 0.3709,
+      "step": 15512
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4790186892347993e-05,
+      "loss": 0.4418,
+      "step": 15513
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.478985105704163e-05,
+      "loss": 0.2725,
+      "step": 15514
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4789515221735262e-05,
+      "loss": 0.5854,
+      "step": 15515
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4789179386428895e-05,
+      "loss": 0.3682,
+      "step": 15516
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4788843551122532e-05,
+      "loss": 0.4047,
+      "step": 15517
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4788507715816165e-05,
+      "loss": 0.2109,
+      "step": 15518
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.47881718805098e-05,
+      "loss": 0.3889,
+      "step": 15519
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4787836045203434e-05,
+      "loss": 0.2758,
+      "step": 15520
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4787500209897067e-05,
+      "loss": 0.4071,
+      "step": 15521
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4787164374590702e-05,
+      "loss": 0.2274,
+      "step": 15522
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4786828539284336e-05,
+      "loss": 0.4474,
+      "step": 15523
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.478649270397797e-05,
+      "loss": 0.2684,
+      "step": 15524
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4786156868671604e-05,
+      "loss": 0.2839,
+      "step": 15525
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.478582103336524e-05,
+      "loss": 0.2148,
+      "step": 15526
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4785485198058873e-05,
+      "loss": 0.3322,
+      "step": 15527
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4785149362752506e-05,
+      "loss": 0.4276,
+      "step": 15528
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4784813527446142e-05,
+      "loss": 0.3204,
+      "step": 15529
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4784477692139775e-05,
+      "loss": 0.222,
+      "step": 15530
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.478414185683341e-05,
+      "loss": 0.2982,
+      "step": 15531
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4783806021527045e-05,
+      "loss": 0.1338,
+      "step": 15532
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.478347018622068e-05,
+      "loss": 0.3093,
+      "step": 15533
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4783134350914312e-05,
+      "loss": 0.256,
+      "step": 15534
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4782798515607949e-05,
+      "loss": 0.4785,
+      "step": 15535
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4782462680301581e-05,
+      "loss": 0.2842,
+      "step": 15536
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4782126844995214e-05,
+      "loss": 0.2868,
+      "step": 15537
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.478179100968885e-05,
+      "loss": 0.2228,
+      "step": 15538
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4781455174382484e-05,
+      "loss": 0.2812,
+      "step": 15539
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4781119339076118e-05,
+      "loss": 0.4668,
+      "step": 15540
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4780783503769753e-05,
+      "loss": 0.3014,
+      "step": 15541
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4780447668463388e-05,
+      "loss": 0.2339,
+      "step": 15542
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.478011183315702e-05,
+      "loss": 0.3601,
+      "step": 15543
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4779775997850655e-05,
+      "loss": 0.3345,
+      "step": 15544
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.477944016254429e-05,
+      "loss": 0.3847,
+      "step": 15545
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4779104327237923e-05,
+      "loss": 0.4085,
+      "step": 15546
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.477876849193156e-05,
+      "loss": 0.2824,
+      "step": 15547
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4778432656625192e-05,
+      "loss": 0.5576,
+      "step": 15548
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4778096821318825e-05,
+      "loss": 0.4495,
+      "step": 15549
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4777760986012461e-05,
+      "loss": 0.3869,
+      "step": 15550
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4777425150706094e-05,
+      "loss": 0.3504,
+      "step": 15551
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4777089315399729e-05,
+      "loss": 0.2591,
+      "step": 15552
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4776753480093364e-05,
+      "loss": 0.3147,
+      "step": 15553
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4776417644786998e-05,
+      "loss": 0.2522,
+      "step": 15554
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4776081809480631e-05,
+      "loss": 0.3518,
+      "step": 15555
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4775745974174268e-05,
+      "loss": 0.4288,
+      "step": 15556
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.47754101388679e-05,
+      "loss": 0.243,
+      "step": 15557
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4775074303561534e-05,
+      "loss": 0.2902,
+      "step": 15558
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.477473846825517e-05,
+      "loss": 0.3773,
+      "step": 15559
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4774402632948803e-05,
+      "loss": 0.2445,
+      "step": 15560
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4774066797642438e-05,
+      "loss": 0.4742,
+      "step": 15561
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4773730962336072e-05,
+      "loss": 0.335,
+      "step": 15562
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4773395127029707e-05,
+      "loss": 0.2834,
+      "step": 15563
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.477305929172334e-05,
+      "loss": 0.3259,
+      "step": 15564
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4772723456416974e-05,
+      "loss": 0.3809,
+      "step": 15565
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4772387621110609e-05,
+      "loss": 0.3977,
+      "step": 15566
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4772051785804242e-05,
+      "loss": 0.364,
+      "step": 15567
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4771715950497878e-05,
+      "loss": 0.3587,
+      "step": 15568
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4771380115191511e-05,
+      "loss": 0.3096,
+      "step": 15569
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4771044279885144e-05,
+      "loss": 0.3414,
+      "step": 15570
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.477070844457878e-05,
+      "loss": 0.2674,
+      "step": 15571
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4770372609272414e-05,
+      "loss": 0.4206,
+      "step": 15572
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4770036773966048e-05,
+      "loss": 0.3246,
+      "step": 15573
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4769700938659683e-05,
+      "loss": 0.4723,
+      "step": 15574
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4769365103353317e-05,
+      "loss": 0.2373,
+      "step": 15575
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.476902926804695e-05,
+      "loss": 0.4258,
+      "step": 15576
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4768693432740587e-05,
+      "loss": 0.3328,
+      "step": 15577
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.476835759743422e-05,
+      "loss": 0.3253,
+      "step": 15578
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4768021762127853e-05,
+      "loss": 0.4788,
+      "step": 15579
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4767685926821489e-05,
+      "loss": 0.2228,
+      "step": 15580
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4767350091515122e-05,
+      "loss": 0.3611,
+      "step": 15581
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4767014256208757e-05,
+      "loss": 0.2624,
+      "step": 15582
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.4766678420902391e-05,
+      "loss": 0.3767,
+      "step": 15583
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4766342585596026e-05,
+      "loss": 0.2994,
+      "step": 15584
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4766006750289659e-05,
+      "loss": 0.3502,
+      "step": 15585
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4765670914983295e-05,
+      "loss": 0.4581,
+      "step": 15586
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4765335079676928e-05,
+      "loss": 0.2963,
+      "step": 15587
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4764999244370561e-05,
+      "loss": 0.2182,
+      "step": 15588
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4764663409064197e-05,
+      "loss": 0.4279,
+      "step": 15589
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.476432757375783e-05,
+      "loss": 0.2049,
+      "step": 15590
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4763991738451463e-05,
+      "loss": 0.5309,
+      "step": 15591
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.47636559031451e-05,
+      "loss": 0.3954,
+      "step": 15592
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4763320067838733e-05,
+      "loss": 0.3461,
+      "step": 15593
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4762984232532367e-05,
+      "loss": 0.3112,
+      "step": 15594
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4762648397226002e-05,
+      "loss": 0.3418,
+      "step": 15595
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4762312561919637e-05,
+      "loss": 0.3805,
+      "step": 15596
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.476197672661327e-05,
+      "loss": 0.2785,
+      "step": 15597
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4761640891306906e-05,
+      "loss": 0.3128,
+      "step": 15598
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4761305056000539e-05,
+      "loss": 0.397,
+      "step": 15599
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4760969220694172e-05,
+      "loss": 0.363,
+      "step": 15600
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4760633385387808e-05,
+      "loss": 0.2366,
+      "step": 15601
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4760297550081441e-05,
+      "loss": 0.4053,
+      "step": 15602
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4759961714775076e-05,
+      "loss": 0.3901,
+      "step": 15603
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.475962587946871e-05,
+      "loss": 0.2255,
+      "step": 15604
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4759290044162345e-05,
+      "loss": 0.1525,
+      "step": 15605
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4758954208855978e-05,
+      "loss": 0.3933,
+      "step": 15606
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4758618373549614e-05,
+      "loss": 0.2106,
+      "step": 15607
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4758282538243247e-05,
+      "loss": 0.2495,
+      "step": 15608
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.475794670293688e-05,
+      "loss": 0.3467,
+      "step": 15609
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4757610867630517e-05,
+      "loss": 0.175,
+      "step": 15610
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.475727503232415e-05,
+      "loss": 0.3391,
+      "step": 15611
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4756939197017783e-05,
+      "loss": 0.546,
+      "step": 15612
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4756603361711417e-05,
+      "loss": 0.1684,
+      "step": 15613
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4756267526405052e-05,
+      "loss": 0.4077,
+      "step": 15614
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4755931691098686e-05,
+      "loss": 0.1388,
+      "step": 15615
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.475559585579232e-05,
+      "loss": 0.4309,
+      "step": 15616
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4755260020485956e-05,
+      "loss": 0.3131,
+      "step": 15617
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4754924185179589e-05,
+      "loss": 0.5844,
+      "step": 15618
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4754588349873222e-05,
+      "loss": 0.3909,
+      "step": 15619
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4754252514566858e-05,
+      "loss": 0.3849,
+      "step": 15620
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4753916679260491e-05,
+      "loss": 0.3655,
+      "step": 15621
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4753580843954126e-05,
+      "loss": 0.3165,
+      "step": 15622
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.475324500864776e-05,
+      "loss": 0.4965,
+      "step": 15623
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4752909173341395e-05,
+      "loss": 0.3899,
+      "step": 15624
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4752573338035028e-05,
+      "loss": 0.1546,
+      "step": 15625
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4752237502728664e-05,
+      "loss": 0.152,
+      "step": 15626
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4751901667422297e-05,
+      "loss": 0.3626,
+      "step": 15627
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.475156583211593e-05,
+      "loss": 0.2338,
+      "step": 15628
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4751229996809566e-05,
+      "loss": 0.3697,
+      "step": 15629
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.47508941615032e-05,
+      "loss": 0.4054,
+      "step": 15630
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4750558326196834e-05,
+      "loss": 0.2958,
+      "step": 15631
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4750222490890469e-05,
+      "loss": 0.2475,
+      "step": 15632
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4749886655584103e-05,
+      "loss": 0.3755,
+      "step": 15633
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4749550820277736e-05,
+      "loss": 0.3203,
+      "step": 15634
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4749214984971371e-05,
+      "loss": 0.2745,
+      "step": 15635
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4748879149665006e-05,
+      "loss": 0.3651,
+      "step": 15636
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4748543314358639e-05,
+      "loss": 0.2952,
+      "step": 15637
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4748207479052275e-05,
+      "loss": 0.4994,
+      "step": 15638
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4747871643745908e-05,
+      "loss": 0.1348,
+      "step": 15639
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.474753580843954e-05,
+      "loss": 0.3945,
+      "step": 15640
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4747199973133177e-05,
+      "loss": 0.2681,
+      "step": 15641
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.474686413782681e-05,
+      "loss": 0.3678,
+      "step": 15642
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4746528302520445e-05,
+      "loss": 0.4421,
+      "step": 15643
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.474619246721408e-05,
+      "loss": 0.4247,
+      "step": 15644
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4745856631907714e-05,
+      "loss": 0.3309,
+      "step": 15645
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4745520796601347e-05,
+      "loss": 0.2193,
+      "step": 15646
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4745184961294983e-05,
+      "loss": 0.3289,
+      "step": 15647
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4744849125988616e-05,
+      "loss": 0.2852,
+      "step": 15648
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.474451329068225e-05,
+      "loss": 0.3807,
+      "step": 15649
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4744177455375886e-05,
+      "loss": 0.289,
+      "step": 15650
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4743841620069519e-05,
+      "loss": 0.2238,
+      "step": 15651
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4743505784763153e-05,
+      "loss": 0.3386,
+      "step": 15652
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4743169949456788e-05,
+      "loss": 0.1965,
+      "step": 15653
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4742834114150422e-05,
+      "loss": 0.3872,
+      "step": 15654
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4742498278844055e-05,
+      "loss": 0.3383,
+      "step": 15655
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.474216244353769e-05,
+      "loss": 0.294,
+      "step": 15656
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4741826608231325e-05,
+      "loss": 0.2951,
+      "step": 15657
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4741490772924958e-05,
+      "loss": 0.3214,
+      "step": 15658
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4741154937618594e-05,
+      "loss": 0.3266,
+      "step": 15659
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4740819102312227e-05,
+      "loss": 0.2486,
+      "step": 15660
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.474048326700586e-05,
+      "loss": 0.3347,
+      "step": 15661
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4740147431699496e-05,
+      "loss": 0.2479,
+      "step": 15662
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.473981159639313e-05,
+      "loss": 0.4022,
+      "step": 15663
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4739475761086764e-05,
+      "loss": 0.2898,
+      "step": 15664
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4739139925780398e-05,
+      "loss": 0.2024,
+      "step": 15665
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4738804090474033e-05,
+      "loss": 0.2247,
+      "step": 15666
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4738468255167666e-05,
+      "loss": 0.19,
+      "step": 15667
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4738132419861302e-05,
+      "loss": 0.3674,
+      "step": 15668
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4737796584554935e-05,
+      "loss": 0.2265,
+      "step": 15669
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4737460749248568e-05,
+      "loss": 0.2683,
+      "step": 15670
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4737124913942205e-05,
+      "loss": 0.3033,
+      "step": 15671
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4736789078635838e-05,
+      "loss": 0.3344,
+      "step": 15672
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4736453243329472e-05,
+      "loss": 0.2591,
+      "step": 15673
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4736117408023107e-05,
+      "loss": 0.466,
+      "step": 15674
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4735781572716742e-05,
+      "loss": 0.2856,
+      "step": 15675
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4735445737410375e-05,
+      "loss": 0.3125,
+      "step": 15676
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4735109902104011e-05,
+      "loss": 0.3563,
+      "step": 15677
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4734774066797644e-05,
+      "loss": 0.3549,
+      "step": 15678
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4734438231491277e-05,
+      "loss": 0.3274,
+      "step": 15679
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4734102396184913e-05,
+      "loss": 0.3142,
+      "step": 15680
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4733766560878546e-05,
+      "loss": 0.3498,
+      "step": 15681
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4733430725572179e-05,
+      "loss": 0.3978,
+      "step": 15682
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4733094890265815e-05,
+      "loss": 0.2723,
+      "step": 15683
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4732759054959448e-05,
+      "loss": 0.3371,
+      "step": 15684
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4732423219653083e-05,
+      "loss": 0.2976,
+      "step": 15685
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4732087384346718e-05,
+      "loss": 0.2862,
+      "step": 15686
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4731751549040352e-05,
+      "loss": 0.4945,
+      "step": 15687
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4731415713733985e-05,
+      "loss": 0.1943,
+      "step": 15688
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4731079878427622e-05,
+      "loss": 0.4208,
+      "step": 15689
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4730744043121255e-05,
+      "loss": 0.1817,
+      "step": 15690
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4730408207814887e-05,
+      "loss": 0.2876,
+      "step": 15691
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4730072372508524e-05,
+      "loss": 0.584,
+      "step": 15692
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4729736537202157e-05,
+      "loss": 0.1874,
+      "step": 15693
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4729400701895791e-05,
+      "loss": 0.3457,
+      "step": 15694
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4729064866589426e-05,
+      "loss": 0.3381,
+      "step": 15695
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.472872903128306e-05,
+      "loss": 0.3954,
+      "step": 15696
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4728393195976694e-05,
+      "loss": 0.4796,
+      "step": 15697
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.472805736067033e-05,
+      "loss": 0.2811,
+      "step": 15698
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4727721525363963e-05,
+      "loss": 0.248,
+      "step": 15699
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4727385690057596e-05,
+      "loss": 0.2666,
+      "step": 15700
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4727049854751232e-05,
+      "loss": 0.2494,
+      "step": 15701
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4726714019444865e-05,
+      "loss": 0.3412,
+      "step": 15702
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4726378184138498e-05,
+      "loss": 0.3294,
+      "step": 15703
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4726042348832134e-05,
+      "loss": 0.3827,
+      "step": 15704
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4725706513525767e-05,
+      "loss": 0.3525,
+      "step": 15705
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4725370678219402e-05,
+      "loss": 0.4116,
+      "step": 15706
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4725034842913037e-05,
+      "loss": 0.3472,
+      "step": 15707
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4724699007606671e-05,
+      "loss": 0.3714,
+      "step": 15708
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4724363172300304e-05,
+      "loss": 0.3684,
+      "step": 15709
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.472402733699394e-05,
+      "loss": 0.432,
+      "step": 15710
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4723691501687574e-05,
+      "loss": 0.3487,
+      "step": 15711
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4723355666381207e-05,
+      "loss": 0.3164,
+      "step": 15712
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4723019831074843e-05,
+      "loss": 0.4055,
+      "step": 15713
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4722683995768476e-05,
+      "loss": 0.343,
+      "step": 15714
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.472234816046211e-05,
+      "loss": 0.3757,
+      "step": 15715
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4722012325155745e-05,
+      "loss": 0.4524,
+      "step": 15716
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.472167648984938e-05,
+      "loss": 0.2423,
+      "step": 15717
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4721340654543013e-05,
+      "loss": 0.3062,
+      "step": 15718
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4721004819236649e-05,
+      "loss": 0.3896,
+      "step": 15719
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4720668983930282e-05,
+      "loss": 0.3451,
+      "step": 15720
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4720333148623915e-05,
+      "loss": 0.2384,
+      "step": 15721
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4719997313317551e-05,
+      "loss": 0.2571,
+      "step": 15722
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4719661478011184e-05,
+      "loss": 0.2291,
+      "step": 15723
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4719325642704817e-05,
+      "loss": 0.2214,
+      "step": 15724
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4718989807398454e-05,
+      "loss": 0.3123,
+      "step": 15725
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4718653972092087e-05,
+      "loss": 0.2267,
+      "step": 15726
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4718318136785721e-05,
+      "loss": 0.2875,
+      "step": 15727
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4717982301479356e-05,
+      "loss": 0.2812,
+      "step": 15728
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.471764646617299e-05,
+      "loss": 0.4436,
+      "step": 15729
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4717310630866623e-05,
+      "loss": 0.4998,
+      "step": 15730
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.471697479556026e-05,
+      "loss": 0.3707,
+      "step": 15731
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4716638960253893e-05,
+      "loss": 0.2927,
+      "step": 15732
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4716303124947526e-05,
+      "loss": 0.201,
+      "step": 15733
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4715967289641162e-05,
+      "loss": 0.3627,
+      "step": 15734
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4715631454334795e-05,
+      "loss": 0.2099,
+      "step": 15735
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.471529561902843e-05,
+      "loss": 0.1446,
+      "step": 15736
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4714959783722064e-05,
+      "loss": 0.2452,
+      "step": 15737
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4714623948415699e-05,
+      "loss": 0.1987,
+      "step": 15738
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4714288113109332e-05,
+      "loss": 0.3527,
+      "step": 15739
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4713952277802968e-05,
+      "loss": 0.2745,
+      "step": 15740
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4713616442496601e-05,
+      "loss": 0.2653,
+      "step": 15741
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4713280607190234e-05,
+      "loss": 0.3216,
+      "step": 15742
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.471294477188387e-05,
+      "loss": 0.4025,
+      "step": 15743
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4712608936577503e-05,
+      "loss": 0.303,
+      "step": 15744
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4712273101271138e-05,
+      "loss": 0.3275,
+      "step": 15745
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4711937265964773e-05,
+      "loss": 0.3302,
+      "step": 15746
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4711601430658406e-05,
+      "loss": 0.3057,
+      "step": 15747
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.471126559535204e-05,
+      "loss": 0.3246,
+      "step": 15748
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4710929760045675e-05,
+      "loss": 0.2975,
+      "step": 15749
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.471059392473931e-05,
+      "loss": 0.2645,
+      "step": 15750
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4710258089432943e-05,
+      "loss": 0.2348,
+      "step": 15751
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4709922254126579e-05,
+      "loss": 0.3619,
+      "step": 15752
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4709586418820212e-05,
+      "loss": 0.3607,
+      "step": 15753
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4709250583513845e-05,
+      "loss": 0.391,
+      "step": 15754
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4708914748207481e-05,
+      "loss": 0.2385,
+      "step": 15755
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4708578912901114e-05,
+      "loss": 0.2872,
+      "step": 15756
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4708243077594749e-05,
+      "loss": 0.2848,
+      "step": 15757
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4707907242288383e-05,
+      "loss": 0.2891,
+      "step": 15758
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4707571406982018e-05,
+      "loss": 0.3743,
+      "step": 15759
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4707235571675651e-05,
+      "loss": 0.3828,
+      "step": 15760
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4706899736369287e-05,
+      "loss": 0.2615,
+      "step": 15761
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.470656390106292e-05,
+      "loss": 0.5482,
+      "step": 15762
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4706228065756553e-05,
+      "loss": 0.2229,
+      "step": 15763
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.470589223045019e-05,
+      "loss": 0.2451,
+      "step": 15764
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4705556395143823e-05,
+      "loss": 0.2767,
+      "step": 15765
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4705220559837457e-05,
+      "loss": 0.296,
+      "step": 15766
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4704884724531092e-05,
+      "loss": 0.4228,
+      "step": 15767
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4704548889224727e-05,
+      "loss": 0.344,
+      "step": 15768
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.470421305391836e-05,
+      "loss": 0.4689,
+      "step": 15769
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4703877218611994e-05,
+      "loss": 0.3122,
+      "step": 15770
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4703541383305629e-05,
+      "loss": 0.5049,
+      "step": 15771
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4703205547999262e-05,
+      "loss": 0.3446,
+      "step": 15772
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4702869712692898e-05,
+      "loss": 0.3026,
+      "step": 15773
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4702533877386531e-05,
+      "loss": 0.3951,
+      "step": 15774
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4702198042080164e-05,
+      "loss": 0.2599,
+      "step": 15775
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.47018622067738e-05,
+      "loss": 0.3735,
+      "step": 15776
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4701526371467433e-05,
+      "loss": 0.3352,
+      "step": 15777
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4701190536161068e-05,
+      "loss": 0.3544,
+      "step": 15778
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4700854700854703e-05,
+      "loss": 0.3384,
+      "step": 15779
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.4700518865548337e-05,
+      "loss": 0.2999,
+      "step": 15780
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.470018303024197e-05,
+      "loss": 0.3086,
+      "step": 15781
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4699847194935606e-05,
+      "loss": 0.2923,
+      "step": 15782
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.469951135962924e-05,
+      "loss": 0.4758,
+      "step": 15783
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4699175524322872e-05,
+      "loss": 0.2898,
+      "step": 15784
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4698839689016509e-05,
+      "loss": 0.4786,
+      "step": 15785
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4698503853710142e-05,
+      "loss": 0.2201,
+      "step": 15786
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4698168018403776e-05,
+      "loss": 0.2141,
+      "step": 15787
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4697832183097411e-05,
+      "loss": 0.3314,
+      "step": 15788
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4697496347791046e-05,
+      "loss": 0.4156,
+      "step": 15789
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4697160512484679e-05,
+      "loss": 0.4451,
+      "step": 15790
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4696824677178312e-05,
+      "loss": 0.2755,
+      "step": 15791
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4696488841871948e-05,
+      "loss": 0.3515,
+      "step": 15792
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.469615300656558e-05,
+      "loss": 0.177,
+      "step": 15793
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4695817171259214e-05,
+      "loss": 0.1883,
+      "step": 15794
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.469548133595285e-05,
+      "loss": 0.3075,
+      "step": 15795
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4695145500646483e-05,
+      "loss": 0.2364,
+      "step": 15796
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4694809665340118e-05,
+      "loss": 0.1858,
+      "step": 15797
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4694473830033752e-05,
+      "loss": 0.2214,
+      "step": 15798
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4694137994727387e-05,
+      "loss": 0.2358,
+      "step": 15799
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.469380215942102e-05,
+      "loss": 0.2999,
+      "step": 15800
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4693466324114656e-05,
+      "loss": 0.2889,
+      "step": 15801
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.469313048880829e-05,
+      "loss": 0.2547,
+      "step": 15802
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4692794653501922e-05,
+      "loss": 0.555,
+      "step": 15803
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4692458818195559e-05,
+      "loss": 0.4846,
+      "step": 15804
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4692122982889192e-05,
+      "loss": 0.3497,
+      "step": 15805
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4691787147582826e-05,
+      "loss": 0.3061,
+      "step": 15806
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.469145131227646e-05,
+      "loss": 0.2616,
+      "step": 15807
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4691115476970095e-05,
+      "loss": 0.3873,
+      "step": 15808
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4690779641663728e-05,
+      "loss": 0.3981,
+      "step": 15809
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4690443806357365e-05,
+      "loss": 0.2526,
+      "step": 15810
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4690107971050998e-05,
+      "loss": 0.2437,
+      "step": 15811
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.468977213574463e-05,
+      "loss": 0.4075,
+      "step": 15812
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4689436300438267e-05,
+      "loss": 0.1748,
+      "step": 15813
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.46891004651319e-05,
+      "loss": 0.2752,
+      "step": 15814
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4688764629825533e-05,
+      "loss": 0.2218,
+      "step": 15815
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.468842879451917e-05,
+      "loss": 0.376,
+      "step": 15816
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4688092959212802e-05,
+      "loss": 0.3573,
+      "step": 15817
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4687757123906437e-05,
+      "loss": 0.2724,
+      "step": 15818
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4687421288600072e-05,
+      "loss": 0.4372,
+      "step": 15819
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4687085453293706e-05,
+      "loss": 0.3369,
+      "step": 15820
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4686749617987339e-05,
+      "loss": 0.3219,
+      "step": 15821
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4686413782680975e-05,
+      "loss": 0.3722,
+      "step": 15822
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4686077947374608e-05,
+      "loss": 0.3131,
+      "step": 15823
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4685742112068241e-05,
+      "loss": 0.3693,
+      "step": 15824
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4685406276761878e-05,
+      "loss": 0.2664,
+      "step": 15825
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.468507044145551e-05,
+      "loss": 0.2986,
+      "step": 15826
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4684734606149145e-05,
+      "loss": 0.3587,
+      "step": 15827
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.468439877084278e-05,
+      "loss": 0.4603,
+      "step": 15828
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4684062935536415e-05,
+      "loss": 0.2996,
+      "step": 15829
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4683727100230048e-05,
+      "loss": 0.2798,
+      "step": 15830
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4683391264923684e-05,
+      "loss": 0.2332,
+      "step": 15831
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4683055429617317e-05,
+      "loss": 0.2975,
+      "step": 15832
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.468271959431095e-05,
+      "loss": 0.3006,
+      "step": 15833
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4682383759004586e-05,
+      "loss": 0.4586,
+      "step": 15834
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4682047923698219e-05,
+      "loss": 0.3266,
+      "step": 15835
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4681712088391854e-05,
+      "loss": 0.194,
+      "step": 15836
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4681376253085488e-05,
+      "loss": 0.2868,
+      "step": 15837
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4681040417779121e-05,
+      "loss": 0.2542,
+      "step": 15838
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4680704582472756e-05,
+      "loss": 0.1869,
+      "step": 15839
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.468036874716639e-05,
+      "loss": 0.3835,
+      "step": 15840
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4680032911860025e-05,
+      "loss": 0.3247,
+      "step": 15841
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4679697076553658e-05,
+      "loss": 0.2489,
+      "step": 15842
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4679361241247295e-05,
+      "loss": 0.3606,
+      "step": 15843
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4679025405940928e-05,
+      "loss": 0.3117,
+      "step": 15844
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.467868957063456e-05,
+      "loss": 0.2599,
+      "step": 15845
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4678353735328197e-05,
+      "loss": 0.3504,
+      "step": 15846
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.467801790002183e-05,
+      "loss": 0.1777,
+      "step": 15847
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4677682064715464e-05,
+      "loss": 0.3235,
+      "step": 15848
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4677346229409099e-05,
+      "loss": 0.3421,
+      "step": 15849
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4677010394102734e-05,
+      "loss": 0.2702,
+      "step": 15850
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4676674558796367e-05,
+      "loss": 0.4102,
+      "step": 15851
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4676338723490003e-05,
+      "loss": 0.2045,
+      "step": 15852
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4676002888183636e-05,
+      "loss": 0.4265,
+      "step": 15853
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4675667052877269e-05,
+      "loss": 0.3464,
+      "step": 15854
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4675331217570905e-05,
+      "loss": 0.2859,
+      "step": 15855
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4674995382264538e-05,
+      "loss": 0.3172,
+      "step": 15856
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4674659546958173e-05,
+      "loss": 0.3686,
+      "step": 15857
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4674323711651808e-05,
+      "loss": 0.2515,
+      "step": 15858
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4673987876345442e-05,
+      "loss": 0.3122,
+      "step": 15859
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4673652041039075e-05,
+      "loss": 0.5444,
+      "step": 15860
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.467331620573271e-05,
+      "loss": 0.25,
+      "step": 15861
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4672980370426344e-05,
+      "loss": 0.2933,
+      "step": 15862
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4672644535119977e-05,
+      "loss": 0.2039,
+      "step": 15863
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4672308699813614e-05,
+      "loss": 0.2801,
+      "step": 15864
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4671972864507247e-05,
+      "loss": 0.3287,
+      "step": 15865
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.467163702920088e-05,
+      "loss": 0.2528,
+      "step": 15866
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4671301193894516e-05,
+      "loss": 0.3438,
+      "step": 15867
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4670965358588149e-05,
+      "loss": 0.3118,
+      "step": 15868
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4670629523281784e-05,
+      "loss": 0.2676,
+      "step": 15869
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4670293687975418e-05,
+      "loss": 0.3215,
+      "step": 15870
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4669957852669053e-05,
+      "loss": 0.1688,
+      "step": 15871
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4669622017362686e-05,
+      "loss": 0.2854,
+      "step": 15872
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4669286182056322e-05,
+      "loss": 0.228,
+      "step": 15873
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4668950346749955e-05,
+      "loss": 0.3131,
+      "step": 15874
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4668614511443588e-05,
+      "loss": 0.2784,
+      "step": 15875
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4668278676137224e-05,
+      "loss": 0.2793,
+      "step": 15876
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4667942840830857e-05,
+      "loss": 0.2492,
+      "step": 15877
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4667607005524492e-05,
+      "loss": 0.3944,
+      "step": 15878
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4667271170218127e-05,
+      "loss": 0.3543,
+      "step": 15879
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4666935334911761e-05,
+      "loss": 0.3394,
+      "step": 15880
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4666599499605394e-05,
+      "loss": 0.2347,
+      "step": 15881
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4666263664299029e-05,
+      "loss": 0.226,
+      "step": 15882
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4665927828992664e-05,
+      "loss": 0.2138,
+      "step": 15883
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4665591993686296e-05,
+      "loss": 0.4089,
+      "step": 15884
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4665256158379933e-05,
+      "loss": 0.2869,
+      "step": 15885
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4664920323073566e-05,
+      "loss": 0.1948,
+      "step": 15886
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4664584487767199e-05,
+      "loss": 0.3923,
+      "step": 15887
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4664248652460835e-05,
+      "loss": 0.3756,
+      "step": 15888
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4663912817154468e-05,
+      "loss": 0.3969,
+      "step": 15889
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4663576981848103e-05,
+      "loss": 0.2569,
+      "step": 15890
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4663241146541737e-05,
+      "loss": 0.2539,
+      "step": 15891
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4662905311235372e-05,
+      "loss": 0.3115,
+      "step": 15892
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4662569475929005e-05,
+      "loss": 0.3471,
+      "step": 15893
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4662233640622641e-05,
+      "loss": 0.2048,
+      "step": 15894
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4661897805316274e-05,
+      "loss": 0.3149,
+      "step": 15895
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4661561970009907e-05,
+      "loss": 0.4154,
+      "step": 15896
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4661226134703544e-05,
+      "loss": 0.4481,
+      "step": 15897
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4660890299397176e-05,
+      "loss": 0.179,
+      "step": 15898
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4660554464090811e-05,
+      "loss": 0.3257,
+      "step": 15899
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4660218628784446e-05,
+      "loss": 0.2937,
+      "step": 15900
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.465988279347808e-05,
+      "loss": 0.4899,
+      "step": 15901
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4659546958171713e-05,
+      "loss": 0.318,
+      "step": 15902
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.465921112286535e-05,
+      "loss": 0.3561,
+      "step": 15903
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4658875287558983e-05,
+      "loss": 0.266,
+      "step": 15904
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4658539452252616e-05,
+      "loss": 0.2192,
+      "step": 15905
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4658203616946252e-05,
+      "loss": 0.391,
+      "step": 15906
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4657867781639885e-05,
+      "loss": 0.3331,
+      "step": 15907
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4657531946333518e-05,
+      "loss": 0.33,
+      "step": 15908
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4657196111027154e-05,
+      "loss": 0.3071,
+      "step": 15909
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4656860275720787e-05,
+      "loss": 0.3278,
+      "step": 15910
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4656524440414422e-05,
+      "loss": 0.2593,
+      "step": 15911
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4656188605108056e-05,
+      "loss": 0.425,
+      "step": 15912
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4655852769801691e-05,
+      "loss": 0.2654,
+      "step": 15913
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4655516934495324e-05,
+      "loss": 0.2826,
+      "step": 15914
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.465518109918896e-05,
+      "loss": 0.4165,
+      "step": 15915
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4654845263882593e-05,
+      "loss": 0.3041,
+      "step": 15916
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4654509428576226e-05,
+      "loss": 0.3995,
+      "step": 15917
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4654173593269863e-05,
+      "loss": 0.2766,
+      "step": 15918
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4653837757963496e-05,
+      "loss": 0.2847,
+      "step": 15919
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.465350192265713e-05,
+      "loss": 0.2283,
+      "step": 15920
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4653166087350765e-05,
+      "loss": 0.2526,
+      "step": 15921
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.46528302520444e-05,
+      "loss": 0.3345,
+      "step": 15922
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4652494416738032e-05,
+      "loss": 0.2498,
+      "step": 15923
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4652158581431669e-05,
+      "loss": 0.2469,
+      "step": 15924
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4651822746125302e-05,
+      "loss": 0.1927,
+      "step": 15925
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4651486910818935e-05,
+      "loss": 0.3593,
+      "step": 15926
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4651151075512571e-05,
+      "loss": 0.3104,
+      "step": 15927
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4650815240206204e-05,
+      "loss": 0.4415,
+      "step": 15928
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4650479404899837e-05,
+      "loss": 0.3489,
+      "step": 15929
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4650143569593473e-05,
+      "loss": 0.3542,
+      "step": 15930
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4649807734287106e-05,
+      "loss": 0.3745,
+      "step": 15931
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4649471898980741e-05,
+      "loss": 0.2211,
+      "step": 15932
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4649136063674376e-05,
+      "loss": 0.4559,
+      "step": 15933
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.464880022836801e-05,
+      "loss": 0.5143,
+      "step": 15934
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4648464393061643e-05,
+      "loss": 0.2081,
+      "step": 15935
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.464812855775528e-05,
+      "loss": 0.3268,
+      "step": 15936
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4647792722448912e-05,
+      "loss": 0.2577,
+      "step": 15937
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4647456887142545e-05,
+      "loss": 0.2528,
+      "step": 15938
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4647121051836182e-05,
+      "loss": 0.3632,
+      "step": 15939
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4646785216529815e-05,
+      "loss": 0.2116,
+      "step": 15940
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.464644938122345e-05,
+      "loss": 0.4032,
+      "step": 15941
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4646113545917084e-05,
+      "loss": 0.3925,
+      "step": 15942
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4645777710610719e-05,
+      "loss": 0.2802,
+      "step": 15943
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4645441875304352e-05,
+      "loss": 0.332,
+      "step": 15944
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4645106039997988e-05,
+      "loss": 0.2955,
+      "step": 15945
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4644770204691621e-05,
+      "loss": 0.2475,
+      "step": 15946
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4644434369385254e-05,
+      "loss": 0.2827,
+      "step": 15947
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.464409853407889e-05,
+      "loss": 0.229,
+      "step": 15948
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4643762698772523e-05,
+      "loss": 0.3218,
+      "step": 15949
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4643426863466156e-05,
+      "loss": 0.5682,
+      "step": 15950
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4643091028159792e-05,
+      "loss": 0.3709,
+      "step": 15951
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4642755192853425e-05,
+      "loss": 0.203,
+      "step": 15952
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.464241935754706e-05,
+      "loss": 0.3883,
+      "step": 15953
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4642083522240695e-05,
+      "loss": 0.3384,
+      "step": 15954
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.464174768693433e-05,
+      "loss": 0.3225,
+      "step": 15955
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4641411851627962e-05,
+      "loss": 0.4,
+      "step": 15956
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4641076016321599e-05,
+      "loss": 0.2955,
+      "step": 15957
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4640740181015232e-05,
+      "loss": 0.2736,
+      "step": 15958
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4640404345708865e-05,
+      "loss": 0.3672,
+      "step": 15959
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4640068510402501e-05,
+      "loss": 0.4651,
+      "step": 15960
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4639732675096134e-05,
+      "loss": 0.3269,
+      "step": 15961
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4639396839789768e-05,
+      "loss": 0.3164,
+      "step": 15962
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4639061004483403e-05,
+      "loss": 0.3922,
+      "step": 15963
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4638725169177038e-05,
+      "loss": 0.2679,
+      "step": 15964
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.463838933387067e-05,
+      "loss": 0.3306,
+      "step": 15965
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4638053498564307e-05,
+      "loss": 0.3021,
+      "step": 15966
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.463771766325794e-05,
+      "loss": 0.2463,
+      "step": 15967
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4637381827951573e-05,
+      "loss": 0.2599,
+      "step": 15968
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.463704599264521e-05,
+      "loss": 0.2346,
+      "step": 15969
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4636710157338842e-05,
+      "loss": 0.3546,
+      "step": 15970
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4636374322032477e-05,
+      "loss": 0.1948,
+      "step": 15971
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.463603848672611e-05,
+      "loss": 0.4503,
+      "step": 15972
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4635702651419745e-05,
+      "loss": 0.2358,
+      "step": 15973
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.463536681611338e-05,
+      "loss": 0.2638,
+      "step": 15974
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4635030980807012e-05,
+      "loss": 0.5799,
+      "step": 15975
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4634695145500648e-05,
+      "loss": 0.3379,
+      "step": 15976
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4634359310194281e-05,
+      "loss": 0.2891,
+      "step": 15977
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4634023474887914e-05,
+      "loss": 0.3634,
+      "step": 15978
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.463368763958155e-05,
+      "loss": 0.2985,
+      "step": 15979
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.4633351804275184e-05,
+      "loss": 0.3145,
+      "step": 15980
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4633015968968818e-05,
+      "loss": 0.3241,
+      "step": 15981
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4632680133662453e-05,
+      "loss": 0.3138,
+      "step": 15982
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4632344298356088e-05,
+      "loss": 0.4163,
+      "step": 15983
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.463200846304972e-05,
+      "loss": 0.2283,
+      "step": 15984
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4631672627743357e-05,
+      "loss": 0.3198,
+      "step": 15985
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.463133679243699e-05,
+      "loss": 0.36,
+      "step": 15986
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4631000957130623e-05,
+      "loss": 0.4571,
+      "step": 15987
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4630665121824259e-05,
+      "loss": 0.2687,
+      "step": 15988
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4630329286517892e-05,
+      "loss": 0.5155,
+      "step": 15989
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4629993451211527e-05,
+      "loss": 0.2971,
+      "step": 15990
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4629657615905161e-05,
+      "loss": 0.4392,
+      "step": 15991
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4629321780598796e-05,
+      "loss": 0.2797,
+      "step": 15992
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4628985945292429e-05,
+      "loss": 0.3992,
+      "step": 15993
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4628650109986065e-05,
+      "loss": 0.4375,
+      "step": 15994
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4628314274679698e-05,
+      "loss": 0.4468,
+      "step": 15995
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4627978439373331e-05,
+      "loss": 0.3274,
+      "step": 15996
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4627642604066968e-05,
+      "loss": 0.3775,
+      "step": 15997
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.46273067687606e-05,
+      "loss": 0.3848,
+      "step": 15998
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4626970933454233e-05,
+      "loss": 0.1985,
+      "step": 15999
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.462663509814787e-05,
+      "loss": 0.3823,
+      "step": 16000
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4626299262841503e-05,
+      "loss": 0.286,
+      "step": 16001
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4625963427535137e-05,
+      "loss": 0.3927,
+      "step": 16002
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4625627592228772e-05,
+      "loss": 0.2832,
+      "step": 16003
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4625291756922407e-05,
+      "loss": 0.2617,
+      "step": 16004
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.462495592161604e-05,
+      "loss": 0.2357,
+      "step": 16005
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4624620086309676e-05,
+      "loss": 0.3294,
+      "step": 16006
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4624284251003309e-05,
+      "loss": 0.3926,
+      "step": 16007
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4623948415696942e-05,
+      "loss": 0.3168,
+      "step": 16008
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4623612580390578e-05,
+      "loss": 0.297,
+      "step": 16009
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4623276745084211e-05,
+      "loss": 0.2325,
+      "step": 16010
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4622940909777846e-05,
+      "loss": 0.2372,
+      "step": 16011
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.462260507447148e-05,
+      "loss": 0.2467,
+      "step": 16012
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4622269239165115e-05,
+      "loss": 0.3842,
+      "step": 16013
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4621933403858748e-05,
+      "loss": 0.2937,
+      "step": 16014
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4621597568552384e-05,
+      "loss": 0.1897,
+      "step": 16015
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4621261733246017e-05,
+      "loss": 0.388,
+      "step": 16016
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.462092589793965e-05,
+      "loss": 0.2827,
+      "step": 16017
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4620590062633287e-05,
+      "loss": 0.1947,
+      "step": 16018
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.462025422732692e-05,
+      "loss": 0.416,
+      "step": 16019
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4619918392020553e-05,
+      "loss": 0.3313,
+      "step": 16020
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4619582556714189e-05,
+      "loss": 0.4007,
+      "step": 16021
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4619246721407822e-05,
+      "loss": 0.2345,
+      "step": 16022
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4618910886101457e-05,
+      "loss": 0.2371,
+      "step": 16023
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4618575050795091e-05,
+      "loss": 0.3385,
+      "step": 16024
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4618239215488726e-05,
+      "loss": 0.1846,
+      "step": 16025
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4617903380182359e-05,
+      "loss": 0.1709,
+      "step": 16026
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4617567544875995e-05,
+      "loss": 0.3408,
+      "step": 16027
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4617231709569628e-05,
+      "loss": 0.321,
+      "step": 16028
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4616895874263261e-05,
+      "loss": 0.336,
+      "step": 16029
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4616560038956897e-05,
+      "loss": 0.2397,
+      "step": 16030
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.461622420365053e-05,
+      "loss": 0.2714,
+      "step": 16031
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4615888368344165e-05,
+      "loss": 0.3037,
+      "step": 16032
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.46155525330378e-05,
+      "loss": 0.4801,
+      "step": 16033
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4615216697731434e-05,
+      "loss": 0.2578,
+      "step": 16034
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4614880862425067e-05,
+      "loss": 0.4331,
+      "step": 16035
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4614545027118704e-05,
+      "loss": 0.3401,
+      "step": 16036
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4614209191812337e-05,
+      "loss": 0.2753,
+      "step": 16037
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.461387335650597e-05,
+      "loss": 0.3813,
+      "step": 16038
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4613537521199606e-05,
+      "loss": 0.3807,
+      "step": 16039
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4613201685893239e-05,
+      "loss": 0.3815,
+      "step": 16040
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4612865850586872e-05,
+      "loss": 0.3914,
+      "step": 16041
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4612530015280508e-05,
+      "loss": 0.303,
+      "step": 16042
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4612194179974141e-05,
+      "loss": 0.2977,
+      "step": 16043
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4611858344667776e-05,
+      "loss": 0.3805,
+      "step": 16044
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.461152250936141e-05,
+      "loss": 0.3609,
+      "step": 16045
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4611186674055045e-05,
+      "loss": 0.1871,
+      "step": 16046
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4610850838748678e-05,
+      "loss": 0.541,
+      "step": 16047
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4610515003442314e-05,
+      "loss": 0.2927,
+      "step": 16048
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4610179168135947e-05,
+      "loss": 0.1905,
+      "step": 16049
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.460984333282958e-05,
+      "loss": 0.2778,
+      "step": 16050
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4609507497523217e-05,
+      "loss": 0.4357,
+      "step": 16051
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.460917166221685e-05,
+      "loss": 0.1892,
+      "step": 16052
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4608835826910484e-05,
+      "loss": 0.303,
+      "step": 16053
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4608499991604119e-05,
+      "loss": 0.3731,
+      "step": 16054
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4608164156297753e-05,
+      "loss": 0.2774,
+      "step": 16055
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4607828320991386e-05,
+      "loss": 0.357,
+      "step": 16056
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4607492485685023e-05,
+      "loss": 0.292,
+      "step": 16057
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4607156650378656e-05,
+      "loss": 0.2449,
+      "step": 16058
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4606820815072289e-05,
+      "loss": 0.3044,
+      "step": 16059
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4606484979765925e-05,
+      "loss": 0.4785,
+      "step": 16060
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4606149144459558e-05,
+      "loss": 0.2918,
+      "step": 16061
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4605813309153193e-05,
+      "loss": 0.3671,
+      "step": 16062
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4605477473846827e-05,
+      "loss": 0.4253,
+      "step": 16063
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.460514163854046e-05,
+      "loss": 0.4821,
+      "step": 16064
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4604805803234095e-05,
+      "loss": 0.4058,
+      "step": 16065
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.460446996792773e-05,
+      "loss": 0.3868,
+      "step": 16066
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4604134132621364e-05,
+      "loss": 0.2764,
+      "step": 16067
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4603798297314997e-05,
+      "loss": 0.2652,
+      "step": 16068
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4603462462008633e-05,
+      "loss": 0.3534,
+      "step": 16069
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4603126626702266e-05,
+      "loss": 0.3403,
+      "step": 16070
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.46027907913959e-05,
+      "loss": 0.236,
+      "step": 16071
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4602454956089536e-05,
+      "loss": 0.286,
+      "step": 16072
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4602119120783169e-05,
+      "loss": 0.183,
+      "step": 16073
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4601783285476803e-05,
+      "loss": 0.389,
+      "step": 16074
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4601447450170438e-05,
+      "loss": 0.4042,
+      "step": 16075
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4601111614864073e-05,
+      "loss": 0.3092,
+      "step": 16076
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4600775779557705e-05,
+      "loss": 0.2789,
+      "step": 16077
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4600439944251342e-05,
+      "loss": 0.3477,
+      "step": 16078
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4600104108944975e-05,
+      "loss": 0.2939,
+      "step": 16079
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4599768273638608e-05,
+      "loss": 0.3818,
+      "step": 16080
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4599432438332244e-05,
+      "loss": 0.3477,
+      "step": 16081
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4599096603025877e-05,
+      "loss": 0.4682,
+      "step": 16082
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4598760767719512e-05,
+      "loss": 0.2852,
+      "step": 16083
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4598424932413146e-05,
+      "loss": 0.3759,
+      "step": 16084
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.459808909710678e-05,
+      "loss": 0.4945,
+      "step": 16085
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4597753261800414e-05,
+      "loss": 0.3321,
+      "step": 16086
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4597417426494049e-05,
+      "loss": 0.3905,
+      "step": 16087
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4597081591187683e-05,
+      "loss": 0.373,
+      "step": 16088
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4596745755881316e-05,
+      "loss": 0.2749,
+      "step": 16089
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4596409920574953e-05,
+      "loss": 0.278,
+      "step": 16090
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4596074085268585e-05,
+      "loss": 0.4631,
+      "step": 16091
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4595738249962218e-05,
+      "loss": 0.2975,
+      "step": 16092
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4595402414655855e-05,
+      "loss": 0.2908,
+      "step": 16093
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4595066579349488e-05,
+      "loss": 0.3353,
+      "step": 16094
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4594730744043122e-05,
+      "loss": 0.3817,
+      "step": 16095
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4594394908736757e-05,
+      "loss": 0.3309,
+      "step": 16096
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4594059073430392e-05,
+      "loss": 0.2033,
+      "step": 16097
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4593723238124025e-05,
+      "loss": 0.3447,
+      "step": 16098
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4593387402817661e-05,
+      "loss": 0.2709,
+      "step": 16099
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4593051567511294e-05,
+      "loss": 0.4146,
+      "step": 16100
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4592715732204927e-05,
+      "loss": 0.324,
+      "step": 16101
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4592379896898563e-05,
+      "loss": 0.4127,
+      "step": 16102
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4592044061592196e-05,
+      "loss": 0.252,
+      "step": 16103
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.459170822628583e-05,
+      "loss": 0.3692,
+      "step": 16104
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4591372390979465e-05,
+      "loss": 0.2534,
+      "step": 16105
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.45910365556731e-05,
+      "loss": 0.2684,
+      "step": 16106
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4590700720366733e-05,
+      "loss": 0.2575,
+      "step": 16107
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4590364885060368e-05,
+      "loss": 0.3105,
+      "step": 16108
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4590029049754002e-05,
+      "loss": 0.215,
+      "step": 16109
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4589693214447635e-05,
+      "loss": 0.228,
+      "step": 16110
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4589357379141272e-05,
+      "loss": 0.2697,
+      "step": 16111
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4589021543834905e-05,
+      "loss": 0.4786,
+      "step": 16112
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4588685708528538e-05,
+      "loss": 0.4679,
+      "step": 16113
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4588349873222174e-05,
+      "loss": 0.3012,
+      "step": 16114
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4588014037915807e-05,
+      "loss": 0.3121,
+      "step": 16115
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4587678202609441e-05,
+      "loss": 0.333,
+      "step": 16116
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4587342367303076e-05,
+      "loss": 0.3353,
+      "step": 16117
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.458700653199671e-05,
+      "loss": 0.1193,
+      "step": 16118
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4586670696690344e-05,
+      "loss": 0.1855,
+      "step": 16119
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.458633486138398e-05,
+      "loss": 0.4242,
+      "step": 16120
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4585999026077613e-05,
+      "loss": 0.3802,
+      "step": 16121
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4585663190771246e-05,
+      "loss": 0.243,
+      "step": 16122
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4585327355464882e-05,
+      "loss": 0.3387,
+      "step": 16123
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4584991520158515e-05,
+      "loss": 0.2775,
+      "step": 16124
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.458465568485215e-05,
+      "loss": 0.3168,
+      "step": 16125
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4584319849545785e-05,
+      "loss": 0.1494,
+      "step": 16126
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.458398401423942e-05,
+      "loss": 0.2315,
+      "step": 16127
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4583648178933052e-05,
+      "loss": 0.4203,
+      "step": 16128
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4583312343626689e-05,
+      "loss": 0.1849,
+      "step": 16129
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4582976508320321e-05,
+      "loss": 0.1638,
+      "step": 16130
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4582640673013954e-05,
+      "loss": 0.3733,
+      "step": 16131
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.458230483770759e-05,
+      "loss": 0.4297,
+      "step": 16132
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4581969002401224e-05,
+      "loss": 0.1843,
+      "step": 16133
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4581633167094857e-05,
+      "loss": 0.4008,
+      "step": 16134
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4581297331788493e-05,
+      "loss": 0.2328,
+      "step": 16135
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4580961496482126e-05,
+      "loss": 0.4202,
+      "step": 16136
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.458062566117576e-05,
+      "loss": 0.5176,
+      "step": 16137
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4580289825869395e-05,
+      "loss": 0.4213,
+      "step": 16138
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.457995399056303e-05,
+      "loss": 0.2993,
+      "step": 16139
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4579618155256663e-05,
+      "loss": 0.2982,
+      "step": 16140
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.45792823199503e-05,
+      "loss": 0.3632,
+      "step": 16141
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4578946484643932e-05,
+      "loss": 0.2537,
+      "step": 16142
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4578610649337565e-05,
+      "loss": 0.253,
+      "step": 16143
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4578274814031201e-05,
+      "loss": 0.4188,
+      "step": 16144
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4577938978724834e-05,
+      "loss": 0.3172,
+      "step": 16145
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4577603143418469e-05,
+      "loss": 0.4113,
+      "step": 16146
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4577267308112104e-05,
+      "loss": 0.4112,
+      "step": 16147
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4576931472805738e-05,
+      "loss": 0.474,
+      "step": 16148
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4576595637499371e-05,
+      "loss": 0.1972,
+      "step": 16149
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4576259802193008e-05,
+      "loss": 0.3447,
+      "step": 16150
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.457592396688664e-05,
+      "loss": 0.2522,
+      "step": 16151
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4575588131580274e-05,
+      "loss": 0.2376,
+      "step": 16152
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4575252296273908e-05,
+      "loss": 0.314,
+      "step": 16153
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4574916460967543e-05,
+      "loss": 0.2461,
+      "step": 16154
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4574580625661176e-05,
+      "loss": 0.3093,
+      "step": 16155
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.457424479035481e-05,
+      "loss": 0.3024,
+      "step": 16156
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4573908955048445e-05,
+      "loss": 0.2562,
+      "step": 16157
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.457357311974208e-05,
+      "loss": 0.3782,
+      "step": 16158
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4573237284435713e-05,
+      "loss": 0.3396,
+      "step": 16159
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4572901449129349e-05,
+      "loss": 0.2511,
+      "step": 16160
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4572565613822982e-05,
+      "loss": 0.3467,
+      "step": 16161
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4572229778516615e-05,
+      "loss": 0.2538,
+      "step": 16162
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4571893943210251e-05,
+      "loss": 0.3094,
+      "step": 16163
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4571558107903884e-05,
+      "loss": 0.2944,
+      "step": 16164
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4571222272597519e-05,
+      "loss": 0.2426,
+      "step": 16165
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4570886437291154e-05,
+      "loss": 0.359,
+      "step": 16166
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4570550601984788e-05,
+      "loss": 0.4759,
+      "step": 16167
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4570214766678421e-05,
+      "loss": 0.2198,
+      "step": 16168
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4569878931372057e-05,
+      "loss": 0.3895,
+      "step": 16169
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.456954309606569e-05,
+      "loss": 0.265,
+      "step": 16170
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4569207260759323e-05,
+      "loss": 0.4538,
+      "step": 16171
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.456887142545296e-05,
+      "loss": 0.3123,
+      "step": 16172
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4568535590146593e-05,
+      "loss": 0.2972,
+      "step": 16173
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4568199754840227e-05,
+      "loss": 0.332,
+      "step": 16174
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4567863919533862e-05,
+      "loss": 0.1986,
+      "step": 16175
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4567528084227495e-05,
+      "loss": 0.2852,
+      "step": 16176
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.456719224892113e-05,
+      "loss": 0.1997,
+      "step": 16177
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.4566856413614764e-05,
+      "loss": 0.3196,
+      "step": 16178
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4566520578308399e-05,
+      "loss": 0.4024,
+      "step": 16179
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4566184743002032e-05,
+      "loss": 0.4168,
+      "step": 16180
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4565848907695668e-05,
+      "loss": 0.379,
+      "step": 16181
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4565513072389301e-05,
+      "loss": 0.14,
+      "step": 16182
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4565177237082934e-05,
+      "loss": 0.2379,
+      "step": 16183
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.456484140177657e-05,
+      "loss": 0.3782,
+      "step": 16184
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4564505566470203e-05,
+      "loss": 0.2001,
+      "step": 16185
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4564169731163838e-05,
+      "loss": 0.3769,
+      "step": 16186
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4563833895857473e-05,
+      "loss": 0.3262,
+      "step": 16187
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4563498060551107e-05,
+      "loss": 0.2683,
+      "step": 16188
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.456316222524474e-05,
+      "loss": 0.2773,
+      "step": 16189
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4562826389938377e-05,
+      "loss": 0.3439,
+      "step": 16190
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.456249055463201e-05,
+      "loss": 0.3153,
+      "step": 16191
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4562154719325643e-05,
+      "loss": 0.3071,
+      "step": 16192
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4561818884019279e-05,
+      "loss": 0.2287,
+      "step": 16193
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4561483048712912e-05,
+      "loss": 0.2506,
+      "step": 16194
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4561147213406546e-05,
+      "loss": 0.3864,
+      "step": 16195
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4560811378100181e-05,
+      "loss": 0.437,
+      "step": 16196
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4560475542793816e-05,
+      "loss": 0.2101,
+      "step": 16197
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4560139707487449e-05,
+      "loss": 0.3969,
+      "step": 16198
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4559803872181083e-05,
+      "loss": 0.3433,
+      "step": 16199
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4559468036874718e-05,
+      "loss": 0.5807,
+      "step": 16200
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4559132201568351e-05,
+      "loss": 0.309,
+      "step": 16201
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4558796366261987e-05,
+      "loss": 0.3892,
+      "step": 16202
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.455846053095562e-05,
+      "loss": 0.2993,
+      "step": 16203
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4558124695649253e-05,
+      "loss": 0.2364,
+      "step": 16204
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.455778886034289e-05,
+      "loss": 0.2907,
+      "step": 16205
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4557453025036522e-05,
+      "loss": 0.2828,
+      "step": 16206
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4557117189730157e-05,
+      "loss": 0.4627,
+      "step": 16207
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4556781354423792e-05,
+      "loss": 0.3051,
+      "step": 16208
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4556445519117426e-05,
+      "loss": 0.2629,
+      "step": 16209
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.455610968381106e-05,
+      "loss": 0.3772,
+      "step": 16210
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4555773848504696e-05,
+      "loss": 0.4906,
+      "step": 16211
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4555438013198329e-05,
+      "loss": 0.2973,
+      "step": 16212
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4555102177891962e-05,
+      "loss": 0.3724,
+      "step": 16213
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4554766342585598e-05,
+      "loss": 0.1584,
+      "step": 16214
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4554430507279231e-05,
+      "loss": 0.3635,
+      "step": 16215
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4554094671972866e-05,
+      "loss": 0.3669,
+      "step": 16216
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.45537588366665e-05,
+      "loss": 0.4647,
+      "step": 16217
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4553423001360135e-05,
+      "loss": 0.2812,
+      "step": 16218
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4553087166053768e-05,
+      "loss": 0.3467,
+      "step": 16219
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4552751330747402e-05,
+      "loss": 0.2594,
+      "step": 16220
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4552415495441037e-05,
+      "loss": 0.2284,
+      "step": 16221
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.455207966013467e-05,
+      "loss": 0.2341,
+      "step": 16222
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4551743824828306e-05,
+      "loss": 0.3421,
+      "step": 16223
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.455140798952194e-05,
+      "loss": 0.3304,
+      "step": 16224
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4551072154215572e-05,
+      "loss": 0.2939,
+      "step": 16225
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4550736318909209e-05,
+      "loss": 0.3486,
+      "step": 16226
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4550400483602842e-05,
+      "loss": 0.2382,
+      "step": 16227
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4550064648296476e-05,
+      "loss": 0.3526,
+      "step": 16228
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4549728812990111e-05,
+      "loss": 0.2231,
+      "step": 16229
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4549392977683746e-05,
+      "loss": 0.1844,
+      "step": 16230
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4549057142377379e-05,
+      "loss": 0.3571,
+      "step": 16231
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4548721307071015e-05,
+      "loss": 0.3158,
+      "step": 16232
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4548385471764648e-05,
+      "loss": 0.339,
+      "step": 16233
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.454804963645828e-05,
+      "loss": 0.3879,
+      "step": 16234
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4547713801151917e-05,
+      "loss": 0.2938,
+      "step": 16235
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.454737796584555e-05,
+      "loss": 0.1828,
+      "step": 16236
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4547042130539185e-05,
+      "loss": 0.2382,
+      "step": 16237
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.454670629523282e-05,
+      "loss": 0.3954,
+      "step": 16238
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4546370459926454e-05,
+      "loss": 0.319,
+      "step": 16239
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4546034624620087e-05,
+      "loss": 0.2931,
+      "step": 16240
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4545698789313723e-05,
+      "loss": 0.4563,
+      "step": 16241
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4545362954007356e-05,
+      "loss": 0.2403,
+      "step": 16242
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.454502711870099e-05,
+      "loss": 0.2399,
+      "step": 16243
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4544691283394626e-05,
+      "loss": 0.3796,
+      "step": 16244
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4544355448088258e-05,
+      "loss": 0.3645,
+      "step": 16245
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4544019612781891e-05,
+      "loss": 0.2734,
+      "step": 16246
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4543683777475528e-05,
+      "loss": 0.4836,
+      "step": 16247
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.454334794216916e-05,
+      "loss": 0.3804,
+      "step": 16248
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4543012106862795e-05,
+      "loss": 0.6082,
+      "step": 16249
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.454267627155643e-05,
+      "loss": 0.3289,
+      "step": 16250
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4542340436250065e-05,
+      "loss": 0.2982,
+      "step": 16251
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4542004600943698e-05,
+      "loss": 0.3268,
+      "step": 16252
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4541668765637334e-05,
+      "loss": 0.442,
+      "step": 16253
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4541332930330967e-05,
+      "loss": 0.1885,
+      "step": 16254
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.45409970950246e-05,
+      "loss": 0.3579,
+      "step": 16255
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4540661259718236e-05,
+      "loss": 0.2959,
+      "step": 16256
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.454032542441187e-05,
+      "loss": 0.2733,
+      "step": 16257
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4539989589105504e-05,
+      "loss": 0.2978,
+      "step": 16258
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4539653753799138e-05,
+      "loss": 0.2494,
+      "step": 16259
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4539317918492773e-05,
+      "loss": 0.3554,
+      "step": 16260
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4538982083186406e-05,
+      "loss": 0.2521,
+      "step": 16261
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4538646247880042e-05,
+      "loss": 0.3006,
+      "step": 16262
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4538310412573675e-05,
+      "loss": 0.3235,
+      "step": 16263
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4537974577267308e-05,
+      "loss": 0.3017,
+      "step": 16264
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4537638741960945e-05,
+      "loss": 0.2179,
+      "step": 16265
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4537302906654578e-05,
+      "loss": 0.3133,
+      "step": 16266
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.453696707134821e-05,
+      "loss": 0.2105,
+      "step": 16267
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4536631236041847e-05,
+      "loss": 0.406,
+      "step": 16268
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.453629540073548e-05,
+      "loss": 0.3436,
+      "step": 16269
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4535959565429115e-05,
+      "loss": 0.4067,
+      "step": 16270
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.453562373012275e-05,
+      "loss": 0.2683,
+      "step": 16271
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4535287894816384e-05,
+      "loss": 0.5414,
+      "step": 16272
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4534952059510017e-05,
+      "loss": 0.2946,
+      "step": 16273
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4534616224203653e-05,
+      "loss": 0.2537,
+      "step": 16274
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4534280388897286e-05,
+      "loss": 0.3402,
+      "step": 16275
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4533944553590919e-05,
+      "loss": 0.4535,
+      "step": 16276
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4533608718284555e-05,
+      "loss": 0.4561,
+      "step": 16277
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4533272882978188e-05,
+      "loss": 0.3656,
+      "step": 16278
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4532937047671823e-05,
+      "loss": 0.3578,
+      "step": 16279
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4532601212365458e-05,
+      "loss": 0.1703,
+      "step": 16280
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4532265377059092e-05,
+      "loss": 0.3258,
+      "step": 16281
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4531929541752725e-05,
+      "loss": 0.2976,
+      "step": 16282
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4531593706446362e-05,
+      "loss": 0.3161,
+      "step": 16283
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4531257871139994e-05,
+      "loss": 0.3673,
+      "step": 16284
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4530922035833627e-05,
+      "loss": 0.328,
+      "step": 16285
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4530586200527264e-05,
+      "loss": 0.3667,
+      "step": 16286
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4530250365220897e-05,
+      "loss": 0.4049,
+      "step": 16287
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4529914529914531e-05,
+      "loss": 0.2091,
+      "step": 16288
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4529578694608166e-05,
+      "loss": 0.3267,
+      "step": 16289
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4529242859301799e-05,
+      "loss": 0.3355,
+      "step": 16290
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4528907023995434e-05,
+      "loss": 0.4534,
+      "step": 16291
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4528571188689068e-05,
+      "loss": 0.2065,
+      "step": 16292
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4528235353382703e-05,
+      "loss": 0.2885,
+      "step": 16293
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4527899518076336e-05,
+      "loss": 0.3273,
+      "step": 16294
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4527563682769972e-05,
+      "loss": 0.3506,
+      "step": 16295
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4527227847463605e-05,
+      "loss": 0.397,
+      "step": 16296
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4526892012157238e-05,
+      "loss": 0.3361,
+      "step": 16297
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4526556176850874e-05,
+      "loss": 0.3585,
+      "step": 16298
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4526220341544507e-05,
+      "loss": 0.3337,
+      "step": 16299
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4525884506238142e-05,
+      "loss": 0.1404,
+      "step": 16300
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4525548670931777e-05,
+      "loss": 0.3829,
+      "step": 16301
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4525212835625411e-05,
+      "loss": 0.3112,
+      "step": 16302
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4524877000319044e-05,
+      "loss": 0.2676,
+      "step": 16303
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.452454116501268e-05,
+      "loss": 0.4549,
+      "step": 16304
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4524205329706314e-05,
+      "loss": 0.3328,
+      "step": 16305
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4523869494399947e-05,
+      "loss": 0.2125,
+      "step": 16306
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4523533659093583e-05,
+      "loss": 0.5908,
+      "step": 16307
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4523197823787216e-05,
+      "loss": 0.3285,
+      "step": 16308
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.452286198848085e-05,
+      "loss": 0.3304,
+      "step": 16309
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4522526153174485e-05,
+      "loss": 0.2842,
+      "step": 16310
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4522190317868118e-05,
+      "loss": 0.2671,
+      "step": 16311
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4521854482561753e-05,
+      "loss": 0.2532,
+      "step": 16312
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4521518647255387e-05,
+      "loss": 0.3948,
+      "step": 16313
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4521182811949022e-05,
+      "loss": 0.2149,
+      "step": 16314
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4520846976642655e-05,
+      "loss": 0.3571,
+      "step": 16315
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4520511141336291e-05,
+      "loss": 0.286,
+      "step": 16316
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4520175306029924e-05,
+      "loss": 0.2145,
+      "step": 16317
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4519839470723557e-05,
+      "loss": 0.3421,
+      "step": 16318
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4519503635417194e-05,
+      "loss": 0.582,
+      "step": 16319
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4519167800110827e-05,
+      "loss": 0.3466,
+      "step": 16320
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4518831964804461e-05,
+      "loss": 0.3446,
+      "step": 16321
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4518496129498096e-05,
+      "loss": 0.4579,
+      "step": 16322
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.451816029419173e-05,
+      "loss": 0.2449,
+      "step": 16323
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4517824458885363e-05,
+      "loss": 0.2387,
+      "step": 16324
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4517488623579e-05,
+      "loss": 0.2979,
+      "step": 16325
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4517152788272633e-05,
+      "loss": 0.3165,
+      "step": 16326
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4516816952966266e-05,
+      "loss": 0.3499,
+      "step": 16327
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4516481117659902e-05,
+      "loss": 0.366,
+      "step": 16328
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4516145282353535e-05,
+      "loss": 0.4178,
+      "step": 16329
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.451580944704717e-05,
+      "loss": 0.3461,
+      "step": 16330
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4515473611740804e-05,
+      "loss": 0.3003,
+      "step": 16331
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4515137776434439e-05,
+      "loss": 0.3452,
+      "step": 16332
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4514801941128072e-05,
+      "loss": 0.264,
+      "step": 16333
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4514466105821705e-05,
+      "loss": 0.2614,
+      "step": 16334
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4514130270515341e-05,
+      "loss": 0.3568,
+      "step": 16335
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4513794435208974e-05,
+      "loss": 0.3365,
+      "step": 16336
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4513458599902607e-05,
+      "loss": 0.3309,
+      "step": 16337
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4513122764596243e-05,
+      "loss": 0.2385,
+      "step": 16338
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4512786929289876e-05,
+      "loss": 0.4435,
+      "step": 16339
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4512451093983511e-05,
+      "loss": 0.5607,
+      "step": 16340
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4512115258677146e-05,
+      "loss": 0.2705,
+      "step": 16341
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.451177942337078e-05,
+      "loss": 0.5509,
+      "step": 16342
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4511443588064413e-05,
+      "loss": 0.3223,
+      "step": 16343
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.451110775275805e-05,
+      "loss": 0.2802,
+      "step": 16344
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4510771917451683e-05,
+      "loss": 0.2018,
+      "step": 16345
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4510436082145316e-05,
+      "loss": 0.2573,
+      "step": 16346
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4510100246838952e-05,
+      "loss": 0.2565,
+      "step": 16347
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4509764411532585e-05,
+      "loss": 0.2343,
+      "step": 16348
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.450942857622622e-05,
+      "loss": 0.3248,
+      "step": 16349
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4509092740919854e-05,
+      "loss": 0.3541,
+      "step": 16350
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4508756905613489e-05,
+      "loss": 0.2893,
+      "step": 16351
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4508421070307122e-05,
+      "loss": 0.3412,
+      "step": 16352
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4508085235000758e-05,
+      "loss": 0.3287,
+      "step": 16353
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4507749399694391e-05,
+      "loss": 0.2504,
+      "step": 16354
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4507413564388024e-05,
+      "loss": 0.2852,
+      "step": 16355
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.450707772908166e-05,
+      "loss": 0.3853,
+      "step": 16356
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4506741893775293e-05,
+      "loss": 0.302,
+      "step": 16357
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4506406058468926e-05,
+      "loss": 0.3448,
+      "step": 16358
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4506070223162563e-05,
+      "loss": 0.2083,
+      "step": 16359
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4505734387856196e-05,
+      "loss": 0.3106,
+      "step": 16360
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.450539855254983e-05,
+      "loss": 0.2633,
+      "step": 16361
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4505062717243465e-05,
+      "loss": 0.4016,
+      "step": 16362
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.45047268819371e-05,
+      "loss": 0.3402,
+      "step": 16363
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4504391046630732e-05,
+      "loss": 0.4107,
+      "step": 16364
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4504055211324369e-05,
+      "loss": 0.5097,
+      "step": 16365
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4503719376018002e-05,
+      "loss": 0.2737,
+      "step": 16366
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4503383540711635e-05,
+      "loss": 0.2927,
+      "step": 16367
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4503047705405271e-05,
+      "loss": 0.3765,
+      "step": 16368
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4502711870098904e-05,
+      "loss": 0.3655,
+      "step": 16369
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4502376034792539e-05,
+      "loss": 0.2431,
+      "step": 16370
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4502040199486173e-05,
+      "loss": 0.3358,
+      "step": 16371
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4501704364179808e-05,
+      "loss": 0.2524,
+      "step": 16372
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.450136852887344e-05,
+      "loss": 0.2993,
+      "step": 16373
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4501032693567077e-05,
+      "loss": 0.3096,
+      "step": 16374
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.450069685826071e-05,
+      "loss": 0.5418,
+      "step": 16375
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.4500361022954343e-05,
+      "loss": 0.3463,
+      "step": 16376
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.450002518764798e-05,
+      "loss": 0.2884,
+      "step": 16377
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4499689352341612e-05,
+      "loss": 0.2074,
+      "step": 16378
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4499353517035245e-05,
+      "loss": 0.3324,
+      "step": 16379
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4499017681728882e-05,
+      "loss": 0.4315,
+      "step": 16380
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4498681846422515e-05,
+      "loss": 0.1666,
+      "step": 16381
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.449834601111615e-05,
+      "loss": 0.2817,
+      "step": 16382
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4498010175809784e-05,
+      "loss": 0.2948,
+      "step": 16383
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4497674340503419e-05,
+      "loss": 0.3366,
+      "step": 16384
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4497338505197052e-05,
+      "loss": 0.3735,
+      "step": 16385
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4497002669890688e-05,
+      "loss": 0.3354,
+      "step": 16386
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.449666683458432e-05,
+      "loss": 0.5044,
+      "step": 16387
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4496330999277954e-05,
+      "loss": 0.5461,
+      "step": 16388
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.449599516397159e-05,
+      "loss": 0.3827,
+      "step": 16389
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4495659328665223e-05,
+      "loss": 0.2686,
+      "step": 16390
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4495323493358858e-05,
+      "loss": 0.5091,
+      "step": 16391
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4494987658052492e-05,
+      "loss": 0.3468,
+      "step": 16392
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4494651822746127e-05,
+      "loss": 0.2888,
+      "step": 16393
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.449431598743976e-05,
+      "loss": 0.3248,
+      "step": 16394
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4493980152133396e-05,
+      "loss": 0.4201,
+      "step": 16395
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.449364431682703e-05,
+      "loss": 0.4136,
+      "step": 16396
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4493308481520662e-05,
+      "loss": 0.3224,
+      "step": 16397
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4492972646214299e-05,
+      "loss": 0.3804,
+      "step": 16398
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4492636810907932e-05,
+      "loss": 0.2429,
+      "step": 16399
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4492300975601566e-05,
+      "loss": 0.3009,
+      "step": 16400
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.44919651402952e-05,
+      "loss": 0.6488,
+      "step": 16401
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4491629304988834e-05,
+      "loss": 0.1919,
+      "step": 16402
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4491293469682468e-05,
+      "loss": 0.2233,
+      "step": 16403
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4490957634376103e-05,
+      "loss": 0.3045,
+      "step": 16404
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4490621799069738e-05,
+      "loss": 0.2463,
+      "step": 16405
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.449028596376337e-05,
+      "loss": 0.2853,
+      "step": 16406
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4489950128457007e-05,
+      "loss": 0.3537,
+      "step": 16407
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.448961429315064e-05,
+      "loss": 0.1537,
+      "step": 16408
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4489278457844273e-05,
+      "loss": 0.2105,
+      "step": 16409
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.448894262253791e-05,
+      "loss": 0.3503,
+      "step": 16410
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4488606787231542e-05,
+      "loss": 0.2656,
+      "step": 16411
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4488270951925177e-05,
+      "loss": 0.2188,
+      "step": 16412
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4487935116618811e-05,
+      "loss": 0.3306,
+      "step": 16413
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4487599281312446e-05,
+      "loss": 0.2307,
+      "step": 16414
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4487263446006079e-05,
+      "loss": 0.3891,
+      "step": 16415
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4486927610699715e-05,
+      "loss": 0.3468,
+      "step": 16416
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4486591775393348e-05,
+      "loss": 0.5178,
+      "step": 16417
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4486255940086981e-05,
+      "loss": 0.4107,
+      "step": 16418
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4485920104780618e-05,
+      "loss": 0.358,
+      "step": 16419
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.448558426947425e-05,
+      "loss": 0.2577,
+      "step": 16420
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4485248434167885e-05,
+      "loss": 0.2843,
+      "step": 16421
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.448491259886152e-05,
+      "loss": 0.4247,
+      "step": 16422
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4484576763555155e-05,
+      "loss": 0.258,
+      "step": 16423
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4484240928248788e-05,
+      "loss": 0.1935,
+      "step": 16424
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4483905092942422e-05,
+      "loss": 0.2619,
+      "step": 16425
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4483569257636057e-05,
+      "loss": 0.2719,
+      "step": 16426
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.448323342232969e-05,
+      "loss": 0.3262,
+      "step": 16427
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4482897587023326e-05,
+      "loss": 0.4528,
+      "step": 16428
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4482561751716959e-05,
+      "loss": 0.2663,
+      "step": 16429
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4482225916410592e-05,
+      "loss": 0.2967,
+      "step": 16430
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4481890081104228e-05,
+      "loss": 0.2722,
+      "step": 16431
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4481554245797861e-05,
+      "loss": 0.4958,
+      "step": 16432
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4481218410491496e-05,
+      "loss": 0.2683,
+      "step": 16433
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.448088257518513e-05,
+      "loss": 0.2897,
+      "step": 16434
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4480546739878765e-05,
+      "loss": 0.4095,
+      "step": 16435
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4480210904572398e-05,
+      "loss": 0.3565,
+      "step": 16436
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4479875069266035e-05,
+      "loss": 0.2755,
+      "step": 16437
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4479539233959668e-05,
+      "loss": 0.2235,
+      "step": 16438
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.44792033986533e-05,
+      "loss": 0.2011,
+      "step": 16439
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4478867563346937e-05,
+      "loss": 0.1618,
+      "step": 16440
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.447853172804057e-05,
+      "loss": 0.3173,
+      "step": 16441
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4478195892734204e-05,
+      "loss": 0.3961,
+      "step": 16442
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4477860057427839e-05,
+      "loss": 0.4378,
+      "step": 16443
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4477524222121474e-05,
+      "loss": 0.2448,
+      "step": 16444
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4477188386815107e-05,
+      "loss": 0.5104,
+      "step": 16445
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4476852551508741e-05,
+      "loss": 0.3418,
+      "step": 16446
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4476516716202376e-05,
+      "loss": 0.2855,
+      "step": 16447
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4476180880896009e-05,
+      "loss": 0.2767,
+      "step": 16448
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4475845045589645e-05,
+      "loss": 0.5101,
+      "step": 16449
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4475509210283278e-05,
+      "loss": 0.2351,
+      "step": 16450
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4475173374976911e-05,
+      "loss": 0.4062,
+      "step": 16451
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4474837539670547e-05,
+      "loss": 0.3792,
+      "step": 16452
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.447450170436418e-05,
+      "loss": 0.3061,
+      "step": 16453
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4474165869057815e-05,
+      "loss": 0.4109,
+      "step": 16454
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.447383003375145e-05,
+      "loss": 0.4586,
+      "step": 16455
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4473494198445084e-05,
+      "loss": 0.3124,
+      "step": 16456
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4473158363138717e-05,
+      "loss": 0.401,
+      "step": 16457
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4472822527832354e-05,
+      "loss": 0.383,
+      "step": 16458
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4472486692525987e-05,
+      "loss": 0.4153,
+      "step": 16459
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.447215085721962e-05,
+      "loss": 0.3652,
+      "step": 16460
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4471815021913256e-05,
+      "loss": 0.3918,
+      "step": 16461
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4471479186606889e-05,
+      "loss": 0.3959,
+      "step": 16462
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4471143351300524e-05,
+      "loss": 0.347,
+      "step": 16463
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4470807515994158e-05,
+      "loss": 0.269,
+      "step": 16464
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4470471680687793e-05,
+      "loss": 0.3921,
+      "step": 16465
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4470135845381426e-05,
+      "loss": 0.3382,
+      "step": 16466
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4469800010075062e-05,
+      "loss": 0.2819,
+      "step": 16467
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4469464174768695e-05,
+      "loss": 0.4827,
+      "step": 16468
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4469128339462328e-05,
+      "loss": 0.3427,
+      "step": 16469
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4468792504155964e-05,
+      "loss": 0.2448,
+      "step": 16470
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4468456668849597e-05,
+      "loss": 0.3214,
+      "step": 16471
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.446812083354323e-05,
+      "loss": 0.4876,
+      "step": 16472
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4467784998236867e-05,
+      "loss": 0.14,
+      "step": 16473
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.44674491629305e-05,
+      "loss": 0.4076,
+      "step": 16474
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4467113327624134e-05,
+      "loss": 0.2926,
+      "step": 16475
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4466777492317769e-05,
+      "loss": 0.265,
+      "step": 16476
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4466441657011404e-05,
+      "loss": 0.2044,
+      "step": 16477
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4466105821705036e-05,
+      "loss": 0.48,
+      "step": 16478
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4465769986398673e-05,
+      "loss": 0.3805,
+      "step": 16479
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4465434151092306e-05,
+      "loss": 0.2214,
+      "step": 16480
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4465098315785939e-05,
+      "loss": 0.3331,
+      "step": 16481
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4464762480479575e-05,
+      "loss": 0.4646,
+      "step": 16482
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4464426645173208e-05,
+      "loss": 0.2406,
+      "step": 16483
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4464090809866843e-05,
+      "loss": 0.3256,
+      "step": 16484
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4463754974560477e-05,
+      "loss": 0.4249,
+      "step": 16485
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4463419139254112e-05,
+      "loss": 0.2711,
+      "step": 16486
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4463083303947745e-05,
+      "loss": 0.5218,
+      "step": 16487
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4462747468641381e-05,
+      "loss": 0.403,
+      "step": 16488
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4462411633335014e-05,
+      "loss": 0.3463,
+      "step": 16489
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4462075798028647e-05,
+      "loss": 0.1589,
+      "step": 16490
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4461739962722283e-05,
+      "loss": 0.1982,
+      "step": 16491
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4461404127415916e-05,
+      "loss": 0.2826,
+      "step": 16492
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.446106829210955e-05,
+      "loss": 0.2859,
+      "step": 16493
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4460732456803186e-05,
+      "loss": 0.2896,
+      "step": 16494
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4460396621496819e-05,
+      "loss": 0.3851,
+      "step": 16495
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4460060786190453e-05,
+      "loss": 0.1637,
+      "step": 16496
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4459724950884088e-05,
+      "loss": 0.3769,
+      "step": 16497
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4459389115577723e-05,
+      "loss": 0.4236,
+      "step": 16498
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4459053280271356e-05,
+      "loss": 0.3388,
+      "step": 16499
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4458717444964992e-05,
+      "loss": 0.24,
+      "step": 16500
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4458381609658625e-05,
+      "loss": 0.3453,
+      "step": 16501
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4458045774352258e-05,
+      "loss": 0.3223,
+      "step": 16502
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4457709939045894e-05,
+      "loss": 0.3434,
+      "step": 16503
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4457374103739527e-05,
+      "loss": 0.3654,
+      "step": 16504
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4457038268433162e-05,
+      "loss": 0.2939,
+      "step": 16505
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4456702433126796e-05,
+      "loss": 0.3325,
+      "step": 16506
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4456366597820431e-05,
+      "loss": 0.4139,
+      "step": 16507
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4456030762514064e-05,
+      "loss": 0.1843,
+      "step": 16508
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.44556949272077e-05,
+      "loss": 0.4337,
+      "step": 16509
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4455359091901333e-05,
+      "loss": 0.1695,
+      "step": 16510
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4455023256594966e-05,
+      "loss": 0.4092,
+      "step": 16511
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4454687421288603e-05,
+      "loss": 0.4307,
+      "step": 16512
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4454351585982236e-05,
+      "loss": 0.1227,
+      "step": 16513
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4454015750675869e-05,
+      "loss": 0.4024,
+      "step": 16514
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4453679915369503e-05,
+      "loss": 0.2523,
+      "step": 16515
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4453344080063138e-05,
+      "loss": 0.378,
+      "step": 16516
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4453008244756772e-05,
+      "loss": 0.3006,
+      "step": 16517
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4452672409450405e-05,
+      "loss": 0.3723,
+      "step": 16518
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4452336574144042e-05,
+      "loss": 0.2762,
+      "step": 16519
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4452000738837675e-05,
+      "loss": 0.3425,
+      "step": 16520
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4451664903531308e-05,
+      "loss": 0.3021,
+      "step": 16521
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4451329068224944e-05,
+      "loss": 0.3548,
+      "step": 16522
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4450993232918577e-05,
+      "loss": 0.2706,
+      "step": 16523
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4450657397612212e-05,
+      "loss": 0.1706,
+      "step": 16524
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4450321562305846e-05,
+      "loss": 0.2379,
+      "step": 16525
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4449985726999481e-05,
+      "loss": 0.3189,
+      "step": 16526
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4449649891693114e-05,
+      "loss": 0.2435,
+      "step": 16527
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.444931405638675e-05,
+      "loss": 0.3489,
+      "step": 16528
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4448978221080383e-05,
+      "loss": 0.178,
+      "step": 16529
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4448642385774016e-05,
+      "loss": 0.4646,
+      "step": 16530
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4448306550467652e-05,
+      "loss": 0.2346,
+      "step": 16531
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4447970715161285e-05,
+      "loss": 0.2072,
+      "step": 16532
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.444763487985492e-05,
+      "loss": 0.3924,
+      "step": 16533
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4447299044548555e-05,
+      "loss": 0.3997,
+      "step": 16534
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.444696320924219e-05,
+      "loss": 0.5288,
+      "step": 16535
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4446627373935822e-05,
+      "loss": 0.17,
+      "step": 16536
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4446291538629457e-05,
+      "loss": 0.4427,
+      "step": 16537
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4445955703323092e-05,
+      "loss": 0.2531,
+      "step": 16538
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4445619868016725e-05,
+      "loss": 0.2245,
+      "step": 16539
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4445284032710361e-05,
+      "loss": 0.2057,
+      "step": 16540
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4444948197403994e-05,
+      "loss": 0.2222,
+      "step": 16541
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4444612362097627e-05,
+      "loss": 0.2755,
+      "step": 16542
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4444276526791263e-05,
+      "loss": 0.2755,
+      "step": 16543
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4443940691484896e-05,
+      "loss": 0.3254,
+      "step": 16544
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.444360485617853e-05,
+      "loss": 0.3452,
+      "step": 16545
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4443269020872165e-05,
+      "loss": 0.2703,
+      "step": 16546
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.44429331855658e-05,
+      "loss": 0.2167,
+      "step": 16547
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4442597350259433e-05,
+      "loss": 0.2073,
+      "step": 16548
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.444226151495307e-05,
+      "loss": 0.291,
+      "step": 16549
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4441925679646702e-05,
+      "loss": 0.3827,
+      "step": 16550
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4441589844340335e-05,
+      "loss": 0.4992,
+      "step": 16551
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4441254009033972e-05,
+      "loss": 0.4099,
+      "step": 16552
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4440918173727605e-05,
+      "loss": 0.3062,
+      "step": 16553
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.444058233842124e-05,
+      "loss": 0.241,
+      "step": 16554
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4440246503114874e-05,
+      "loss": 0.2025,
+      "step": 16555
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4439910667808508e-05,
+      "loss": 0.473,
+      "step": 16556
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4439574832502141e-05,
+      "loss": 0.2961,
+      "step": 16557
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4439238997195778e-05,
+      "loss": 0.502,
+      "step": 16558
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.443890316188941e-05,
+      "loss": 0.4205,
+      "step": 16559
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4438567326583044e-05,
+      "loss": 0.2388,
+      "step": 16560
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.443823149127668e-05,
+      "loss": 0.4749,
+      "step": 16561
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4437895655970313e-05,
+      "loss": 0.2051,
+      "step": 16562
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4437559820663946e-05,
+      "loss": 0.4168,
+      "step": 16563
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4437223985357582e-05,
+      "loss": 0.4656,
+      "step": 16564
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4436888150051215e-05,
+      "loss": 0.2597,
+      "step": 16565
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.443655231474485e-05,
+      "loss": 0.2395,
+      "step": 16566
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4436216479438485e-05,
+      "loss": 0.3474,
+      "step": 16567
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.443588064413212e-05,
+      "loss": 0.3068,
+      "step": 16568
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4435544808825752e-05,
+      "loss": 0.2932,
+      "step": 16569
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4435208973519388e-05,
+      "loss": 0.3107,
+      "step": 16570
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4434873138213021e-05,
+      "loss": 0.237,
+      "step": 16571
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4434537302906654e-05,
+      "loss": 0.4667,
+      "step": 16572
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.443420146760029e-05,
+      "loss": 0.3006,
+      "step": 16573
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4433865632293924e-05,
+      "loss": 0.2486,
+      "step": 16574
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.4433529796987558e-05,
+      "loss": 0.2535,
+      "step": 16575
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4433193961681193e-05,
+      "loss": 0.3458,
+      "step": 16576
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4432858126374828e-05,
+      "loss": 0.236,
+      "step": 16577
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.443252229106846e-05,
+      "loss": 0.2498,
+      "step": 16578
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4432186455762097e-05,
+      "loss": 0.2808,
+      "step": 16579
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.443185062045573e-05,
+      "loss": 0.2535,
+      "step": 16580
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4431514785149363e-05,
+      "loss": 0.3132,
+      "step": 16581
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4431178949842999e-05,
+      "loss": 0.2711,
+      "step": 16582
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4430843114536632e-05,
+      "loss": 0.3082,
+      "step": 16583
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4430507279230265e-05,
+      "loss": 0.4594,
+      "step": 16584
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4430171443923901e-05,
+      "loss": 0.2537,
+      "step": 16585
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4429835608617534e-05,
+      "loss": 0.3206,
+      "step": 16586
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4429499773311169e-05,
+      "loss": 0.3534,
+      "step": 16587
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4429163938004804e-05,
+      "loss": 0.2933,
+      "step": 16588
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4428828102698438e-05,
+      "loss": 0.2539,
+      "step": 16589
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4428492267392071e-05,
+      "loss": 0.2649,
+      "step": 16590
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4428156432085708e-05,
+      "loss": 0.1785,
+      "step": 16591
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.442782059677934e-05,
+      "loss": 0.374,
+      "step": 16592
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4427484761472973e-05,
+      "loss": 0.2794,
+      "step": 16593
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.442714892616661e-05,
+      "loss": 0.2545,
+      "step": 16594
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4426813090860243e-05,
+      "loss": 0.2456,
+      "step": 16595
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4426477255553877e-05,
+      "loss": 0.197,
+      "step": 16596
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4426141420247512e-05,
+      "loss": 0.2652,
+      "step": 16597
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4425805584941147e-05,
+      "loss": 0.3281,
+      "step": 16598
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.442546974963478e-05,
+      "loss": 0.203,
+      "step": 16599
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4425133914328416e-05,
+      "loss": 0.3025,
+      "step": 16600
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4424798079022049e-05,
+      "loss": 0.1865,
+      "step": 16601
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4424462243715682e-05,
+      "loss": 0.3098,
+      "step": 16602
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4424126408409318e-05,
+      "loss": 0.409,
+      "step": 16603
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4423790573102951e-05,
+      "loss": 0.4285,
+      "step": 16604
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4423454737796584e-05,
+      "loss": 0.2881,
+      "step": 16605
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.442311890249022e-05,
+      "loss": 0.3841,
+      "step": 16606
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4422783067183853e-05,
+      "loss": 0.2423,
+      "step": 16607
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4422447231877488e-05,
+      "loss": 0.2482,
+      "step": 16608
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4422111396571123e-05,
+      "loss": 0.2902,
+      "step": 16609
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4421775561264757e-05,
+      "loss": 0.3372,
+      "step": 16610
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.442143972595839e-05,
+      "loss": 0.4162,
+      "step": 16611
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4421103890652027e-05,
+      "loss": 0.19,
+      "step": 16612
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.442076805534566e-05,
+      "loss": 0.3343,
+      "step": 16613
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4420432220039293e-05,
+      "loss": 0.1415,
+      "step": 16614
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4420096384732929e-05,
+      "loss": 0.3201,
+      "step": 16615
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4419760549426562e-05,
+      "loss": 0.3135,
+      "step": 16616
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4419424714120197e-05,
+      "loss": 0.4066,
+      "step": 16617
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4419088878813831e-05,
+      "loss": 0.3838,
+      "step": 16618
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4418753043507466e-05,
+      "loss": 0.163,
+      "step": 16619
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4418417208201099e-05,
+      "loss": 0.1645,
+      "step": 16620
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4418081372894735e-05,
+      "loss": 0.2178,
+      "step": 16621
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4417745537588368e-05,
+      "loss": 0.2284,
+      "step": 16622
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4417409702282001e-05,
+      "loss": 0.2302,
+      "step": 16623
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4417073866975637e-05,
+      "loss": 0.322,
+      "step": 16624
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.441673803166927e-05,
+      "loss": 0.4521,
+      "step": 16625
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4416402196362905e-05,
+      "loss": 0.5739,
+      "step": 16626
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.441606636105654e-05,
+      "loss": 0.2923,
+      "step": 16627
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4415730525750173e-05,
+      "loss": 0.226,
+      "step": 16628
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4415394690443807e-05,
+      "loss": 0.4012,
+      "step": 16629
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4415058855137442e-05,
+      "loss": 0.1299,
+      "step": 16630
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4414723019831077e-05,
+      "loss": 0.3016,
+      "step": 16631
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.441438718452471e-05,
+      "loss": 0.2823,
+      "step": 16632
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4414051349218346e-05,
+      "loss": 0.251,
+      "step": 16633
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4413715513911979e-05,
+      "loss": 0.5066,
+      "step": 16634
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4413379678605612e-05,
+      "loss": 0.4277,
+      "step": 16635
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4413043843299248e-05,
+      "loss": 0.2096,
+      "step": 16636
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4412708007992881e-05,
+      "loss": 0.2766,
+      "step": 16637
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4412372172686516e-05,
+      "loss": 0.2231,
+      "step": 16638
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.441203633738015e-05,
+      "loss": 0.2343,
+      "step": 16639
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4411700502073785e-05,
+      "loss": 0.4999,
+      "step": 16640
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4411364666767418e-05,
+      "loss": 0.3307,
+      "step": 16641
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4411028831461054e-05,
+      "loss": 0.169,
+      "step": 16642
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4410692996154687e-05,
+      "loss": 0.1828,
+      "step": 16643
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.441035716084832e-05,
+      "loss": 0.3507,
+      "step": 16644
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4410021325541957e-05,
+      "loss": 0.2677,
+      "step": 16645
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.440968549023559e-05,
+      "loss": 0.3291,
+      "step": 16646
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4409349654929224e-05,
+      "loss": 0.2707,
+      "step": 16647
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4409013819622859e-05,
+      "loss": 0.2202,
+      "step": 16648
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4408677984316492e-05,
+      "loss": 0.2864,
+      "step": 16649
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4408342149010126e-05,
+      "loss": 0.4337,
+      "step": 16650
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4408006313703761e-05,
+      "loss": 0.2469,
+      "step": 16651
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4407670478397396e-05,
+      "loss": 0.5076,
+      "step": 16652
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4407334643091029e-05,
+      "loss": 0.4138,
+      "step": 16653
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4406998807784665e-05,
+      "loss": 0.3895,
+      "step": 16654
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4406662972478298e-05,
+      "loss": 0.3866,
+      "step": 16655
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4406327137171931e-05,
+      "loss": 0.3471,
+      "step": 16656
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4405991301865567e-05,
+      "loss": 0.255,
+      "step": 16657
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.44056554665592e-05,
+      "loss": 0.2972,
+      "step": 16658
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4405319631252835e-05,
+      "loss": 0.3788,
+      "step": 16659
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.440498379594647e-05,
+      "loss": 0.2316,
+      "step": 16660
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4404647960640104e-05,
+      "loss": 0.1768,
+      "step": 16661
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4404312125333737e-05,
+      "loss": 0.3204,
+      "step": 16662
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4403976290027373e-05,
+      "loss": 0.3604,
+      "step": 16663
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4403640454721006e-05,
+      "loss": 0.3151,
+      "step": 16664
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.440330461941464e-05,
+      "loss": 0.2733,
+      "step": 16665
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4402968784108276e-05,
+      "loss": 0.3163,
+      "step": 16666
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4402632948801909e-05,
+      "loss": 0.4341,
+      "step": 16667
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4402297113495543e-05,
+      "loss": 0.2897,
+      "step": 16668
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4401961278189178e-05,
+      "loss": 0.1739,
+      "step": 16669
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4401625442882813e-05,
+      "loss": 0.3524,
+      "step": 16670
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4401289607576445e-05,
+      "loss": 0.4462,
+      "step": 16671
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.440095377227008e-05,
+      "loss": 0.3516,
+      "step": 16672
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4400617936963715e-05,
+      "loss": 0.3626,
+      "step": 16673
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4400282101657348e-05,
+      "loss": 0.2382,
+      "step": 16674
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4399946266350984e-05,
+      "loss": 0.3443,
+      "step": 16675
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4399610431044617e-05,
+      "loss": 0.2108,
+      "step": 16676
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.439927459573825e-05,
+      "loss": 0.2178,
+      "step": 16677
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4398938760431886e-05,
+      "loss": 0.3392,
+      "step": 16678
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.439860292512552e-05,
+      "loss": 0.261,
+      "step": 16679
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4398267089819154e-05,
+      "loss": 0.4281,
+      "step": 16680
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4397931254512789e-05,
+      "loss": 0.2687,
+      "step": 16681
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4397595419206423e-05,
+      "loss": 0.3338,
+      "step": 16682
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4397259583900056e-05,
+      "loss": 0.4133,
+      "step": 16683
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4396923748593693e-05,
+      "loss": 0.4146,
+      "step": 16684
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4396587913287325e-05,
+      "loss": 0.2649,
+      "step": 16685
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4396252077980958e-05,
+      "loss": 0.2899,
+      "step": 16686
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4395916242674595e-05,
+      "loss": 0.3463,
+      "step": 16687
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4395580407368228e-05,
+      "loss": 0.3307,
+      "step": 16688
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4395244572061862e-05,
+      "loss": 0.2297,
+      "step": 16689
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4394908736755497e-05,
+      "loss": 0.4543,
+      "step": 16690
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4394572901449132e-05,
+      "loss": 0.3248,
+      "step": 16691
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4394237066142765e-05,
+      "loss": 0.449,
+      "step": 16692
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4393901230836398e-05,
+      "loss": 0.3414,
+      "step": 16693
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4393565395530034e-05,
+      "loss": 0.146,
+      "step": 16694
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4393229560223667e-05,
+      "loss": 0.2962,
+      "step": 16695
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.43928937249173e-05,
+      "loss": 0.2978,
+      "step": 16696
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4392557889610936e-05,
+      "loss": 0.569,
+      "step": 16697
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4392222054304569e-05,
+      "loss": 0.3495,
+      "step": 16698
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4391886218998204e-05,
+      "loss": 0.3435,
+      "step": 16699
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4391550383691838e-05,
+      "loss": 0.4127,
+      "step": 16700
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4391214548385473e-05,
+      "loss": 0.5322,
+      "step": 16701
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4390878713079106e-05,
+      "loss": 0.2339,
+      "step": 16702
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4390542877772742e-05,
+      "loss": 0.2291,
+      "step": 16703
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4390207042466375e-05,
+      "loss": 0.3241,
+      "step": 16704
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4389871207160008e-05,
+      "loss": 0.5624,
+      "step": 16705
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4389535371853645e-05,
+      "loss": 0.2722,
+      "step": 16706
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4389199536547278e-05,
+      "loss": 0.3292,
+      "step": 16707
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4388863701240912e-05,
+      "loss": 0.1517,
+      "step": 16708
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4388527865934547e-05,
+      "loss": 0.2552,
+      "step": 16709
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4388192030628181e-05,
+      "loss": 0.3921,
+      "step": 16710
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4387856195321814e-05,
+      "loss": 0.1621,
+      "step": 16711
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.438752036001545e-05,
+      "loss": 0.3914,
+      "step": 16712
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4387184524709084e-05,
+      "loss": 0.2634,
+      "step": 16713
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4386848689402717e-05,
+      "loss": 0.2576,
+      "step": 16714
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4386512854096353e-05,
+      "loss": 0.2891,
+      "step": 16715
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4386177018789986e-05,
+      "loss": 0.4695,
+      "step": 16716
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.438584118348362e-05,
+      "loss": 0.5312,
+      "step": 16717
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4385505348177255e-05,
+      "loss": 0.1877,
+      "step": 16718
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4385169512870888e-05,
+      "loss": 0.2795,
+      "step": 16719
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4384833677564523e-05,
+      "loss": 0.3399,
+      "step": 16720
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4384497842258158e-05,
+      "loss": 0.3547,
+      "step": 16721
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4384162006951792e-05,
+      "loss": 0.2011,
+      "step": 16722
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4383826171645425e-05,
+      "loss": 0.3939,
+      "step": 16723
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4383490336339061e-05,
+      "loss": 0.2731,
+      "step": 16724
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4383154501032694e-05,
+      "loss": 0.4374,
+      "step": 16725
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4382818665726327e-05,
+      "loss": 0.2683,
+      "step": 16726
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4382482830419964e-05,
+      "loss": 0.44,
+      "step": 16727
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4382146995113597e-05,
+      "loss": 0.4159,
+      "step": 16728
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4381811159807231e-05,
+      "loss": 0.2464,
+      "step": 16729
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4381475324500866e-05,
+      "loss": 0.3057,
+      "step": 16730
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.43811394891945e-05,
+      "loss": 0.3081,
+      "step": 16731
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4380803653888134e-05,
+      "loss": 0.3149,
+      "step": 16732
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.438046781858177e-05,
+      "loss": 0.3354,
+      "step": 16733
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4380131983275403e-05,
+      "loss": 0.3524,
+      "step": 16734
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4379796147969036e-05,
+      "loss": 0.2924,
+      "step": 16735
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4379460312662672e-05,
+      "loss": 0.232,
+      "step": 16736
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4379124477356305e-05,
+      "loss": 0.3834,
+      "step": 16737
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.437878864204994e-05,
+      "loss": 0.2258,
+      "step": 16738
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4378452806743574e-05,
+      "loss": 0.2777,
+      "step": 16739
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4378116971437207e-05,
+      "loss": 0.3056,
+      "step": 16740
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4377781136130842e-05,
+      "loss": 0.2271,
+      "step": 16741
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4377445300824477e-05,
+      "loss": 0.3387,
+      "step": 16742
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4377109465518111e-05,
+      "loss": 0.2917,
+      "step": 16743
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4376773630211744e-05,
+      "loss": 0.2098,
+      "step": 16744
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.437643779490538e-05,
+      "loss": 0.3402,
+      "step": 16745
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4376101959599014e-05,
+      "loss": 0.173,
+      "step": 16746
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4375766124292647e-05,
+      "loss": 0.2652,
+      "step": 16747
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4375430288986283e-05,
+      "loss": 0.3214,
+      "step": 16748
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4375094453679916e-05,
+      "loss": 0.3476,
+      "step": 16749
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.437475861837355e-05,
+      "loss": 0.2782,
+      "step": 16750
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4374422783067185e-05,
+      "loss": 0.2654,
+      "step": 16751
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.437408694776082e-05,
+      "loss": 0.4424,
+      "step": 16752
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4373751112454453e-05,
+      "loss": 0.1522,
+      "step": 16753
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4373415277148089e-05,
+      "loss": 0.2665,
+      "step": 16754
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4373079441841722e-05,
+      "loss": 0.3458,
+      "step": 16755
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4372743606535355e-05,
+      "loss": 0.3478,
+      "step": 16756
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4372407771228991e-05,
+      "loss": 0.2794,
+      "step": 16757
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4372071935922624e-05,
+      "loss": 0.4828,
+      "step": 16758
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4371736100616259e-05,
+      "loss": 0.4023,
+      "step": 16759
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4371400265309894e-05,
+      "loss": 0.35,
+      "step": 16760
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4371064430003528e-05,
+      "loss": 0.1879,
+      "step": 16761
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4370728594697161e-05,
+      "loss": 0.4176,
+      "step": 16762
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4370392759390796e-05,
+      "loss": 0.2548,
+      "step": 16763
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.437005692408443e-05,
+      "loss": 0.2269,
+      "step": 16764
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4369721088778063e-05,
+      "loss": 0.2342,
+      "step": 16765
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.43693852534717e-05,
+      "loss": 0.3126,
+      "step": 16766
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4369049418165333e-05,
+      "loss": 0.2128,
+      "step": 16767
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4368713582858966e-05,
+      "loss": 0.2833,
+      "step": 16768
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4368377747552602e-05,
+      "loss": 0.2782,
+      "step": 16769
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4368041912246235e-05,
+      "loss": 0.5948,
+      "step": 16770
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.436770607693987e-05,
+      "loss": 0.3691,
+      "step": 16771
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4367370241633504e-05,
+      "loss": 0.1995,
+      "step": 16772
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4367034406327139e-05,
+      "loss": 0.281,
+      "step": 16773
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.4366698571020772e-05,
+      "loss": 0.6861,
+      "step": 16774
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4366362735714408e-05,
+      "loss": 0.3852,
+      "step": 16775
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4366026900408041e-05,
+      "loss": 0.2917,
+      "step": 16776
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4365691065101674e-05,
+      "loss": 0.263,
+      "step": 16777
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.436535522979531e-05,
+      "loss": 0.2346,
+      "step": 16778
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4365019394488943e-05,
+      "loss": 0.3228,
+      "step": 16779
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4364683559182578e-05,
+      "loss": 0.3238,
+      "step": 16780
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4364347723876213e-05,
+      "loss": 0.2308,
+      "step": 16781
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4364011888569847e-05,
+      "loss": 0.1227,
+      "step": 16782
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.436367605326348e-05,
+      "loss": 0.4559,
+      "step": 16783
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4363340217957117e-05,
+      "loss": 0.3071,
+      "step": 16784
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.436300438265075e-05,
+      "loss": 0.3364,
+      "step": 16785
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4362668547344383e-05,
+      "loss": 0.3226,
+      "step": 16786
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4362332712038019e-05,
+      "loss": 0.4963,
+      "step": 16787
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4361996876731652e-05,
+      "loss": 0.3424,
+      "step": 16788
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4361661041425285e-05,
+      "loss": 0.3298,
+      "step": 16789
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4361325206118921e-05,
+      "loss": 0.1387,
+      "step": 16790
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4360989370812554e-05,
+      "loss": 0.3993,
+      "step": 16791
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4360653535506189e-05,
+      "loss": 0.1944,
+      "step": 16792
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4360317700199823e-05,
+      "loss": 0.3069,
+      "step": 16793
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4359981864893458e-05,
+      "loss": 0.2229,
+      "step": 16794
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4359646029587091e-05,
+      "loss": 0.3041,
+      "step": 16795
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4359310194280727e-05,
+      "loss": 0.4388,
+      "step": 16796
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.435897435897436e-05,
+      "loss": 0.255,
+      "step": 16797
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4358638523667993e-05,
+      "loss": 0.4397,
+      "step": 16798
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.435830268836163e-05,
+      "loss": 0.482,
+      "step": 16799
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4357966853055262e-05,
+      "loss": 0.3279,
+      "step": 16800
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4357631017748897e-05,
+      "loss": 0.1697,
+      "step": 16801
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4357295182442532e-05,
+      "loss": 0.3053,
+      "step": 16802
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4356959347136166e-05,
+      "loss": 0.4006,
+      "step": 16803
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.43566235118298e-05,
+      "loss": 0.4373,
+      "step": 16804
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4356287676523436e-05,
+      "loss": 0.3445,
+      "step": 16805
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4355951841217069e-05,
+      "loss": 0.2448,
+      "step": 16806
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4355616005910702e-05,
+      "loss": 0.3284,
+      "step": 16807
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4355280170604338e-05,
+      "loss": 0.4731,
+      "step": 16808
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4354944335297971e-05,
+      "loss": 0.4196,
+      "step": 16809
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4354608499991604e-05,
+      "loss": 0.2916,
+      "step": 16810
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.435427266468524e-05,
+      "loss": 0.169,
+      "step": 16811
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4353936829378873e-05,
+      "loss": 0.3255,
+      "step": 16812
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4353600994072508e-05,
+      "loss": 0.4299,
+      "step": 16813
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4353265158766142e-05,
+      "loss": 0.3195,
+      "step": 16814
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4352929323459777e-05,
+      "loss": 0.3654,
+      "step": 16815
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.435259348815341e-05,
+      "loss": 0.1374,
+      "step": 16816
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4352257652847046e-05,
+      "loss": 0.3944,
+      "step": 16817
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.435192181754068e-05,
+      "loss": 0.3204,
+      "step": 16818
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4351585982234312e-05,
+      "loss": 0.3766,
+      "step": 16819
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4351250146927949e-05,
+      "loss": 0.4628,
+      "step": 16820
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4350914311621582e-05,
+      "loss": 0.2398,
+      "step": 16821
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4350578476315216e-05,
+      "loss": 0.2438,
+      "step": 16822
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4350242641008851e-05,
+      "loss": 0.2194,
+      "step": 16823
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4349906805702486e-05,
+      "loss": 0.3373,
+      "step": 16824
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4349570970396119e-05,
+      "loss": 0.228,
+      "step": 16825
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4349235135089755e-05,
+      "loss": 0.2055,
+      "step": 16826
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4348899299783388e-05,
+      "loss": 0.2496,
+      "step": 16827
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.434856346447702e-05,
+      "loss": 0.2797,
+      "step": 16828
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4348227629170657e-05,
+      "loss": 0.2328,
+      "step": 16829
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.434789179386429e-05,
+      "loss": 0.3652,
+      "step": 16830
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4347555958557923e-05,
+      "loss": 0.3102,
+      "step": 16831
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.434722012325156e-05,
+      "loss": 0.3582,
+      "step": 16832
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4346884287945192e-05,
+      "loss": 0.4277,
+      "step": 16833
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4346548452638827e-05,
+      "loss": 0.3075,
+      "step": 16834
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4346212617332462e-05,
+      "loss": 0.2769,
+      "step": 16835
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4345876782026096e-05,
+      "loss": 0.2934,
+      "step": 16836
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.434554094671973e-05,
+      "loss": 0.5808,
+      "step": 16837
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4345205111413366e-05,
+      "loss": 0.3,
+      "step": 16838
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4344869276106998e-05,
+      "loss": 0.2717,
+      "step": 16839
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4344533440800631e-05,
+      "loss": 0.3116,
+      "step": 16840
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4344197605494268e-05,
+      "loss": 0.4148,
+      "step": 16841
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.43438617701879e-05,
+      "loss": 0.2561,
+      "step": 16842
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4343525934881535e-05,
+      "loss": 0.3936,
+      "step": 16843
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.434319009957517e-05,
+      "loss": 0.3576,
+      "step": 16844
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4342854264268805e-05,
+      "loss": 0.3216,
+      "step": 16845
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4342518428962438e-05,
+      "loss": 0.2031,
+      "step": 16846
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4342182593656074e-05,
+      "loss": 0.3923,
+      "step": 16847
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4341846758349707e-05,
+      "loss": 0.4146,
+      "step": 16848
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.434151092304334e-05,
+      "loss": 0.1908,
+      "step": 16849
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4341175087736976e-05,
+      "loss": 0.3482,
+      "step": 16850
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.434083925243061e-05,
+      "loss": 0.4452,
+      "step": 16851
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4340503417124244e-05,
+      "loss": 0.4765,
+      "step": 16852
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4340167581817878e-05,
+      "loss": 0.3106,
+      "step": 16853
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4339831746511511e-05,
+      "loss": 0.3067,
+      "step": 16854
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4339495911205146e-05,
+      "loss": 0.332,
+      "step": 16855
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.433916007589878e-05,
+      "loss": 0.3382,
+      "step": 16856
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4338824240592415e-05,
+      "loss": 0.3364,
+      "step": 16857
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4338488405286048e-05,
+      "loss": 0.2238,
+      "step": 16858
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4338152569979685e-05,
+      "loss": 0.2098,
+      "step": 16859
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4337816734673318e-05,
+      "loss": 0.4094,
+      "step": 16860
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.433748089936695e-05,
+      "loss": 0.2887,
+      "step": 16861
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4337145064060587e-05,
+      "loss": 0.3969,
+      "step": 16862
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.433680922875422e-05,
+      "loss": 0.2627,
+      "step": 16863
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4336473393447855e-05,
+      "loss": 0.3389,
+      "step": 16864
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.433613755814149e-05,
+      "loss": 0.4686,
+      "step": 16865
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4335801722835124e-05,
+      "loss": 0.4576,
+      "step": 16866
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4335465887528757e-05,
+      "loss": 0.344,
+      "step": 16867
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4335130052222393e-05,
+      "loss": 0.2941,
+      "step": 16868
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4334794216916026e-05,
+      "loss": 0.3538,
+      "step": 16869
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4334458381609659e-05,
+      "loss": 0.3389,
+      "step": 16870
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4334122546303295e-05,
+      "loss": 0.2401,
+      "step": 16871
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4333786710996928e-05,
+      "loss": 0.3473,
+      "step": 16872
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4333450875690563e-05,
+      "loss": 0.4061,
+      "step": 16873
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4333115040384196e-05,
+      "loss": 0.3241,
+      "step": 16874
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.433277920507783e-05,
+      "loss": 0.4814,
+      "step": 16875
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4332443369771465e-05,
+      "loss": 0.5507,
+      "step": 16876
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4332107534465098e-05,
+      "loss": 0.2379,
+      "step": 16877
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4331771699158734e-05,
+      "loss": 0.3567,
+      "step": 16878
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4331435863852367e-05,
+      "loss": 0.1862,
+      "step": 16879
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4331100028546e-05,
+      "loss": 0.2335,
+      "step": 16880
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4330764193239637e-05,
+      "loss": 0.2347,
+      "step": 16881
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.433042835793327e-05,
+      "loss": 0.3594,
+      "step": 16882
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4330092522626904e-05,
+      "loss": 0.2222,
+      "step": 16883
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4329756687320539e-05,
+      "loss": 0.2816,
+      "step": 16884
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4329420852014174e-05,
+      "loss": 0.4022,
+      "step": 16885
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4329085016707807e-05,
+      "loss": 0.339,
+      "step": 16886
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4328749181401443e-05,
+      "loss": 0.5035,
+      "step": 16887
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4328413346095076e-05,
+      "loss": 0.2511,
+      "step": 16888
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4328077510788709e-05,
+      "loss": 0.2053,
+      "step": 16889
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4327741675482345e-05,
+      "loss": 0.2687,
+      "step": 16890
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4327405840175978e-05,
+      "loss": 0.3349,
+      "step": 16891
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4327070004869613e-05,
+      "loss": 0.1811,
+      "step": 16892
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4326734169563247e-05,
+      "loss": 0.2953,
+      "step": 16893
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4326398334256882e-05,
+      "loss": 0.2014,
+      "step": 16894
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4326062498950515e-05,
+      "loss": 0.458,
+      "step": 16895
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4325726663644151e-05,
+      "loss": 0.3282,
+      "step": 16896
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4325390828337784e-05,
+      "loss": 0.3033,
+      "step": 16897
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4325054993031417e-05,
+      "loss": 0.2593,
+      "step": 16898
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4324719157725054e-05,
+      "loss": 0.411,
+      "step": 16899
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4324383322418687e-05,
+      "loss": 0.2135,
+      "step": 16900
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.432404748711232e-05,
+      "loss": 0.3602,
+      "step": 16901
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4323711651805956e-05,
+      "loss": 0.3404,
+      "step": 16902
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4323375816499589e-05,
+      "loss": 0.2557,
+      "step": 16903
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4323039981193223e-05,
+      "loss": 0.317,
+      "step": 16904
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4322704145886858e-05,
+      "loss": 0.2988,
+      "step": 16905
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4322368310580493e-05,
+      "loss": 0.3766,
+      "step": 16906
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4322032475274126e-05,
+      "loss": 0.3508,
+      "step": 16907
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4321696639967762e-05,
+      "loss": 0.3858,
+      "step": 16908
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4321360804661395e-05,
+      "loss": 0.202,
+      "step": 16909
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4321024969355028e-05,
+      "loss": 0.3437,
+      "step": 16910
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4320689134048664e-05,
+      "loss": 0.3675,
+      "step": 16911
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4320353298742297e-05,
+      "loss": 0.2866,
+      "step": 16912
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4320017463435932e-05,
+      "loss": 0.3524,
+      "step": 16913
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4319681628129567e-05,
+      "loss": 0.3647,
+      "step": 16914
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4319345792823201e-05,
+      "loss": 0.2475,
+      "step": 16915
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4319009957516834e-05,
+      "loss": 0.2538,
+      "step": 16916
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.431867412221047e-05,
+      "loss": 0.3157,
+      "step": 16917
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4318338286904103e-05,
+      "loss": 0.3241,
+      "step": 16918
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4318002451597736e-05,
+      "loss": 0.4153,
+      "step": 16919
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4317666616291373e-05,
+      "loss": 0.2756,
+      "step": 16920
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4317330780985006e-05,
+      "loss": 0.326,
+      "step": 16921
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4316994945678639e-05,
+      "loss": 0.3427,
+      "step": 16922
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4316659110372275e-05,
+      "loss": 0.3544,
+      "step": 16923
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4316323275065908e-05,
+      "loss": 0.3226,
+      "step": 16924
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4315987439759543e-05,
+      "loss": 0.3227,
+      "step": 16925
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4315651604453177e-05,
+      "loss": 0.2003,
+      "step": 16926
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4315315769146812e-05,
+      "loss": 0.4502,
+      "step": 16927
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4314979933840445e-05,
+      "loss": 0.366,
+      "step": 16928
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4314644098534081e-05,
+      "loss": 0.431,
+      "step": 16929
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4314308263227714e-05,
+      "loss": 0.3604,
+      "step": 16930
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4313972427921347e-05,
+      "loss": 0.4127,
+      "step": 16931
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4313636592614983e-05,
+      "loss": 0.2294,
+      "step": 16932
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4313300757308616e-05,
+      "loss": 0.2933,
+      "step": 16933
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4312964922002251e-05,
+      "loss": 0.4487,
+      "step": 16934
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4312629086695886e-05,
+      "loss": 0.3966,
+      "step": 16935
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.431229325138952e-05,
+      "loss": 0.2424,
+      "step": 16936
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4311957416083153e-05,
+      "loss": 0.3888,
+      "step": 16937
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.431162158077679e-05,
+      "loss": 0.4492,
+      "step": 16938
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4311285745470423e-05,
+      "loss": 0.2569,
+      "step": 16939
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4310949910164056e-05,
+      "loss": 0.2498,
+      "step": 16940
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4310614074857692e-05,
+      "loss": 0.3369,
+      "step": 16941
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4310278239551325e-05,
+      "loss": 0.4932,
+      "step": 16942
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.430994240424496e-05,
+      "loss": 0.2741,
+      "step": 16943
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4309606568938594e-05,
+      "loss": 0.265,
+      "step": 16944
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4309270733632227e-05,
+      "loss": 0.3682,
+      "step": 16945
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4308934898325862e-05,
+      "loss": 0.2164,
+      "step": 16946
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4308599063019496e-05,
+      "loss": 0.3385,
+      "step": 16947
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4308263227713131e-05,
+      "loss": 0.3725,
+      "step": 16948
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4307927392406764e-05,
+      "loss": 0.3627,
+      "step": 16949
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.43075915571004e-05,
+      "loss": 0.4049,
+      "step": 16950
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4307255721794033e-05,
+      "loss": 0.3952,
+      "step": 16951
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4306919886487666e-05,
+      "loss": 0.2668,
+      "step": 16952
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4306584051181303e-05,
+      "loss": 0.2723,
+      "step": 16953
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4306248215874936e-05,
+      "loss": 0.1896,
+      "step": 16954
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.430591238056857e-05,
+      "loss": 0.3078,
+      "step": 16955
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4305576545262205e-05,
+      "loss": 0.3487,
+      "step": 16956
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.430524070995584e-05,
+      "loss": 0.3511,
+      "step": 16957
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4304904874649472e-05,
+      "loss": 0.2293,
+      "step": 16958
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4304569039343109e-05,
+      "loss": 0.3118,
+      "step": 16959
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4304233204036742e-05,
+      "loss": 0.3263,
+      "step": 16960
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4303897368730375e-05,
+      "loss": 0.5008,
+      "step": 16961
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4303561533424011e-05,
+      "loss": 0.2255,
+      "step": 16962
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4303225698117644e-05,
+      "loss": 0.5369,
+      "step": 16963
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4302889862811279e-05,
+      "loss": 0.2838,
+      "step": 16964
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4302554027504913e-05,
+      "loss": 0.2966,
+      "step": 16965
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4302218192198546e-05,
+      "loss": 0.2705,
+      "step": 16966
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.430188235689218e-05,
+      "loss": 0.3491,
+      "step": 16967
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4301546521585815e-05,
+      "loss": 0.3595,
+      "step": 16968
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.430121068627945e-05,
+      "loss": 0.2778,
+      "step": 16969
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4300874850973083e-05,
+      "loss": 0.3927,
+      "step": 16970
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.430053901566672e-05,
+      "loss": 0.1989,
+      "step": 16971
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.4300203180360352e-05,
+      "loss": 0.269,
+      "step": 16972
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4299867345053985e-05,
+      "loss": 0.4568,
+      "step": 16973
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4299531509747622e-05,
+      "loss": 0.5348,
+      "step": 16974
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4299195674441255e-05,
+      "loss": 0.4224,
+      "step": 16975
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.429885983913489e-05,
+      "loss": 0.4398,
+      "step": 16976
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4298524003828524e-05,
+      "loss": 0.2134,
+      "step": 16977
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4298188168522159e-05,
+      "loss": 0.287,
+      "step": 16978
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4297852333215792e-05,
+      "loss": 0.3511,
+      "step": 16979
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4297516497909428e-05,
+      "loss": 0.2837,
+      "step": 16980
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.429718066260306e-05,
+      "loss": 0.3441,
+      "step": 16981
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4296844827296694e-05,
+      "loss": 0.4044,
+      "step": 16982
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.429650899199033e-05,
+      "loss": 0.3668,
+      "step": 16983
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4296173156683963e-05,
+      "loss": 0.3397,
+      "step": 16984
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4295837321377598e-05,
+      "loss": 0.2526,
+      "step": 16985
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4295501486071232e-05,
+      "loss": 0.2753,
+      "step": 16986
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4295165650764867e-05,
+      "loss": 0.238,
+      "step": 16987
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.42948298154585e-05,
+      "loss": 0.2664,
+      "step": 16988
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4294493980152135e-05,
+      "loss": 0.2706,
+      "step": 16989
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.429415814484577e-05,
+      "loss": 0.4019,
+      "step": 16990
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4293822309539402e-05,
+      "loss": 0.2333,
+      "step": 16991
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4293486474233039e-05,
+      "loss": 0.2325,
+      "step": 16992
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4293150638926672e-05,
+      "loss": 0.2714,
+      "step": 16993
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4292814803620304e-05,
+      "loss": 0.1843,
+      "step": 16994
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.429247896831394e-05,
+      "loss": 0.3421,
+      "step": 16995
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4292143133007574e-05,
+      "loss": 0.4335,
+      "step": 16996
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4291807297701208e-05,
+      "loss": 0.3197,
+      "step": 16997
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4291471462394843e-05,
+      "loss": 0.4902,
+      "step": 16998
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4291135627088478e-05,
+      "loss": 0.2491,
+      "step": 16999
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.429079979178211e-05,
+      "loss": 0.2224,
+      "step": 17000
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4290463956475747e-05,
+      "loss": 0.3367,
+      "step": 17001
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.429012812116938e-05,
+      "loss": 0.3263,
+      "step": 17002
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4289792285863013e-05,
+      "loss": 0.4021,
+      "step": 17003
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.428945645055665e-05,
+      "loss": 0.2598,
+      "step": 17004
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4289120615250282e-05,
+      "loss": 0.3154,
+      "step": 17005
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4288784779943917e-05,
+      "loss": 0.295,
+      "step": 17006
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4288448944637551e-05,
+      "loss": 0.3832,
+      "step": 17007
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4288113109331186e-05,
+      "loss": 0.2853,
+      "step": 17008
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4287777274024819e-05,
+      "loss": 0.263,
+      "step": 17009
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4287441438718454e-05,
+      "loss": 0.2586,
+      "step": 17010
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4287105603412088e-05,
+      "loss": 0.341,
+      "step": 17011
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4286769768105721e-05,
+      "loss": 0.2991,
+      "step": 17012
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4286433932799358e-05,
+      "loss": 0.3086,
+      "step": 17013
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.428609809749299e-05,
+      "loss": 0.2474,
+      "step": 17014
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4285762262186624e-05,
+      "loss": 0.2452,
+      "step": 17015
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.428542642688026e-05,
+      "loss": 0.35,
+      "step": 17016
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4285090591573893e-05,
+      "loss": 0.315,
+      "step": 17017
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4284754756267528e-05,
+      "loss": 0.223,
+      "step": 17018
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4284418920961162e-05,
+      "loss": 0.122,
+      "step": 17019
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4284083085654797e-05,
+      "loss": 0.3484,
+      "step": 17020
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.428374725034843e-05,
+      "loss": 0.3693,
+      "step": 17021
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4283411415042066e-05,
+      "loss": 0.2336,
+      "step": 17022
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4283075579735699e-05,
+      "loss": 0.2099,
+      "step": 17023
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4282739744429332e-05,
+      "loss": 0.1841,
+      "step": 17024
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4282403909122968e-05,
+      "loss": 0.2035,
+      "step": 17025
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4282068073816601e-05,
+      "loss": 0.2213,
+      "step": 17026
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4281732238510236e-05,
+      "loss": 0.2322,
+      "step": 17027
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.428139640320387e-05,
+      "loss": 0.4376,
+      "step": 17028
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4281060567897505e-05,
+      "loss": 0.3445,
+      "step": 17029
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4280724732591138e-05,
+      "loss": 0.4858,
+      "step": 17030
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4280388897284775e-05,
+      "loss": 0.3507,
+      "step": 17031
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4280053061978408e-05,
+      "loss": 0.2815,
+      "step": 17032
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.427971722667204e-05,
+      "loss": 0.36,
+      "step": 17033
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4279381391365677e-05,
+      "loss": 0.4103,
+      "step": 17034
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.427904555605931e-05,
+      "loss": 0.212,
+      "step": 17035
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4278709720752943e-05,
+      "loss": 0.3244,
+      "step": 17036
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4278373885446579e-05,
+      "loss": 0.3424,
+      "step": 17037
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4278038050140212e-05,
+      "loss": 0.3605,
+      "step": 17038
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4277702214833847e-05,
+      "loss": 0.3125,
+      "step": 17039
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4277366379527481e-05,
+      "loss": 0.4473,
+      "step": 17040
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4277030544221116e-05,
+      "loss": 0.531,
+      "step": 17041
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4276694708914749e-05,
+      "loss": 0.2853,
+      "step": 17042
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4276358873608385e-05,
+      "loss": 0.245,
+      "step": 17043
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4276023038302018e-05,
+      "loss": 0.3512,
+      "step": 17044
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4275687202995651e-05,
+      "loss": 0.4433,
+      "step": 17045
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4275351367689287e-05,
+      "loss": 0.1971,
+      "step": 17046
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.427501553238292e-05,
+      "loss": 0.2908,
+      "step": 17047
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4274679697076555e-05,
+      "loss": 0.2968,
+      "step": 17048
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.427434386177019e-05,
+      "loss": 0.1799,
+      "step": 17049
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4274008026463824e-05,
+      "loss": 0.3515,
+      "step": 17050
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4273672191157457e-05,
+      "loss": 0.3362,
+      "step": 17051
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4273336355851094e-05,
+      "loss": 0.499,
+      "step": 17052
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4273000520544727e-05,
+      "loss": 0.2511,
+      "step": 17053
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.427266468523836e-05,
+      "loss": 0.3233,
+      "step": 17054
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4272328849931994e-05,
+      "loss": 0.1918,
+      "step": 17055
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4271993014625629e-05,
+      "loss": 0.208,
+      "step": 17056
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4271657179319262e-05,
+      "loss": 0.3341,
+      "step": 17057
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4271321344012896e-05,
+      "loss": 0.4397,
+      "step": 17058
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4270985508706531e-05,
+      "loss": 0.4034,
+      "step": 17059
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4270649673400166e-05,
+      "loss": 0.252,
+      "step": 17060
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4270313838093799e-05,
+      "loss": 0.2722,
+      "step": 17061
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4269978002787435e-05,
+      "loss": 0.3171,
+      "step": 17062
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4269642167481068e-05,
+      "loss": 0.2414,
+      "step": 17063
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4269306332174701e-05,
+      "loss": 0.2345,
+      "step": 17064
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4268970496868337e-05,
+      "loss": 0.3112,
+      "step": 17065
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.426863466156197e-05,
+      "loss": 0.2957,
+      "step": 17066
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4268298826255605e-05,
+      "loss": 0.29,
+      "step": 17067
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.426796299094924e-05,
+      "loss": 0.2722,
+      "step": 17068
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4267627155642874e-05,
+      "loss": 0.2603,
+      "step": 17069
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4267291320336507e-05,
+      "loss": 0.3605,
+      "step": 17070
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4266955485030144e-05,
+      "loss": 0.3813,
+      "step": 17071
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4266619649723776e-05,
+      "loss": 0.4567,
+      "step": 17072
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.426628381441741e-05,
+      "loss": 0.2161,
+      "step": 17073
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4265947979111046e-05,
+      "loss": 0.4464,
+      "step": 17074
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4265612143804679e-05,
+      "loss": 0.1917,
+      "step": 17075
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4265276308498313e-05,
+      "loss": 0.2393,
+      "step": 17076
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4264940473191948e-05,
+      "loss": 0.4437,
+      "step": 17077
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4264604637885583e-05,
+      "loss": 0.3659,
+      "step": 17078
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4264268802579216e-05,
+      "loss": 0.2726,
+      "step": 17079
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.426393296727285e-05,
+      "loss": 0.5416,
+      "step": 17080
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4263597131966485e-05,
+      "loss": 0.2591,
+      "step": 17081
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4263261296660118e-05,
+      "loss": 0.3219,
+      "step": 17082
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4262925461353754e-05,
+      "loss": 0.2248,
+      "step": 17083
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4262589626047387e-05,
+      "loss": 0.3176,
+      "step": 17084
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.426225379074102e-05,
+      "loss": 0.316,
+      "step": 17085
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4261917955434656e-05,
+      "loss": 0.2702,
+      "step": 17086
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.426158212012829e-05,
+      "loss": 0.2107,
+      "step": 17087
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4261246284821924e-05,
+      "loss": 0.3558,
+      "step": 17088
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4260910449515559e-05,
+      "loss": 0.2157,
+      "step": 17089
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4260574614209193e-05,
+      "loss": 0.345,
+      "step": 17090
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4260238778902826e-05,
+      "loss": 0.2827,
+      "step": 17091
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4259902943596463e-05,
+      "loss": 0.3301,
+      "step": 17092
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4259567108290096e-05,
+      "loss": 0.2942,
+      "step": 17093
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4259231272983729e-05,
+      "loss": 0.3062,
+      "step": 17094
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4258895437677365e-05,
+      "loss": 0.3397,
+      "step": 17095
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4258559602370998e-05,
+      "loss": 0.3068,
+      "step": 17096
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4258223767064632e-05,
+      "loss": 0.1683,
+      "step": 17097
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4257887931758267e-05,
+      "loss": 0.5103,
+      "step": 17098
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4257552096451902e-05,
+      "loss": 0.2561,
+      "step": 17099
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4257216261145535e-05,
+      "loss": 0.3496,
+      "step": 17100
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.425688042583917e-05,
+      "loss": 0.3738,
+      "step": 17101
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4256544590532804e-05,
+      "loss": 0.2645,
+      "step": 17102
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4256208755226437e-05,
+      "loss": 0.3313,
+      "step": 17103
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4255872919920073e-05,
+      "loss": 0.2582,
+      "step": 17104
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4255537084613706e-05,
+      "loss": 0.3579,
+      "step": 17105
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.425520124930734e-05,
+      "loss": 0.3006,
+      "step": 17106
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4254865414000976e-05,
+      "loss": 0.4734,
+      "step": 17107
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4254529578694609e-05,
+      "loss": 0.2964,
+      "step": 17108
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4254193743388243e-05,
+      "loss": 0.2962,
+      "step": 17109
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4253857908081878e-05,
+      "loss": 0.2488,
+      "step": 17110
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4253522072775512e-05,
+      "loss": 0.3027,
+      "step": 17111
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4253186237469145e-05,
+      "loss": 0.2545,
+      "step": 17112
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4252850402162782e-05,
+      "loss": 0.2789,
+      "step": 17113
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4252514566856415e-05,
+      "loss": 0.5124,
+      "step": 17114
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4252178731550048e-05,
+      "loss": 0.2522,
+      "step": 17115
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4251842896243684e-05,
+      "loss": 0.5642,
+      "step": 17116
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4251507060937317e-05,
+      "loss": 0.3483,
+      "step": 17117
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4251171225630952e-05,
+      "loss": 0.4131,
+      "step": 17118
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4250835390324586e-05,
+      "loss": 0.3164,
+      "step": 17119
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4250499555018221e-05,
+      "loss": 0.4585,
+      "step": 17120
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4250163719711854e-05,
+      "loss": 0.3959,
+      "step": 17121
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.424982788440549e-05,
+      "loss": 0.237,
+      "step": 17122
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4249492049099123e-05,
+      "loss": 0.3498,
+      "step": 17123
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4249156213792756e-05,
+      "loss": 0.1688,
+      "step": 17124
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4248820378486392e-05,
+      "loss": 0.3275,
+      "step": 17125
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4248484543180025e-05,
+      "loss": 0.35,
+      "step": 17126
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4248148707873658e-05,
+      "loss": 0.4381,
+      "step": 17127
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4247812872567295e-05,
+      "loss": 0.2678,
+      "step": 17128
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4247477037260928e-05,
+      "loss": 0.2286,
+      "step": 17129
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4247141201954562e-05,
+      "loss": 0.2457,
+      "step": 17130
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4246805366648197e-05,
+      "loss": 0.2291,
+      "step": 17131
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4246469531341832e-05,
+      "loss": 0.3877,
+      "step": 17132
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4246133696035465e-05,
+      "loss": 0.2086,
+      "step": 17133
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4245797860729101e-05,
+      "loss": 0.4218,
+      "step": 17134
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4245462025422734e-05,
+      "loss": 0.1741,
+      "step": 17135
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4245126190116367e-05,
+      "loss": 0.4077,
+      "step": 17136
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4244790354810003e-05,
+      "loss": 0.2087,
+      "step": 17137
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4244454519503636e-05,
+      "loss": 0.4164,
+      "step": 17138
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.424411868419727e-05,
+      "loss": 0.3658,
+      "step": 17139
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4243782848890905e-05,
+      "loss": 0.3027,
+      "step": 17140
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.424344701358454e-05,
+      "loss": 0.3388,
+      "step": 17141
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4243111178278173e-05,
+      "loss": 0.246,
+      "step": 17142
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.424277534297181e-05,
+      "loss": 0.5073,
+      "step": 17143
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4242439507665442e-05,
+      "loss": 0.4425,
+      "step": 17144
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4242103672359075e-05,
+      "loss": 0.2457,
+      "step": 17145
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4241767837052712e-05,
+      "loss": 0.3596,
+      "step": 17146
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4241432001746345e-05,
+      "loss": 0.3238,
+      "step": 17147
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4241096166439977e-05,
+      "loss": 0.2742,
+      "step": 17148
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4240760331133614e-05,
+      "loss": 0.2528,
+      "step": 17149
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4240424495827247e-05,
+      "loss": 0.3654,
+      "step": 17150
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4240088660520881e-05,
+      "loss": 0.3239,
+      "step": 17151
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4239752825214516e-05,
+      "loss": 0.4052,
+      "step": 17152
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.423941698990815e-05,
+      "loss": 0.3595,
+      "step": 17153
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4239081154601784e-05,
+      "loss": 0.3818,
+      "step": 17154
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.423874531929542e-05,
+      "loss": 0.2953,
+      "step": 17155
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4238409483989053e-05,
+      "loss": 0.1802,
+      "step": 17156
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4238073648682686e-05,
+      "loss": 0.3705,
+      "step": 17157
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4237737813376322e-05,
+      "loss": 0.4716,
+      "step": 17158
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4237401978069955e-05,
+      "loss": 0.3284,
+      "step": 17159
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.423706614276359e-05,
+      "loss": 0.2605,
+      "step": 17160
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4236730307457225e-05,
+      "loss": 0.3634,
+      "step": 17161
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4236394472150859e-05,
+      "loss": 0.3201,
+      "step": 17162
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4236058636844492e-05,
+      "loss": 0.3145,
+      "step": 17163
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4235722801538128e-05,
+      "loss": 0.3668,
+      "step": 17164
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4235386966231761e-05,
+      "loss": 0.4136,
+      "step": 17165
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4235051130925394e-05,
+      "loss": 0.314,
+      "step": 17166
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.423471529561903e-05,
+      "loss": 0.3832,
+      "step": 17167
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4234379460312664e-05,
+      "loss": 0.2836,
+      "step": 17168
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4234043625006297e-05,
+      "loss": 0.4331,
+      "step": 17169
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4233707789699933e-05,
+      "loss": 0.1673,
+      "step": 17170
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.4233371954393566e-05,
+      "loss": 0.3185,
+      "step": 17171
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.42330361190872e-05,
+      "loss": 0.4348,
+      "step": 17172
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4232700283780835e-05,
+      "loss": 0.3193,
+      "step": 17173
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.423236444847447e-05,
+      "loss": 0.421,
+      "step": 17174
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4232028613168103e-05,
+      "loss": 0.2073,
+      "step": 17175
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4231692777861739e-05,
+      "loss": 0.3779,
+      "step": 17176
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4231356942555372e-05,
+      "loss": 0.5377,
+      "step": 17177
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4231021107249005e-05,
+      "loss": 0.2895,
+      "step": 17178
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4230685271942641e-05,
+      "loss": 0.4344,
+      "step": 17179
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4230349436636274e-05,
+      "loss": 0.372,
+      "step": 17180
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4230013601329909e-05,
+      "loss": 0.3682,
+      "step": 17181
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4229677766023544e-05,
+      "loss": 0.2609,
+      "step": 17182
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4229341930717178e-05,
+      "loss": 0.3659,
+      "step": 17183
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4229006095410811e-05,
+      "loss": 0.2341,
+      "step": 17184
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4228670260104448e-05,
+      "loss": 0.2267,
+      "step": 17185
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.422833442479808e-05,
+      "loss": 0.3359,
+      "step": 17186
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4227998589491713e-05,
+      "loss": 0.406,
+      "step": 17187
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.422766275418535e-05,
+      "loss": 0.3489,
+      "step": 17188
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4227326918878983e-05,
+      "loss": 0.2334,
+      "step": 17189
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4226991083572617e-05,
+      "loss": 0.1745,
+      "step": 17190
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4226655248266252e-05,
+      "loss": 0.4066,
+      "step": 17191
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4226319412959885e-05,
+      "loss": 0.3441,
+      "step": 17192
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.422598357765352e-05,
+      "loss": 0.5784,
+      "step": 17193
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4225647742347154e-05,
+      "loss": 0.3098,
+      "step": 17194
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4225311907040789e-05,
+      "loss": 0.2307,
+      "step": 17195
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4224976071734422e-05,
+      "loss": 0.2076,
+      "step": 17196
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4224640236428058e-05,
+      "loss": 0.4351,
+      "step": 17197
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4224304401121691e-05,
+      "loss": 0.2969,
+      "step": 17198
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4223968565815324e-05,
+      "loss": 0.1815,
+      "step": 17199
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.422363273050896e-05,
+      "loss": 0.2853,
+      "step": 17200
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4223296895202593e-05,
+      "loss": 0.1883,
+      "step": 17201
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4222961059896228e-05,
+      "loss": 0.5124,
+      "step": 17202
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4222625224589863e-05,
+      "loss": 0.2912,
+      "step": 17203
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4222289389283497e-05,
+      "loss": 0.2326,
+      "step": 17204
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.422195355397713e-05,
+      "loss": 0.3204,
+      "step": 17205
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4221617718670767e-05,
+      "loss": 0.275,
+      "step": 17206
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.42212818833644e-05,
+      "loss": 0.3377,
+      "step": 17207
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4220946048058033e-05,
+      "loss": 0.1985,
+      "step": 17208
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4220610212751669e-05,
+      "loss": 0.2462,
+      "step": 17209
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4220274377445302e-05,
+      "loss": 0.3173,
+      "step": 17210
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4219938542138937e-05,
+      "loss": 0.2505,
+      "step": 17211
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4219602706832571e-05,
+      "loss": 0.3698,
+      "step": 17212
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4219266871526206e-05,
+      "loss": 0.3224,
+      "step": 17213
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4218931036219839e-05,
+      "loss": 0.2098,
+      "step": 17214
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4218595200913473e-05,
+      "loss": 0.2557,
+      "step": 17215
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4218259365607108e-05,
+      "loss": 0.1501,
+      "step": 17216
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4217923530300741e-05,
+      "loss": 0.287,
+      "step": 17217
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4217587694994377e-05,
+      "loss": 0.3425,
+      "step": 17218
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.421725185968801e-05,
+      "loss": 0.3228,
+      "step": 17219
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4216916024381643e-05,
+      "loss": 0.2897,
+      "step": 17220
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.421658018907528e-05,
+      "loss": 0.356,
+      "step": 17221
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4216244353768913e-05,
+      "loss": 0.2557,
+      "step": 17222
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4215908518462547e-05,
+      "loss": 0.2736,
+      "step": 17223
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4215572683156182e-05,
+      "loss": 0.2667,
+      "step": 17224
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4215236847849817e-05,
+      "loss": 0.4396,
+      "step": 17225
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.421490101254345e-05,
+      "loss": 0.2877,
+      "step": 17226
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4214565177237086e-05,
+      "loss": 0.5434,
+      "step": 17227
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4214229341930719e-05,
+      "loss": 0.5068,
+      "step": 17228
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4213893506624352e-05,
+      "loss": 0.3174,
+      "step": 17229
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4213557671317988e-05,
+      "loss": 0.3893,
+      "step": 17230
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4213221836011621e-05,
+      "loss": 0.3275,
+      "step": 17231
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4212886000705256e-05,
+      "loss": 0.3214,
+      "step": 17232
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.421255016539889e-05,
+      "loss": 0.2451,
+      "step": 17233
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4212214330092525e-05,
+      "loss": 0.3134,
+      "step": 17234
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4211878494786158e-05,
+      "loss": 0.1931,
+      "step": 17235
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4211542659479791e-05,
+      "loss": 0.2488,
+      "step": 17236
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4211206824173427e-05,
+      "loss": 0.2768,
+      "step": 17237
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.421087098886706e-05,
+      "loss": 0.4006,
+      "step": 17238
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4210535153560693e-05,
+      "loss": 0.2659,
+      "step": 17239
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.421019931825433e-05,
+      "loss": 0.4224,
+      "step": 17240
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4209863482947962e-05,
+      "loss": 0.3588,
+      "step": 17241
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4209527647641597e-05,
+      "loss": 0.2813,
+      "step": 17242
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4209191812335232e-05,
+      "loss": 0.255,
+      "step": 17243
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4208855977028866e-05,
+      "loss": 0.3396,
+      "step": 17244
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.42085201417225e-05,
+      "loss": 0.3804,
+      "step": 17245
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4208184306416136e-05,
+      "loss": 0.369,
+      "step": 17246
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4207848471109769e-05,
+      "loss": 0.1293,
+      "step": 17247
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4207512635803402e-05,
+      "loss": 0.3272,
+      "step": 17248
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4207176800497038e-05,
+      "loss": 0.3831,
+      "step": 17249
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4206840965190671e-05,
+      "loss": 0.3612,
+      "step": 17250
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4206505129884305e-05,
+      "loss": 0.2648,
+      "step": 17251
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.420616929457794e-05,
+      "loss": 0.3703,
+      "step": 17252
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4205833459271575e-05,
+      "loss": 0.316,
+      "step": 17253
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4205497623965208e-05,
+      "loss": 0.2908,
+      "step": 17254
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4205161788658844e-05,
+      "loss": 0.2363,
+      "step": 17255
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4204825953352477e-05,
+      "loss": 0.2191,
+      "step": 17256
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.420449011804611e-05,
+      "loss": 0.5231,
+      "step": 17257
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4204154282739746e-05,
+      "loss": 0.2467,
+      "step": 17258
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.420381844743338e-05,
+      "loss": 0.2209,
+      "step": 17259
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4203482612127012e-05,
+      "loss": 0.318,
+      "step": 17260
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4203146776820649e-05,
+      "loss": 0.3053,
+      "step": 17261
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4202810941514282e-05,
+      "loss": 0.3198,
+      "step": 17262
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4202475106207916e-05,
+      "loss": 0.3796,
+      "step": 17263
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.420213927090155e-05,
+      "loss": 0.31,
+      "step": 17264
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4201803435595185e-05,
+      "loss": 0.3761,
+      "step": 17265
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4201467600288818e-05,
+      "loss": 0.3019,
+      "step": 17266
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4201131764982455e-05,
+      "loss": 0.3468,
+      "step": 17267
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4200795929676088e-05,
+      "loss": 0.1836,
+      "step": 17268
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.420046009436972e-05,
+      "loss": 0.3483,
+      "step": 17269
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4200124259063357e-05,
+      "loss": 0.2756,
+      "step": 17270
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.419978842375699e-05,
+      "loss": 0.4338,
+      "step": 17271
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4199452588450625e-05,
+      "loss": 0.2454,
+      "step": 17272
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.419911675314426e-05,
+      "loss": 0.3194,
+      "step": 17273
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4198780917837894e-05,
+      "loss": 0.42,
+      "step": 17274
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4198445082531527e-05,
+      "loss": 0.4448,
+      "step": 17275
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4198109247225163e-05,
+      "loss": 0.3906,
+      "step": 17276
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4197773411918796e-05,
+      "loss": 0.2465,
+      "step": 17277
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4197437576612429e-05,
+      "loss": 0.2624,
+      "step": 17278
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4197101741306065e-05,
+      "loss": 0.2253,
+      "step": 17279
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4196765905999698e-05,
+      "loss": 0.3063,
+      "step": 17280
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4196430070693333e-05,
+      "loss": 0.184,
+      "step": 17281
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4196094235386968e-05,
+      "loss": 0.4939,
+      "step": 17282
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.41957584000806e-05,
+      "loss": 0.2017,
+      "step": 17283
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4195422564774235e-05,
+      "loss": 0.2088,
+      "step": 17284
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.419508672946787e-05,
+      "loss": 0.4721,
+      "step": 17285
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4194750894161505e-05,
+      "loss": 0.2802,
+      "step": 17286
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4194415058855138e-05,
+      "loss": 0.3401,
+      "step": 17287
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4194079223548774e-05,
+      "loss": 0.2415,
+      "step": 17288
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4193743388242407e-05,
+      "loss": 0.3548,
+      "step": 17289
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.419340755293604e-05,
+      "loss": 0.4133,
+      "step": 17290
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4193071717629676e-05,
+      "loss": 0.4036,
+      "step": 17291
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4192735882323309e-05,
+      "loss": 0.2646,
+      "step": 17292
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4192400047016944e-05,
+      "loss": 0.3349,
+      "step": 17293
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4192064211710578e-05,
+      "loss": 0.2842,
+      "step": 17294
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4191728376404213e-05,
+      "loss": 0.265,
+      "step": 17295
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4191392541097846e-05,
+      "loss": 0.2847,
+      "step": 17296
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4191056705791482e-05,
+      "loss": 0.2963,
+      "step": 17297
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4190720870485115e-05,
+      "loss": 0.2947,
+      "step": 17298
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4190385035178748e-05,
+      "loss": 0.3445,
+      "step": 17299
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4190049199872385e-05,
+      "loss": 0.3566,
+      "step": 17300
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4189713364566018e-05,
+      "loss": 0.1457,
+      "step": 17301
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4189377529259652e-05,
+      "loss": 0.3334,
+      "step": 17302
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4189041693953287e-05,
+      "loss": 0.3968,
+      "step": 17303
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.418870585864692e-05,
+      "loss": 0.2745,
+      "step": 17304
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4188370023340554e-05,
+      "loss": 0.1837,
+      "step": 17305
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4188034188034189e-05,
+      "loss": 0.234,
+      "step": 17306
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4187698352727824e-05,
+      "loss": 0.2527,
+      "step": 17307
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4187362517421457e-05,
+      "loss": 0.2537,
+      "step": 17308
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4187026682115093e-05,
+      "loss": 0.2716,
+      "step": 17309
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4186690846808726e-05,
+      "loss": 0.3891,
+      "step": 17310
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4186355011502359e-05,
+      "loss": 0.2041,
+      "step": 17311
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4186019176195995e-05,
+      "loss": 0.3056,
+      "step": 17312
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4185683340889628e-05,
+      "loss": 0.2452,
+      "step": 17313
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4185347505583263e-05,
+      "loss": 0.4615,
+      "step": 17314
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4185011670276898e-05,
+      "loss": 0.2947,
+      "step": 17315
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4184675834970532e-05,
+      "loss": 0.286,
+      "step": 17316
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4184339999664165e-05,
+      "loss": 0.3462,
+      "step": 17317
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4184004164357801e-05,
+      "loss": 0.4384,
+      "step": 17318
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4183668329051434e-05,
+      "loss": 0.2148,
+      "step": 17319
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4183332493745067e-05,
+      "loss": 0.3012,
+      "step": 17320
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4182996658438704e-05,
+      "loss": 0.3103,
+      "step": 17321
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4182660823132337e-05,
+      "loss": 0.2862,
+      "step": 17322
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4182324987825971e-05,
+      "loss": 0.4018,
+      "step": 17323
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4181989152519606e-05,
+      "loss": 0.438,
+      "step": 17324
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.418165331721324e-05,
+      "loss": 0.5103,
+      "step": 17325
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4181317481906874e-05,
+      "loss": 0.3137,
+      "step": 17326
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4180981646600508e-05,
+      "loss": 0.2301,
+      "step": 17327
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4180645811294143e-05,
+      "loss": 0.3817,
+      "step": 17328
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4180309975987776e-05,
+      "loss": 0.1953,
+      "step": 17329
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4179974140681412e-05,
+      "loss": 0.4498,
+      "step": 17330
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4179638305375045e-05,
+      "loss": 0.2788,
+      "step": 17331
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4179302470068678e-05,
+      "loss": 0.3155,
+      "step": 17332
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4178966634762314e-05,
+      "loss": 0.3226,
+      "step": 17333
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4178630799455947e-05,
+      "loss": 0.3655,
+      "step": 17334
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4178294964149582e-05,
+      "loss": 0.2253,
+      "step": 17335
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4177959128843217e-05,
+      "loss": 0.2896,
+      "step": 17336
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4177623293536851e-05,
+      "loss": 0.2582,
+      "step": 17337
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4177287458230484e-05,
+      "loss": 0.2868,
+      "step": 17338
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.417695162292412e-05,
+      "loss": 0.487,
+      "step": 17339
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4176615787617754e-05,
+      "loss": 0.2282,
+      "step": 17340
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4176279952311386e-05,
+      "loss": 0.4879,
+      "step": 17341
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4175944117005023e-05,
+      "loss": 0.3033,
+      "step": 17342
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4175608281698656e-05,
+      "loss": 0.3575,
+      "step": 17343
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.417527244639229e-05,
+      "loss": 0.1694,
+      "step": 17344
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4174936611085925e-05,
+      "loss": 0.258,
+      "step": 17345
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.417460077577956e-05,
+      "loss": 0.2903,
+      "step": 17346
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4174264940473193e-05,
+      "loss": 0.2791,
+      "step": 17347
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4173929105166829e-05,
+      "loss": 0.2935,
+      "step": 17348
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4173593269860462e-05,
+      "loss": 0.2411,
+      "step": 17349
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4173257434554095e-05,
+      "loss": 0.2837,
+      "step": 17350
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4172921599247731e-05,
+      "loss": 0.2302,
+      "step": 17351
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4172585763941364e-05,
+      "loss": 0.3106,
+      "step": 17352
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4172249928634997e-05,
+      "loss": 0.1959,
+      "step": 17353
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4171914093328634e-05,
+      "loss": 0.2388,
+      "step": 17354
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4171578258022266e-05,
+      "loss": 0.2558,
+      "step": 17355
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4171242422715901e-05,
+      "loss": 0.248,
+      "step": 17356
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4170906587409536e-05,
+      "loss": 0.3901,
+      "step": 17357
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.417057075210317e-05,
+      "loss": 0.2685,
+      "step": 17358
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4170234916796803e-05,
+      "loss": 0.119,
+      "step": 17359
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.416989908149044e-05,
+      "loss": 0.2061,
+      "step": 17360
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4169563246184073e-05,
+      "loss": 0.451,
+      "step": 17361
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4169227410877706e-05,
+      "loss": 0.3065,
+      "step": 17362
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4168891575571342e-05,
+      "loss": 0.2509,
+      "step": 17363
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4168555740264975e-05,
+      "loss": 0.426,
+      "step": 17364
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.416821990495861e-05,
+      "loss": 0.5031,
+      "step": 17365
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4167884069652244e-05,
+      "loss": 0.3209,
+      "step": 17366
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4167548234345879e-05,
+      "loss": 0.3247,
+      "step": 17367
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4167212399039512e-05,
+      "loss": 0.2698,
+      "step": 17368
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.4166876563733148e-05,
+      "loss": 0.2475,
+      "step": 17369
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4166540728426781e-05,
+      "loss": 0.4052,
+      "step": 17370
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4166204893120414e-05,
+      "loss": 0.3827,
+      "step": 17371
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.416586905781405e-05,
+      "loss": 0.3325,
+      "step": 17372
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4165533222507683e-05,
+      "loss": 0.2597,
+      "step": 17373
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4165197387201316e-05,
+      "loss": 0.226,
+      "step": 17374
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4164861551894953e-05,
+      "loss": 0.4445,
+      "step": 17375
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4164525716588586e-05,
+      "loss": 0.2528,
+      "step": 17376
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.416418988128222e-05,
+      "loss": 0.2783,
+      "step": 17377
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4163854045975855e-05,
+      "loss": 0.3495,
+      "step": 17378
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.416351821066949e-05,
+      "loss": 0.353,
+      "step": 17379
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4163182375363122e-05,
+      "loss": 0.3329,
+      "step": 17380
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4162846540056759e-05,
+      "loss": 0.3033,
+      "step": 17381
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4162510704750392e-05,
+      "loss": 0.2466,
+      "step": 17382
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4162174869444025e-05,
+      "loss": 0.3756,
+      "step": 17383
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4161839034137661e-05,
+      "loss": 0.3703,
+      "step": 17384
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4161503198831294e-05,
+      "loss": 0.3312,
+      "step": 17385
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4161167363524929e-05,
+      "loss": 0.3529,
+      "step": 17386
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4160831528218563e-05,
+      "loss": 0.3066,
+      "step": 17387
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4160495692912198e-05,
+      "loss": 0.2354,
+      "step": 17388
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4160159857605831e-05,
+      "loss": 0.439,
+      "step": 17389
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4159824022299467e-05,
+      "loss": 0.3877,
+      "step": 17390
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.41594881869931e-05,
+      "loss": 0.2617,
+      "step": 17391
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4159152351686733e-05,
+      "loss": 0.3264,
+      "step": 17392
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.415881651638037e-05,
+      "loss": 0.2386,
+      "step": 17393
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4158480681074002e-05,
+      "loss": 0.333,
+      "step": 17394
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4158144845767635e-05,
+      "loss": 0.3471,
+      "step": 17395
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4157809010461272e-05,
+      "loss": 0.2639,
+      "step": 17396
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4157473175154905e-05,
+      "loss": 0.2891,
+      "step": 17397
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.415713733984854e-05,
+      "loss": 0.3728,
+      "step": 17398
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4156801504542174e-05,
+      "loss": 0.3231,
+      "step": 17399
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4156465669235809e-05,
+      "loss": 0.234,
+      "step": 17400
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4156129833929442e-05,
+      "loss": 0.3012,
+      "step": 17401
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4155793998623078e-05,
+      "loss": 0.4022,
+      "step": 17402
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4155458163316711e-05,
+      "loss": 0.2725,
+      "step": 17403
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4155122328010344e-05,
+      "loss": 0.3435,
+      "step": 17404
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.415478649270398e-05,
+      "loss": 0.3642,
+      "step": 17405
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4154450657397613e-05,
+      "loss": 0.3158,
+      "step": 17406
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4154114822091248e-05,
+      "loss": 0.2388,
+      "step": 17407
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4153778986784882e-05,
+      "loss": 0.4321,
+      "step": 17408
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4153443151478517e-05,
+      "loss": 0.2434,
+      "step": 17409
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.415310731617215e-05,
+      "loss": 0.2634,
+      "step": 17410
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4152771480865786e-05,
+      "loss": 0.3449,
+      "step": 17411
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.415243564555942e-05,
+      "loss": 0.1788,
+      "step": 17412
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4152099810253052e-05,
+      "loss": 0.2395,
+      "step": 17413
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4151763974946689e-05,
+      "loss": 0.2798,
+      "step": 17414
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4151428139640322e-05,
+      "loss": 0.1139,
+      "step": 17415
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4151092304333956e-05,
+      "loss": 0.2309,
+      "step": 17416
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.415075646902759e-05,
+      "loss": 0.392,
+      "step": 17417
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4150420633721224e-05,
+      "loss": 0.4082,
+      "step": 17418
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4150084798414858e-05,
+      "loss": 0.1989,
+      "step": 17419
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4149748963108491e-05,
+      "loss": 0.4093,
+      "step": 17420
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4149413127802128e-05,
+      "loss": 0.4594,
+      "step": 17421
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.414907729249576e-05,
+      "loss": 0.2147,
+      "step": 17422
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4148741457189394e-05,
+      "loss": 0.3735,
+      "step": 17423
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.414840562188303e-05,
+      "loss": 0.2427,
+      "step": 17424
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4148069786576663e-05,
+      "loss": 0.3465,
+      "step": 17425
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4147733951270298e-05,
+      "loss": 0.3757,
+      "step": 17426
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4147398115963932e-05,
+      "loss": 0.3127,
+      "step": 17427
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4147062280657567e-05,
+      "loss": 0.5426,
+      "step": 17428
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.41467264453512e-05,
+      "loss": 0.2711,
+      "step": 17429
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4146390610044836e-05,
+      "loss": 0.2987,
+      "step": 17430
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.414605477473847e-05,
+      "loss": 0.2608,
+      "step": 17431
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4145718939432102e-05,
+      "loss": 0.5615,
+      "step": 17432
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4145383104125738e-05,
+      "loss": 0.293,
+      "step": 17433
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4145047268819371e-05,
+      "loss": 0.3453,
+      "step": 17434
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4144711433513006e-05,
+      "loss": 0.2197,
+      "step": 17435
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.414437559820664e-05,
+      "loss": 0.1986,
+      "step": 17436
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4144039762900275e-05,
+      "loss": 0.198,
+      "step": 17437
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4143703927593908e-05,
+      "loss": 0.2648,
+      "step": 17438
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4143368092287543e-05,
+      "loss": 0.2853,
+      "step": 17439
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4143032256981178e-05,
+      "loss": 0.3879,
+      "step": 17440
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.414269642167481e-05,
+      "loss": 0.2072,
+      "step": 17441
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4142360586368447e-05,
+      "loss": 0.3951,
+      "step": 17442
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.414202475106208e-05,
+      "loss": 0.3942,
+      "step": 17443
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4141688915755713e-05,
+      "loss": 0.197,
+      "step": 17444
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.414135308044935e-05,
+      "loss": 0.2089,
+      "step": 17445
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4141017245142982e-05,
+      "loss": 0.3512,
+      "step": 17446
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4140681409836617e-05,
+      "loss": 0.2964,
+      "step": 17447
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4140345574530251e-05,
+      "loss": 0.273,
+      "step": 17448
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4140009739223886e-05,
+      "loss": 0.3632,
+      "step": 17449
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4139673903917519e-05,
+      "loss": 0.3269,
+      "step": 17450
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4139338068611155e-05,
+      "loss": 0.2367,
+      "step": 17451
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4139002233304788e-05,
+      "loss": 0.3161,
+      "step": 17452
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4138666397998421e-05,
+      "loss": 0.2542,
+      "step": 17453
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4138330562692058e-05,
+      "loss": 0.2666,
+      "step": 17454
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.413799472738569e-05,
+      "loss": 0.1804,
+      "step": 17455
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4137658892079325e-05,
+      "loss": 0.5509,
+      "step": 17456
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.413732305677296e-05,
+      "loss": 0.2799,
+      "step": 17457
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4136987221466594e-05,
+      "loss": 0.3067,
+      "step": 17458
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4136651386160227e-05,
+      "loss": 0.2192,
+      "step": 17459
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4136315550853864e-05,
+      "loss": 0.3552,
+      "step": 17460
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4135979715547497e-05,
+      "loss": 0.3696,
+      "step": 17461
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.413564388024113e-05,
+      "loss": 0.2702,
+      "step": 17462
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4135308044934766e-05,
+      "loss": 0.2308,
+      "step": 17463
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4134972209628399e-05,
+      "loss": 0.4274,
+      "step": 17464
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4134636374322032e-05,
+      "loss": 0.2417,
+      "step": 17465
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4134300539015668e-05,
+      "loss": 0.255,
+      "step": 17466
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4133964703709301e-05,
+      "loss": 0.3382,
+      "step": 17467
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4133628868402936e-05,
+      "loss": 0.183,
+      "step": 17468
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.413329303309657e-05,
+      "loss": 0.1581,
+      "step": 17469
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4132957197790205e-05,
+      "loss": 0.2303,
+      "step": 17470
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4132621362483838e-05,
+      "loss": 0.2644,
+      "step": 17471
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4132285527177474e-05,
+      "loss": 0.2382,
+      "step": 17472
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4131949691871107e-05,
+      "loss": 0.3188,
+      "step": 17473
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.413161385656474e-05,
+      "loss": 0.2643,
+      "step": 17474
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4131278021258377e-05,
+      "loss": 0.4303,
+      "step": 17475
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.413094218595201e-05,
+      "loss": 0.3855,
+      "step": 17476
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4130606350645644e-05,
+      "loss": 0.2262,
+      "step": 17477
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4130270515339279e-05,
+      "loss": 0.2483,
+      "step": 17478
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4129934680032914e-05,
+      "loss": 0.2861,
+      "step": 17479
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4129598844726547e-05,
+      "loss": 0.278,
+      "step": 17480
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4129263009420183e-05,
+      "loss": 0.4722,
+      "step": 17481
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4128927174113816e-05,
+      "loss": 0.2269,
+      "step": 17482
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4128591338807449e-05,
+      "loss": 0.3783,
+      "step": 17483
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4128255503501085e-05,
+      "loss": 0.5519,
+      "step": 17484
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4127919668194718e-05,
+      "loss": 0.3918,
+      "step": 17485
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4127583832888351e-05,
+      "loss": 0.3743,
+      "step": 17486
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4127247997581987e-05,
+      "loss": 0.292,
+      "step": 17487
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.412691216227562e-05,
+      "loss": 0.1694,
+      "step": 17488
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4126576326969255e-05,
+      "loss": 0.4235,
+      "step": 17489
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.412624049166289e-05,
+      "loss": 0.4048,
+      "step": 17490
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4125904656356524e-05,
+      "loss": 0.3371,
+      "step": 17491
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4125568821050157e-05,
+      "loss": 0.3142,
+      "step": 17492
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4125232985743794e-05,
+      "loss": 0.4342,
+      "step": 17493
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4124897150437427e-05,
+      "loss": 0.3887,
+      "step": 17494
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.412456131513106e-05,
+      "loss": 0.4679,
+      "step": 17495
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4124225479824696e-05,
+      "loss": 0.4993,
+      "step": 17496
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4123889644518329e-05,
+      "loss": 0.3258,
+      "step": 17497
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4123553809211963e-05,
+      "loss": 0.376,
+      "step": 17498
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4123217973905598e-05,
+      "loss": 0.2201,
+      "step": 17499
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4122882138599233e-05,
+      "loss": 0.2092,
+      "step": 17500
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4122546303292866e-05,
+      "loss": 0.3107,
+      "step": 17501
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4122210467986502e-05,
+      "loss": 0.3214,
+      "step": 17502
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4121874632680135e-05,
+      "loss": 0.2326,
+      "step": 17503
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4121538797373768e-05,
+      "loss": 0.1896,
+      "step": 17504
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4121202962067404e-05,
+      "loss": 0.3251,
+      "step": 17505
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4120867126761037e-05,
+      "loss": 0.2845,
+      "step": 17506
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4120531291454672e-05,
+      "loss": 0.3514,
+      "step": 17507
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4120195456148307e-05,
+      "loss": 0.3729,
+      "step": 17508
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.411985962084194e-05,
+      "loss": 0.4868,
+      "step": 17509
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4119523785535574e-05,
+      "loss": 0.2384,
+      "step": 17510
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4119187950229209e-05,
+      "loss": 0.3883,
+      "step": 17511
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4118852114922843e-05,
+      "loss": 0.3408,
+      "step": 17512
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4118516279616476e-05,
+      "loss": 0.4268,
+      "step": 17513
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4118180444310113e-05,
+      "loss": 0.4959,
+      "step": 17514
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4117844609003746e-05,
+      "loss": 0.4013,
+      "step": 17515
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4117508773697379e-05,
+      "loss": 0.1986,
+      "step": 17516
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4117172938391015e-05,
+      "loss": 0.2409,
+      "step": 17517
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4116837103084648e-05,
+      "loss": 0.2517,
+      "step": 17518
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4116501267778283e-05,
+      "loss": 0.3092,
+      "step": 17519
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4116165432471917e-05,
+      "loss": 0.5774,
+      "step": 17520
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4115829597165552e-05,
+      "loss": 0.2853,
+      "step": 17521
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4115493761859185e-05,
+      "loss": 0.267,
+      "step": 17522
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4115157926552821e-05,
+      "loss": 0.5169,
+      "step": 17523
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4114822091246454e-05,
+      "loss": 0.2483,
+      "step": 17524
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4114486255940087e-05,
+      "loss": 0.3497,
+      "step": 17525
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4114150420633723e-05,
+      "loss": 0.4645,
+      "step": 17526
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4113814585327356e-05,
+      "loss": 0.2987,
+      "step": 17527
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4113478750020991e-05,
+      "loss": 0.2331,
+      "step": 17528
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4113142914714626e-05,
+      "loss": 0.2618,
+      "step": 17529
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4112807079408259e-05,
+      "loss": 0.3383,
+      "step": 17530
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4112471244101893e-05,
+      "loss": 0.3328,
+      "step": 17531
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4112135408795528e-05,
+      "loss": 0.2539,
+      "step": 17532
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4111799573489163e-05,
+      "loss": 0.2439,
+      "step": 17533
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4111463738182796e-05,
+      "loss": 0.3417,
+      "step": 17534
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4111127902876432e-05,
+      "loss": 0.2682,
+      "step": 17535
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4110792067570065e-05,
+      "loss": 0.2305,
+      "step": 17536
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4110456232263698e-05,
+      "loss": 0.3526,
+      "step": 17537
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4110120396957334e-05,
+      "loss": 0.3669,
+      "step": 17538
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4109784561650967e-05,
+      "loss": 0.3,
+      "step": 17539
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4109448726344602e-05,
+      "loss": 0.2105,
+      "step": 17540
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4109112891038236e-05,
+      "loss": 0.2461,
+      "step": 17541
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4108777055731871e-05,
+      "loss": 0.517,
+      "step": 17542
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4108441220425504e-05,
+      "loss": 0.23,
+      "step": 17543
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.410810538511914e-05,
+      "loss": 0.2663,
+      "step": 17544
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4107769549812773e-05,
+      "loss": 0.1859,
+      "step": 17545
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4107433714506406e-05,
+      "loss": 0.2622,
+      "step": 17546
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4107097879200043e-05,
+      "loss": 0.2169,
+      "step": 17547
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4106762043893675e-05,
+      "loss": 0.3307,
+      "step": 17548
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.410642620858731e-05,
+      "loss": 0.5307,
+      "step": 17549
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4106090373280945e-05,
+      "loss": 0.2616,
+      "step": 17550
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.410575453797458e-05,
+      "loss": 0.3052,
+      "step": 17551
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4105418702668212e-05,
+      "loss": 0.1796,
+      "step": 17552
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4105082867361847e-05,
+      "loss": 0.3848,
+      "step": 17553
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4104747032055482e-05,
+      "loss": 0.2256,
+      "step": 17554
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4104411196749115e-05,
+      "loss": 0.2452,
+      "step": 17555
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4104075361442751e-05,
+      "loss": 0.2725,
+      "step": 17556
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4103739526136384e-05,
+      "loss": 0.3575,
+      "step": 17557
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4103403690830017e-05,
+      "loss": 0.3977,
+      "step": 17558
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4103067855523653e-05,
+      "loss": 0.2367,
+      "step": 17559
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4102732020217286e-05,
+      "loss": 0.3357,
+      "step": 17560
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.410239618491092e-05,
+      "loss": 0.3459,
+      "step": 17561
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4102060349604555e-05,
+      "loss": 0.4138,
+      "step": 17562
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.410172451429819e-05,
+      "loss": 0.2723,
+      "step": 17563
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4101388678991823e-05,
+      "loss": 0.3421,
+      "step": 17564
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.410105284368546e-05,
+      "loss": 0.3382,
+      "step": 17565
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4100717008379092e-05,
+      "loss": 0.3492,
+      "step": 17566
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4100381173072725e-05,
+      "loss": 0.2209,
+      "step": 17567
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.4100045337766362e-05,
+      "loss": 0.2157,
+      "step": 17568
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4099709502459995e-05,
+      "loss": 0.3383,
+      "step": 17569
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.409937366715363e-05,
+      "loss": 0.4084,
+      "step": 17570
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4099037831847264e-05,
+      "loss": 0.3055,
+      "step": 17571
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4098701996540899e-05,
+      "loss": 0.3341,
+      "step": 17572
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4098366161234532e-05,
+      "loss": 0.2893,
+      "step": 17573
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4098030325928166e-05,
+      "loss": 0.2483,
+      "step": 17574
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.40976944906218e-05,
+      "loss": 0.3754,
+      "step": 17575
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4097358655315434e-05,
+      "loss": 0.2009,
+      "step": 17576
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.409702282000907e-05,
+      "loss": 0.3529,
+      "step": 17577
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4096686984702703e-05,
+      "loss": 0.3304,
+      "step": 17578
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4096351149396336e-05,
+      "loss": 0.1951,
+      "step": 17579
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4096015314089972e-05,
+      "loss": 0.4473,
+      "step": 17580
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4095679478783605e-05,
+      "loss": 0.4156,
+      "step": 17581
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.409534364347724e-05,
+      "loss": 0.3715,
+      "step": 17582
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4095007808170875e-05,
+      "loss": 0.3702,
+      "step": 17583
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.409467197286451e-05,
+      "loss": 0.517,
+      "step": 17584
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4094336137558142e-05,
+      "loss": 0.2661,
+      "step": 17585
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4094000302251779e-05,
+      "loss": 0.5313,
+      "step": 17586
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4093664466945411e-05,
+      "loss": 0.3688,
+      "step": 17587
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4093328631639044e-05,
+      "loss": 0.272,
+      "step": 17588
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.409299279633268e-05,
+      "loss": 0.419,
+      "step": 17589
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4092656961026314e-05,
+      "loss": 0.275,
+      "step": 17590
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4092321125719948e-05,
+      "loss": 0.4425,
+      "step": 17591
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4091985290413583e-05,
+      "loss": 0.3641,
+      "step": 17592
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4091649455107218e-05,
+      "loss": 0.3586,
+      "step": 17593
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.409131361980085e-05,
+      "loss": 0.2745,
+      "step": 17594
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4090977784494487e-05,
+      "loss": 0.2462,
+      "step": 17595
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.409064194918812e-05,
+      "loss": 0.3014,
+      "step": 17596
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4090306113881753e-05,
+      "loss": 0.3573,
+      "step": 17597
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4089970278575388e-05,
+      "loss": 0.2245,
+      "step": 17598
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4089634443269022e-05,
+      "loss": 0.3156,
+      "step": 17599
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4089298607962655e-05,
+      "loss": 0.3884,
+      "step": 17600
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.408896277265629e-05,
+      "loss": 0.3819,
+      "step": 17601
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4088626937349924e-05,
+      "loss": 0.2742,
+      "step": 17602
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4088291102043559e-05,
+      "loss": 0.2808,
+      "step": 17603
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4087955266737192e-05,
+      "loss": 0.3346,
+      "step": 17604
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4087619431430828e-05,
+      "loss": 0.2369,
+      "step": 17605
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4087283596124461e-05,
+      "loss": 0.4153,
+      "step": 17606
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4086947760818094e-05,
+      "loss": 0.3323,
+      "step": 17607
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.408661192551173e-05,
+      "loss": 0.295,
+      "step": 17608
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4086276090205364e-05,
+      "loss": 0.3098,
+      "step": 17609
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4085940254898998e-05,
+      "loss": 0.2898,
+      "step": 17610
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4085604419592633e-05,
+      "loss": 0.3084,
+      "step": 17611
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4085268584286268e-05,
+      "loss": 0.258,
+      "step": 17612
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.40849327489799e-05,
+      "loss": 0.2071,
+      "step": 17613
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4084596913673537e-05,
+      "loss": 0.2747,
+      "step": 17614
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.408426107836717e-05,
+      "loss": 0.2222,
+      "step": 17615
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4083925243060803e-05,
+      "loss": 0.3379,
+      "step": 17616
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4083589407754439e-05,
+      "loss": 0.1826,
+      "step": 17617
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4083253572448072e-05,
+      "loss": 0.2209,
+      "step": 17618
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4082917737141707e-05,
+      "loss": 0.1846,
+      "step": 17619
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4082581901835341e-05,
+      "loss": 0.3741,
+      "step": 17620
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4082246066528974e-05,
+      "loss": 0.2613,
+      "step": 17621
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4081910231222609e-05,
+      "loss": 0.1977,
+      "step": 17622
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4081574395916244e-05,
+      "loss": 0.4195,
+      "step": 17623
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4081238560609878e-05,
+      "loss": 0.2112,
+      "step": 17624
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4080902725303511e-05,
+      "loss": 0.3091,
+      "step": 17625
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4080566889997147e-05,
+      "loss": 0.29,
+      "step": 17626
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.408023105469078e-05,
+      "loss": 0.2735,
+      "step": 17627
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4079895219384413e-05,
+      "loss": 0.5093,
+      "step": 17628
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.407955938407805e-05,
+      "loss": 0.3164,
+      "step": 17629
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4079223548771683e-05,
+      "loss": 0.2829,
+      "step": 17630
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4078887713465317e-05,
+      "loss": 0.2833,
+      "step": 17631
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4078551878158952e-05,
+      "loss": 0.3247,
+      "step": 17632
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4078216042852587e-05,
+      "loss": 0.3926,
+      "step": 17633
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.407788020754622e-05,
+      "loss": 0.4552,
+      "step": 17634
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4077544372239856e-05,
+      "loss": 0.3688,
+      "step": 17635
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4077208536933489e-05,
+      "loss": 0.3083,
+      "step": 17636
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4076872701627122e-05,
+      "loss": 0.3133,
+      "step": 17637
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4076536866320758e-05,
+      "loss": 0.2895,
+      "step": 17638
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4076201031014391e-05,
+      "loss": 0.3669,
+      "step": 17639
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4075865195708026e-05,
+      "loss": 0.1918,
+      "step": 17640
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.407552936040166e-05,
+      "loss": 0.4091,
+      "step": 17641
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4075193525095295e-05,
+      "loss": 0.4636,
+      "step": 17642
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4074857689788928e-05,
+      "loss": 0.3474,
+      "step": 17643
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4074521854482563e-05,
+      "loss": 0.3429,
+      "step": 17644
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4074186019176197e-05,
+      "loss": 0.1893,
+      "step": 17645
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.407385018386983e-05,
+      "loss": 0.3958,
+      "step": 17646
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4073514348563467e-05,
+      "loss": 0.3525,
+      "step": 17647
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.40731785132571e-05,
+      "loss": 0.3245,
+      "step": 17648
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4072842677950733e-05,
+      "loss": 0.2913,
+      "step": 17649
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4072506842644369e-05,
+      "loss": 0.3776,
+      "step": 17650
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4072171007338002e-05,
+      "loss": 0.2625,
+      "step": 17651
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4071835172031636e-05,
+      "loss": 0.5124,
+      "step": 17652
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4071499336725271e-05,
+      "loss": 0.3563,
+      "step": 17653
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4071163501418906e-05,
+      "loss": 0.3454,
+      "step": 17654
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4070827666112539e-05,
+      "loss": 0.3125,
+      "step": 17655
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4070491830806175e-05,
+      "loss": 0.3333,
+      "step": 17656
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4070155995499808e-05,
+      "loss": 0.2157,
+      "step": 17657
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4069820160193441e-05,
+      "loss": 0.2438,
+      "step": 17658
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4069484324887077e-05,
+      "loss": 0.3299,
+      "step": 17659
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.406914848958071e-05,
+      "loss": 0.3197,
+      "step": 17660
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4068812654274345e-05,
+      "loss": 0.3583,
+      "step": 17661
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.406847681896798e-05,
+      "loss": 0.3159,
+      "step": 17662
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4068140983661614e-05,
+      "loss": 0.4086,
+      "step": 17663
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4067805148355247e-05,
+      "loss": 0.2253,
+      "step": 17664
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4067469313048882e-05,
+      "loss": 0.2186,
+      "step": 17665
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4067133477742516e-05,
+      "loss": 0.3246,
+      "step": 17666
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.406679764243615e-05,
+      "loss": 0.2302,
+      "step": 17667
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4066461807129786e-05,
+      "loss": 0.2627,
+      "step": 17668
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4066125971823419e-05,
+      "loss": 0.3088,
+      "step": 17669
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4065790136517052e-05,
+      "loss": 0.2916,
+      "step": 17670
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4065454301210688e-05,
+      "loss": 0.2605,
+      "step": 17671
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4065118465904321e-05,
+      "loss": 0.2721,
+      "step": 17672
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4064782630597956e-05,
+      "loss": 0.1963,
+      "step": 17673
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.406444679529159e-05,
+      "loss": 0.3558,
+      "step": 17674
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4064110959985225e-05,
+      "loss": 0.1444,
+      "step": 17675
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4063775124678858e-05,
+      "loss": 0.3027,
+      "step": 17676
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4063439289372494e-05,
+      "loss": 0.277,
+      "step": 17677
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4063103454066127e-05,
+      "loss": 0.3107,
+      "step": 17678
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.406276761875976e-05,
+      "loss": 0.3588,
+      "step": 17679
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4062431783453396e-05,
+      "loss": 0.3084,
+      "step": 17680
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.406209594814703e-05,
+      "loss": 0.3245,
+      "step": 17681
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4061760112840664e-05,
+      "loss": 0.2287,
+      "step": 17682
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4061424277534299e-05,
+      "loss": 0.4199,
+      "step": 17683
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4061088442227933e-05,
+      "loss": 0.2583,
+      "step": 17684
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4060752606921566e-05,
+      "loss": 0.355,
+      "step": 17685
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4060416771615203e-05,
+      "loss": 0.3236,
+      "step": 17686
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4060080936308836e-05,
+      "loss": 0.2371,
+      "step": 17687
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4059745101002469e-05,
+      "loss": 0.3431,
+      "step": 17688
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4059409265696105e-05,
+      "loss": 0.3566,
+      "step": 17689
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4059073430389738e-05,
+      "loss": 0.2141,
+      "step": 17690
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.405873759508337e-05,
+      "loss": 0.3973,
+      "step": 17691
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4058401759777007e-05,
+      "loss": 0.2897,
+      "step": 17692
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.405806592447064e-05,
+      "loss": 0.3949,
+      "step": 17693
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4057730089164275e-05,
+      "loss": 0.2511,
+      "step": 17694
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.405739425385791e-05,
+      "loss": 0.2068,
+      "step": 17695
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4057058418551544e-05,
+      "loss": 0.3911,
+      "step": 17696
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4056722583245177e-05,
+      "loss": 0.422,
+      "step": 17697
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4056386747938813e-05,
+      "loss": 0.2903,
+      "step": 17698
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4056050912632446e-05,
+      "loss": 0.2692,
+      "step": 17699
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.405571507732608e-05,
+      "loss": 0.2696,
+      "step": 17700
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4055379242019716e-05,
+      "loss": 0.3499,
+      "step": 17701
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4055043406713349e-05,
+      "loss": 0.4486,
+      "step": 17702
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4054707571406983e-05,
+      "loss": 0.3152,
+      "step": 17703
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4054371736100618e-05,
+      "loss": 0.3409,
+      "step": 17704
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4054035900794252e-05,
+      "loss": 0.1963,
+      "step": 17705
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4053700065487885e-05,
+      "loss": 0.2718,
+      "step": 17706
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4053364230181522e-05,
+      "loss": 0.1888,
+      "step": 17707
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4053028394875155e-05,
+      "loss": 0.2642,
+      "step": 17708
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4052692559568788e-05,
+      "loss": 0.1886,
+      "step": 17709
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4052356724262424e-05,
+      "loss": 0.2453,
+      "step": 17710
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4052020888956057e-05,
+      "loss": 0.5188,
+      "step": 17711
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.405168505364969e-05,
+      "loss": 0.1954,
+      "step": 17712
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4051349218343326e-05,
+      "loss": 0.3727,
+      "step": 17713
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.405101338303696e-05,
+      "loss": 0.5189,
+      "step": 17714
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4050677547730594e-05,
+      "loss": 0.3228,
+      "step": 17715
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4050341712424228e-05,
+      "loss": 0.3885,
+      "step": 17716
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4050005877117863e-05,
+      "loss": 0.3793,
+      "step": 17717
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4049670041811496e-05,
+      "loss": 0.3588,
+      "step": 17718
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4049334206505132e-05,
+      "loss": 0.4815,
+      "step": 17719
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4048998371198765e-05,
+      "loss": 0.1794,
+      "step": 17720
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4048662535892398e-05,
+      "loss": 0.3155,
+      "step": 17721
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4048326700586035e-05,
+      "loss": 0.3033,
+      "step": 17722
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4047990865279668e-05,
+      "loss": 0.1994,
+      "step": 17723
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4047655029973302e-05,
+      "loss": 0.3162,
+      "step": 17724
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4047319194666937e-05,
+      "loss": 0.3068,
+      "step": 17725
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4046983359360572e-05,
+      "loss": 0.2591,
+      "step": 17726
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4046647524054205e-05,
+      "loss": 0.4095,
+      "step": 17727
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4046311688747841e-05,
+      "loss": 0.3127,
+      "step": 17728
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4045975853441474e-05,
+      "loss": 0.5076,
+      "step": 17729
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4045640018135107e-05,
+      "loss": 0.229,
+      "step": 17730
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4045304182828743e-05,
+      "loss": 0.273,
+      "step": 17731
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4044968347522376e-05,
+      "loss": 0.3304,
+      "step": 17732
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.404463251221601e-05,
+      "loss": 0.2656,
+      "step": 17733
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4044296676909645e-05,
+      "loss": 0.4492,
+      "step": 17734
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4043960841603278e-05,
+      "loss": 0.541,
+      "step": 17735
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4043625006296913e-05,
+      "loss": 0.3412,
+      "step": 17736
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4043289170990548e-05,
+      "loss": 0.3243,
+      "step": 17737
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4042953335684182e-05,
+      "loss": 0.2541,
+      "step": 17738
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4042617500377815e-05,
+      "loss": 0.2796,
+      "step": 17739
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4042281665071452e-05,
+      "loss": 0.2352,
+      "step": 17740
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4041945829765085e-05,
+      "loss": 0.4408,
+      "step": 17741
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4041609994458717e-05,
+      "loss": 0.2293,
+      "step": 17742
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4041274159152354e-05,
+      "loss": 0.3909,
+      "step": 17743
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4040938323845987e-05,
+      "loss": 0.4137,
+      "step": 17744
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4040602488539621e-05,
+      "loss": 0.4658,
+      "step": 17745
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4040266653233256e-05,
+      "loss": 0.3642,
+      "step": 17746
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.403993081792689e-05,
+      "loss": 0.3191,
+      "step": 17747
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4039594982620524e-05,
+      "loss": 0.3138,
+      "step": 17748
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.403925914731416e-05,
+      "loss": 0.2002,
+      "step": 17749
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4038923312007793e-05,
+      "loss": 0.2587,
+      "step": 17750
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4038587476701426e-05,
+      "loss": 0.5452,
+      "step": 17751
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4038251641395062e-05,
+      "loss": 0.3988,
+      "step": 17752
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4037915806088695e-05,
+      "loss": 0.3511,
+      "step": 17753
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.403757997078233e-05,
+      "loss": 0.1693,
+      "step": 17754
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4037244135475964e-05,
+      "loss": 0.2513,
+      "step": 17755
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4036908300169597e-05,
+      "loss": 0.3238,
+      "step": 17756
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4036572464863232e-05,
+      "loss": 0.2167,
+      "step": 17757
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4036236629556867e-05,
+      "loss": 0.2714,
+      "step": 17758
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4035900794250501e-05,
+      "loss": 0.2267,
+      "step": 17759
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4035564958944134e-05,
+      "loss": 0.3301,
+      "step": 17760
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.403522912363777e-05,
+      "loss": 0.2771,
+      "step": 17761
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4034893288331404e-05,
+      "loss": 0.2208,
+      "step": 17762
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4034557453025037e-05,
+      "loss": 0.2427,
+      "step": 17763
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4034221617718673e-05,
+      "loss": 0.3519,
+      "step": 17764
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.4033885782412306e-05,
+      "loss": 0.1822,
+      "step": 17765
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.403354994710594e-05,
+      "loss": 0.3507,
+      "step": 17766
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4033214111799575e-05,
+      "loss": 0.4581,
+      "step": 17767
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.403287827649321e-05,
+      "loss": 0.2336,
+      "step": 17768
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4032542441186843e-05,
+      "loss": 0.3566,
+      "step": 17769
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4032206605880479e-05,
+      "loss": 0.3321,
+      "step": 17770
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4031870770574112e-05,
+      "loss": 0.2456,
+      "step": 17771
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4031534935267745e-05,
+      "loss": 0.3091,
+      "step": 17772
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4031199099961381e-05,
+      "loss": 0.3279,
+      "step": 17773
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4030863264655014e-05,
+      "loss": 0.3279,
+      "step": 17774
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4030527429348649e-05,
+      "loss": 0.3362,
+      "step": 17775
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4030191594042282e-05,
+      "loss": 0.3369,
+      "step": 17776
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4029855758735918e-05,
+      "loss": 0.392,
+      "step": 17777
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4029519923429551e-05,
+      "loss": 0.2107,
+      "step": 17778
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4029184088123184e-05,
+      "loss": 0.3133,
+      "step": 17779
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.402884825281682e-05,
+      "loss": 0.2769,
+      "step": 17780
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4028512417510453e-05,
+      "loss": 0.2261,
+      "step": 17781
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4028176582204086e-05,
+      "loss": 0.1893,
+      "step": 17782
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4027840746897723e-05,
+      "loss": 0.2189,
+      "step": 17783
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4027504911591356e-05,
+      "loss": 0.3269,
+      "step": 17784
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.402716907628499e-05,
+      "loss": 0.3039,
+      "step": 17785
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4026833240978625e-05,
+      "loss": 0.3401,
+      "step": 17786
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.402649740567226e-05,
+      "loss": 0.2895,
+      "step": 17787
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4026161570365893e-05,
+      "loss": 0.2763,
+      "step": 17788
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4025825735059529e-05,
+      "loss": 0.2347,
+      "step": 17789
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4025489899753162e-05,
+      "loss": 0.2479,
+      "step": 17790
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4025154064446795e-05,
+      "loss": 0.262,
+      "step": 17791
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4024818229140431e-05,
+      "loss": 0.391,
+      "step": 17792
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4024482393834064e-05,
+      "loss": 0.4739,
+      "step": 17793
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4024146558527699e-05,
+      "loss": 0.4919,
+      "step": 17794
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4023810723221333e-05,
+      "loss": 0.2599,
+      "step": 17795
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4023474887914968e-05,
+      "loss": 0.3459,
+      "step": 17796
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4023139052608601e-05,
+      "loss": 0.6338,
+      "step": 17797
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4022803217302237e-05,
+      "loss": 0.3561,
+      "step": 17798
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.402246738199587e-05,
+      "loss": 0.2844,
+      "step": 17799
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4022131546689503e-05,
+      "loss": 0.2158,
+      "step": 17800
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.402179571138314e-05,
+      "loss": 0.2941,
+      "step": 17801
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4021459876076773e-05,
+      "loss": 0.3434,
+      "step": 17802
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4021124040770406e-05,
+      "loss": 0.3515,
+      "step": 17803
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4020788205464042e-05,
+      "loss": 0.1461,
+      "step": 17804
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4020452370157675e-05,
+      "loss": 0.3417,
+      "step": 17805
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.402011653485131e-05,
+      "loss": 0.3216,
+      "step": 17806
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4019780699544944e-05,
+      "loss": 0.3239,
+      "step": 17807
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4019444864238579e-05,
+      "loss": 0.1847,
+      "step": 17808
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4019109028932212e-05,
+      "loss": 0.221,
+      "step": 17809
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4018773193625848e-05,
+      "loss": 0.3398,
+      "step": 17810
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4018437358319481e-05,
+      "loss": 0.3887,
+      "step": 17811
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4018101523013114e-05,
+      "loss": 0.3167,
+      "step": 17812
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.401776568770675e-05,
+      "loss": 0.4214,
+      "step": 17813
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4017429852400383e-05,
+      "loss": 0.466,
+      "step": 17814
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4017094017094018e-05,
+      "loss": 0.3461,
+      "step": 17815
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4016758181787653e-05,
+      "loss": 0.1417,
+      "step": 17816
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4016422346481287e-05,
+      "loss": 0.2509,
+      "step": 17817
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.401608651117492e-05,
+      "loss": 0.3061,
+      "step": 17818
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4015750675868557e-05,
+      "loss": 0.2296,
+      "step": 17819
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.401541484056219e-05,
+      "loss": 0.2296,
+      "step": 17820
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4015079005255822e-05,
+      "loss": 0.2451,
+      "step": 17821
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4014743169949459e-05,
+      "loss": 0.223,
+      "step": 17822
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4014407334643092e-05,
+      "loss": 0.4272,
+      "step": 17823
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4014071499336725e-05,
+      "loss": 0.2512,
+      "step": 17824
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4013735664030361e-05,
+      "loss": 0.3856,
+      "step": 17825
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4013399828723994e-05,
+      "loss": 0.3284,
+      "step": 17826
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4013063993417629e-05,
+      "loss": 0.4998,
+      "step": 17827
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4012728158111263e-05,
+      "loss": 0.296,
+      "step": 17828
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4012392322804898e-05,
+      "loss": 0.2966,
+      "step": 17829
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4012056487498531e-05,
+      "loss": 0.2827,
+      "step": 17830
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4011720652192167e-05,
+      "loss": 0.2321,
+      "step": 17831
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.40113848168858e-05,
+      "loss": 0.2946,
+      "step": 17832
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4011048981579433e-05,
+      "loss": 0.2648,
+      "step": 17833
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.401071314627307e-05,
+      "loss": 0.435,
+      "step": 17834
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4010377310966702e-05,
+      "loss": 0.3819,
+      "step": 17835
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4010041475660337e-05,
+      "loss": 0.398,
+      "step": 17836
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4009705640353972e-05,
+      "loss": 0.3296,
+      "step": 17837
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4009369805047606e-05,
+      "loss": 0.2333,
+      "step": 17838
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.400903396974124e-05,
+      "loss": 0.1961,
+      "step": 17839
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4008698134434876e-05,
+      "loss": 0.2753,
+      "step": 17840
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4008362299128509e-05,
+      "loss": 0.3739,
+      "step": 17841
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4008026463822142e-05,
+      "loss": 0.4155,
+      "step": 17842
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4007690628515778e-05,
+      "loss": 0.4691,
+      "step": 17843
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.400735479320941e-05,
+      "loss": 0.3477,
+      "step": 17844
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4007018957903045e-05,
+      "loss": 0.3304,
+      "step": 17845
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.400668312259668e-05,
+      "loss": 0.3172,
+      "step": 17846
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4006347287290313e-05,
+      "loss": 0.3091,
+      "step": 17847
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4006011451983948e-05,
+      "loss": 0.3053,
+      "step": 17848
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4005675616677582e-05,
+      "loss": 0.3049,
+      "step": 17849
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4005339781371217e-05,
+      "loss": 0.2657,
+      "step": 17850
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.400500394606485e-05,
+      "loss": 0.3152,
+      "step": 17851
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4004668110758486e-05,
+      "loss": 0.3946,
+      "step": 17852
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.400433227545212e-05,
+      "loss": 0.3449,
+      "step": 17853
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4003996440145752e-05,
+      "loss": 0.3216,
+      "step": 17854
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4003660604839389e-05,
+      "loss": 0.3564,
+      "step": 17855
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4003324769533022e-05,
+      "loss": 0.2234,
+      "step": 17856
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4002988934226656e-05,
+      "loss": 0.2211,
+      "step": 17857
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.400265309892029e-05,
+      "loss": 0.3487,
+      "step": 17858
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4002317263613925e-05,
+      "loss": 0.2973,
+      "step": 17859
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4001981428307558e-05,
+      "loss": 0.2245,
+      "step": 17860
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4001645593001195e-05,
+      "loss": 0.4593,
+      "step": 17861
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4001309757694828e-05,
+      "loss": 0.262,
+      "step": 17862
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.400097392238846e-05,
+      "loss": 0.3379,
+      "step": 17863
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.4000638087082097e-05,
+      "loss": 0.5081,
+      "step": 17864
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.400030225177573e-05,
+      "loss": 0.2172,
+      "step": 17865
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3999966416469365e-05,
+      "loss": 0.3614,
+      "step": 17866
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3999630581163e-05,
+      "loss": 0.2766,
+      "step": 17867
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3999294745856634e-05,
+      "loss": 0.3678,
+      "step": 17868
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3998958910550267e-05,
+      "loss": 0.315,
+      "step": 17869
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3998623075243902e-05,
+      "loss": 0.346,
+      "step": 17870
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3998287239937536e-05,
+      "loss": 0.2423,
+      "step": 17871
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3997951404631169e-05,
+      "loss": 0.2795,
+      "step": 17872
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3997615569324805e-05,
+      "loss": 0.3037,
+      "step": 17873
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3997279734018438e-05,
+      "loss": 0.3803,
+      "step": 17874
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3996943898712071e-05,
+      "loss": 0.5081,
+      "step": 17875
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3996608063405708e-05,
+      "loss": 0.2355,
+      "step": 17876
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.399627222809934e-05,
+      "loss": 0.2052,
+      "step": 17877
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3995936392792975e-05,
+      "loss": 0.2361,
+      "step": 17878
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.399560055748661e-05,
+      "loss": 0.1552,
+      "step": 17879
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3995264722180245e-05,
+      "loss": 0.1893,
+      "step": 17880
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3994928886873878e-05,
+      "loss": 0.4096,
+      "step": 17881
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3994593051567514e-05,
+      "loss": 0.3367,
+      "step": 17882
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3994257216261147e-05,
+      "loss": 0.3803,
+      "step": 17883
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.399392138095478e-05,
+      "loss": 0.3644,
+      "step": 17884
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3993585545648416e-05,
+      "loss": 0.3803,
+      "step": 17885
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3993249710342049e-05,
+      "loss": 0.2106,
+      "step": 17886
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3992913875035684e-05,
+      "loss": 0.3128,
+      "step": 17887
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3992578039729318e-05,
+      "loss": 0.4899,
+      "step": 17888
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3992242204422953e-05,
+      "loss": 0.2664,
+      "step": 17889
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3991906369116586e-05,
+      "loss": 0.268,
+      "step": 17890
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.399157053381022e-05,
+      "loss": 0.4173,
+      "step": 17891
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3991234698503855e-05,
+      "loss": 0.3278,
+      "step": 17892
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3990898863197488e-05,
+      "loss": 0.27,
+      "step": 17893
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3990563027891125e-05,
+      "loss": 0.4057,
+      "step": 17894
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3990227192584758e-05,
+      "loss": 0.283,
+      "step": 17895
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.398989135727839e-05,
+      "loss": 0.3044,
+      "step": 17896
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3989555521972027e-05,
+      "loss": 0.3501,
+      "step": 17897
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.398921968666566e-05,
+      "loss": 0.2786,
+      "step": 17898
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3988883851359294e-05,
+      "loss": 0.2427,
+      "step": 17899
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3988548016052929e-05,
+      "loss": 0.4466,
+      "step": 17900
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3988212180746564e-05,
+      "loss": 0.4022,
+      "step": 17901
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3987876345440197e-05,
+      "loss": 0.3019,
+      "step": 17902
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3987540510133833e-05,
+      "loss": 0.3953,
+      "step": 17903
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3987204674827466e-05,
+      "loss": 0.3969,
+      "step": 17904
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3986868839521099e-05,
+      "loss": 0.2961,
+      "step": 17905
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3986533004214735e-05,
+      "loss": 0.2667,
+      "step": 17906
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3986197168908368e-05,
+      "loss": 0.2951,
+      "step": 17907
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3985861333602003e-05,
+      "loss": 0.3745,
+      "step": 17908
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3985525498295638e-05,
+      "loss": 0.4185,
+      "step": 17909
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3985189662989272e-05,
+      "loss": 0.4175,
+      "step": 17910
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3984853827682905e-05,
+      "loss": 0.2011,
+      "step": 17911
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3984517992376541e-05,
+      "loss": 0.237,
+      "step": 17912
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3984182157070174e-05,
+      "loss": 0.4016,
+      "step": 17913
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3983846321763807e-05,
+      "loss": 0.3648,
+      "step": 17914
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3983510486457444e-05,
+      "loss": 0.2563,
+      "step": 17915
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3983174651151077e-05,
+      "loss": 0.2161,
+      "step": 17916
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.398283881584471e-05,
+      "loss": 0.2207,
+      "step": 17917
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3982502980538346e-05,
+      "loss": 0.2851,
+      "step": 17918
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3982167145231979e-05,
+      "loss": 0.3929,
+      "step": 17919
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3981831309925614e-05,
+      "loss": 0.4259,
+      "step": 17920
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3981495474619248e-05,
+      "loss": 0.3831,
+      "step": 17921
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3981159639312883e-05,
+      "loss": 0.3961,
+      "step": 17922
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3980823804006516e-05,
+      "loss": 0.3755,
+      "step": 17923
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3980487968700152e-05,
+      "loss": 0.3694,
+      "step": 17924
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3980152133393785e-05,
+      "loss": 0.3962,
+      "step": 17925
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3979816298087418e-05,
+      "loss": 0.3088,
+      "step": 17926
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3979480462781054e-05,
+      "loss": 0.2511,
+      "step": 17927
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3979144627474687e-05,
+      "loss": 0.2438,
+      "step": 17928
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3978808792168322e-05,
+      "loss": 0.2122,
+      "step": 17929
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3978472956861957e-05,
+      "loss": 0.3453,
+      "step": 17930
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3978137121555591e-05,
+      "loss": 0.1552,
+      "step": 17931
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3977801286249224e-05,
+      "loss": 0.3939,
+      "step": 17932
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.397746545094286e-05,
+      "loss": 0.2935,
+      "step": 17933
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3977129615636494e-05,
+      "loss": 0.3471,
+      "step": 17934
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3976793780330126e-05,
+      "loss": 0.1975,
+      "step": 17935
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3976457945023763e-05,
+      "loss": 0.3255,
+      "step": 17936
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3976122109717396e-05,
+      "loss": 0.2611,
+      "step": 17937
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3975786274411029e-05,
+      "loss": 0.2952,
+      "step": 17938
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3975450439104665e-05,
+      "loss": 0.314,
+      "step": 17939
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3975114603798298e-05,
+      "loss": 0.3642,
+      "step": 17940
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3974778768491933e-05,
+      "loss": 0.205,
+      "step": 17941
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3974442933185567e-05,
+      "loss": 0.2158,
+      "step": 17942
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3974107097879202e-05,
+      "loss": 0.548,
+      "step": 17943
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3973771262572835e-05,
+      "loss": 0.2503,
+      "step": 17944
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3973435427266471e-05,
+      "loss": 0.2218,
+      "step": 17945
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3973099591960104e-05,
+      "loss": 0.3965,
+      "step": 17946
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3972763756653737e-05,
+      "loss": 0.3981,
+      "step": 17947
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3972427921347374e-05,
+      "loss": 0.3077,
+      "step": 17948
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3972092086041006e-05,
+      "loss": 0.2621,
+      "step": 17949
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3971756250734641e-05,
+      "loss": 0.295,
+      "step": 17950
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3971420415428276e-05,
+      "loss": 0.3191,
+      "step": 17951
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.397108458012191e-05,
+      "loss": 0.4092,
+      "step": 17952
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3970748744815543e-05,
+      "loss": 0.4027,
+      "step": 17953
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.397041290950918e-05,
+      "loss": 0.2453,
+      "step": 17954
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3970077074202813e-05,
+      "loss": 0.1695,
+      "step": 17955
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3969741238896446e-05,
+      "loss": 0.3297,
+      "step": 17956
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.396940540359008e-05,
+      "loss": 0.3125,
+      "step": 17957
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3969069568283715e-05,
+      "loss": 0.1757,
+      "step": 17958
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3968733732977348e-05,
+      "loss": 0.3728,
+      "step": 17959
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3968397897670983e-05,
+      "loss": 0.3717,
+      "step": 17960
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3968062062364617e-05,
+      "loss": 0.5198,
+      "step": 17961
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3967726227058252e-05,
+      "loss": 0.3281,
+      "step": 17962
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3967390391751885e-05,
+      "loss": 0.2044,
+      "step": 17963
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3967054556445521e-05,
+      "loss": 0.3019,
+      "step": 17964
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.3966718721139154e-05,
+      "loss": 0.3668,
+      "step": 17965
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3966382885832787e-05,
+      "loss": 0.2674,
+      "step": 17966
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3966047050526423e-05,
+      "loss": 0.3538,
+      "step": 17967
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3965711215220056e-05,
+      "loss": 0.2522,
+      "step": 17968
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3965375379913691e-05,
+      "loss": 0.3638,
+      "step": 17969
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3965039544607326e-05,
+      "loss": 0.2378,
+      "step": 17970
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.396470370930096e-05,
+      "loss": 0.2405,
+      "step": 17971
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3964367873994593e-05,
+      "loss": 0.2825,
+      "step": 17972
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.396403203868823e-05,
+      "loss": 0.3672,
+      "step": 17973
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3963696203381862e-05,
+      "loss": 0.3866,
+      "step": 17974
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3963360368075495e-05,
+      "loss": 0.3992,
+      "step": 17975
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3963024532769132e-05,
+      "loss": 0.2933,
+      "step": 17976
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3962688697462765e-05,
+      "loss": 0.2486,
+      "step": 17977
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.39623528621564e-05,
+      "loss": 0.1317,
+      "step": 17978
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3962017026850034e-05,
+      "loss": 0.3248,
+      "step": 17979
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3961681191543669e-05,
+      "loss": 0.3438,
+      "step": 17980
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3961345356237302e-05,
+      "loss": 0.5458,
+      "step": 17981
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3961009520930936e-05,
+      "loss": 0.2998,
+      "step": 17982
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3960673685624571e-05,
+      "loss": 0.2689,
+      "step": 17983
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3960337850318204e-05,
+      "loss": 0.2943,
+      "step": 17984
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.396000201501184e-05,
+      "loss": 0.3178,
+      "step": 17985
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3959666179705473e-05,
+      "loss": 0.2546,
+      "step": 17986
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3959330344399106e-05,
+      "loss": 0.5196,
+      "step": 17987
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3958994509092742e-05,
+      "loss": 0.3997,
+      "step": 17988
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3958658673786375e-05,
+      "loss": 0.4881,
+      "step": 17989
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.395832283848001e-05,
+      "loss": 0.3369,
+      "step": 17990
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3957987003173645e-05,
+      "loss": 0.2921,
+      "step": 17991
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.395765116786728e-05,
+      "loss": 0.1614,
+      "step": 17992
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3957315332560912e-05,
+      "loss": 0.2933,
+      "step": 17993
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3956979497254549e-05,
+      "loss": 0.1753,
+      "step": 17994
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3956643661948182e-05,
+      "loss": 0.443,
+      "step": 17995
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3956307826641815e-05,
+      "loss": 0.3865,
+      "step": 17996
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3955971991335451e-05,
+      "loss": 0.3681,
+      "step": 17997
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3955636156029084e-05,
+      "loss": 0.2399,
+      "step": 17998
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3955300320722719e-05,
+      "loss": 0.2632,
+      "step": 17999
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3954964485416353e-05,
+      "loss": 0.1639,
+      "step": 18000
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3954628650109988e-05,
+      "loss": 0.2076,
+      "step": 18001
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.395429281480362e-05,
+      "loss": 0.48,
+      "step": 18002
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3953956979497257e-05,
+      "loss": 0.4144,
+      "step": 18003
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.395362114419089e-05,
+      "loss": 0.4897,
+      "step": 18004
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3953285308884523e-05,
+      "loss": 0.2824,
+      "step": 18005
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.395294947357816e-05,
+      "loss": 0.3469,
+      "step": 18006
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3952613638271792e-05,
+      "loss": 0.3003,
+      "step": 18007
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3952277802965425e-05,
+      "loss": 0.247,
+      "step": 18008
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3951941967659062e-05,
+      "loss": 0.4627,
+      "step": 18009
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3951606132352695e-05,
+      "loss": 0.208,
+      "step": 18010
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.395127029704633e-05,
+      "loss": 0.525,
+      "step": 18011
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3950934461739964e-05,
+      "loss": 0.2233,
+      "step": 18012
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3950598626433598e-05,
+      "loss": 0.2606,
+      "step": 18013
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3950262791127231e-05,
+      "loss": 0.4683,
+      "step": 18014
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3949926955820868e-05,
+      "loss": 0.2587,
+      "step": 18015
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.39495911205145e-05,
+      "loss": 0.4974,
+      "step": 18016
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3949255285208134e-05,
+      "loss": 0.4142,
+      "step": 18017
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.394891944990177e-05,
+      "loss": 0.4007,
+      "step": 18018
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3948583614595403e-05,
+      "loss": 0.3974,
+      "step": 18019
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3948247779289038e-05,
+      "loss": 0.2379,
+      "step": 18020
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3947911943982672e-05,
+      "loss": 0.4484,
+      "step": 18021
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3947576108676307e-05,
+      "loss": 0.3086,
+      "step": 18022
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.394724027336994e-05,
+      "loss": 0.4002,
+      "step": 18023
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3946904438063576e-05,
+      "loss": 0.3516,
+      "step": 18024
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.394656860275721e-05,
+      "loss": 0.3165,
+      "step": 18025
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3946232767450842e-05,
+      "loss": 0.3993,
+      "step": 18026
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3945896932144478e-05,
+      "loss": 0.4969,
+      "step": 18027
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3945561096838111e-05,
+      "loss": 0.3628,
+      "step": 18028
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3945225261531744e-05,
+      "loss": 0.2886,
+      "step": 18029
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.394488942622538e-05,
+      "loss": 0.4462,
+      "step": 18030
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3944553590919014e-05,
+      "loss": 0.3072,
+      "step": 18031
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3944217755612648e-05,
+      "loss": 0.4583,
+      "step": 18032
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3943881920306283e-05,
+      "loss": 0.3791,
+      "step": 18033
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3943546084999918e-05,
+      "loss": 0.259,
+      "step": 18034
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.394321024969355e-05,
+      "loss": 0.3643,
+      "step": 18035
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3942874414387187e-05,
+      "loss": 0.3059,
+      "step": 18036
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.394253857908082e-05,
+      "loss": 0.4478,
+      "step": 18037
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3942202743774453e-05,
+      "loss": 0.2833,
+      "step": 18038
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3941866908468089e-05,
+      "loss": 0.2652,
+      "step": 18039
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3941531073161722e-05,
+      "loss": 0.6772,
+      "step": 18040
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3941195237855357e-05,
+      "loss": 0.2182,
+      "step": 18041
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3940859402548991e-05,
+      "loss": 0.2792,
+      "step": 18042
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3940523567242626e-05,
+      "loss": 0.4775,
+      "step": 18043
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3940187731936259e-05,
+      "loss": 0.291,
+      "step": 18044
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3939851896629895e-05,
+      "loss": 0.2843,
+      "step": 18045
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3939516061323528e-05,
+      "loss": 0.2793,
+      "step": 18046
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3939180226017161e-05,
+      "loss": 0.2843,
+      "step": 18047
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3938844390710798e-05,
+      "loss": 0.2888,
+      "step": 18048
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.393850855540443e-05,
+      "loss": 0.3694,
+      "step": 18049
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3938172720098063e-05,
+      "loss": 0.4183,
+      "step": 18050
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.39378368847917e-05,
+      "loss": 0.2505,
+      "step": 18051
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3937501049485333e-05,
+      "loss": 0.3169,
+      "step": 18052
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3937165214178967e-05,
+      "loss": 0.2812,
+      "step": 18053
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3936829378872602e-05,
+      "loss": 0.287,
+      "step": 18054
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3936493543566237e-05,
+      "loss": 0.4026,
+      "step": 18055
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.393615770825987e-05,
+      "loss": 0.2777,
+      "step": 18056
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3935821872953506e-05,
+      "loss": 0.2563,
+      "step": 18057
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3935486037647139e-05,
+      "loss": 0.3268,
+      "step": 18058
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3935150202340772e-05,
+      "loss": 0.3134,
+      "step": 18059
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3934814367034408e-05,
+      "loss": 0.2746,
+      "step": 18060
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3934478531728041e-05,
+      "loss": 0.3695,
+      "step": 18061
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3934142696421676e-05,
+      "loss": 0.2572,
+      "step": 18062
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.393380686111531e-05,
+      "loss": 0.2367,
+      "step": 18063
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3933471025808945e-05,
+      "loss": 0.2381,
+      "step": 18064
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3933135190502578e-05,
+      "loss": 0.4245,
+      "step": 18065
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3932799355196214e-05,
+      "loss": 0.3732,
+      "step": 18066
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3932463519889847e-05,
+      "loss": 0.2376,
+      "step": 18067
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.393212768458348e-05,
+      "loss": 0.2505,
+      "step": 18068
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3931791849277117e-05,
+      "loss": 0.3088,
+      "step": 18069
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.393145601397075e-05,
+      "loss": 0.1488,
+      "step": 18070
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3931120178664384e-05,
+      "loss": 0.2372,
+      "step": 18071
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3930784343358019e-05,
+      "loss": 0.4854,
+      "step": 18072
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3930448508051652e-05,
+      "loss": 0.1896,
+      "step": 18073
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3930112672745287e-05,
+      "loss": 0.3142,
+      "step": 18074
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3929776837438921e-05,
+      "loss": 0.491,
+      "step": 18075
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3929441002132556e-05,
+      "loss": 0.284,
+      "step": 18076
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3929105166826189e-05,
+      "loss": 0.3372,
+      "step": 18077
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3928769331519825e-05,
+      "loss": 0.4315,
+      "step": 18078
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3928433496213458e-05,
+      "loss": 0.2984,
+      "step": 18079
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3928097660907091e-05,
+      "loss": 0.293,
+      "step": 18080
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3927761825600727e-05,
+      "loss": 0.3937,
+      "step": 18081
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.392742599029436e-05,
+      "loss": 0.3923,
+      "step": 18082
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3927090154987995e-05,
+      "loss": 0.4212,
+      "step": 18083
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.392675431968163e-05,
+      "loss": 0.3951,
+      "step": 18084
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3926418484375264e-05,
+      "loss": 0.3543,
+      "step": 18085
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3926082649068897e-05,
+      "loss": 0.3562,
+      "step": 18086
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3925746813762534e-05,
+      "loss": 0.3479,
+      "step": 18087
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3925410978456167e-05,
+      "loss": 0.3125,
+      "step": 18088
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.39250751431498e-05,
+      "loss": 0.3127,
+      "step": 18089
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3924739307843436e-05,
+      "loss": 0.4716,
+      "step": 18090
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3924403472537069e-05,
+      "loss": 0.2032,
+      "step": 18091
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3924067637230703e-05,
+      "loss": 0.2038,
+      "step": 18092
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3923731801924338e-05,
+      "loss": 0.388,
+      "step": 18093
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3923395966617971e-05,
+      "loss": 0.2489,
+      "step": 18094
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3923060131311606e-05,
+      "loss": 0.229,
+      "step": 18095
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.392272429600524e-05,
+      "loss": 0.2962,
+      "step": 18096
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3922388460698875e-05,
+      "loss": 0.3751,
+      "step": 18097
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3922052625392508e-05,
+      "loss": 0.1922,
+      "step": 18098
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3921716790086144e-05,
+      "loss": 0.4988,
+      "step": 18099
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3921380954779777e-05,
+      "loss": 0.2579,
+      "step": 18100
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.392104511947341e-05,
+      "loss": 0.4106,
+      "step": 18101
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3920709284167047e-05,
+      "loss": 0.2389,
+      "step": 18102
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.392037344886068e-05,
+      "loss": 0.3456,
+      "step": 18103
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3920037613554314e-05,
+      "loss": 0.4726,
+      "step": 18104
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3919701778247949e-05,
+      "loss": 0.3197,
+      "step": 18105
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3919365942941583e-05,
+      "loss": 0.2716,
+      "step": 18106
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3919030107635216e-05,
+      "loss": 0.2495,
+      "step": 18107
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3918694272328853e-05,
+      "loss": 0.2665,
+      "step": 18108
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3918358437022486e-05,
+      "loss": 0.2998,
+      "step": 18109
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3918022601716119e-05,
+      "loss": 0.2986,
+      "step": 18110
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3917686766409755e-05,
+      "loss": 0.5252,
+      "step": 18111
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3917350931103388e-05,
+      "loss": 0.2328,
+      "step": 18112
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3917015095797023e-05,
+      "loss": 0.2849,
+      "step": 18113
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3916679260490657e-05,
+      "loss": 0.2957,
+      "step": 18114
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3916343425184292e-05,
+      "loss": 0.3135,
+      "step": 18115
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3916007589877925e-05,
+      "loss": 0.3439,
+      "step": 18116
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.391567175457156e-05,
+      "loss": 0.2172,
+      "step": 18117
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3915335919265194e-05,
+      "loss": 0.2379,
+      "step": 18118
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3915000083958827e-05,
+      "loss": 0.3668,
+      "step": 18119
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3914664248652463e-05,
+      "loss": 0.311,
+      "step": 18120
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3914328413346096e-05,
+      "loss": 0.2901,
+      "step": 18121
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.391399257803973e-05,
+      "loss": 0.4233,
+      "step": 18122
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3913656742733366e-05,
+      "loss": 0.3701,
+      "step": 18123
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3913320907426999e-05,
+      "loss": 0.3949,
+      "step": 18124
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3912985072120633e-05,
+      "loss": 0.1835,
+      "step": 18125
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3912649236814268e-05,
+      "loss": 0.2873,
+      "step": 18126
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3912313401507903e-05,
+      "loss": 0.4018,
+      "step": 18127
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3911977566201535e-05,
+      "loss": 0.305,
+      "step": 18128
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3911641730895172e-05,
+      "loss": 0.288,
+      "step": 18129
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3911305895588805e-05,
+      "loss": 0.2865,
+      "step": 18130
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3910970060282438e-05,
+      "loss": 0.3721,
+      "step": 18131
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3910634224976074e-05,
+      "loss": 0.2857,
+      "step": 18132
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3910298389669707e-05,
+      "loss": 0.4173,
+      "step": 18133
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3909962554363342e-05,
+      "loss": 0.3582,
+      "step": 18134
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3909626719056976e-05,
+      "loss": 0.2862,
+      "step": 18135
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3909290883750611e-05,
+      "loss": 0.4197,
+      "step": 18136
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3908955048444244e-05,
+      "loss": 0.3607,
+      "step": 18137
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3908619213137877e-05,
+      "loss": 0.4827,
+      "step": 18138
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3908283377831513e-05,
+      "loss": 0.4915,
+      "step": 18139
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3907947542525146e-05,
+      "loss": 0.2419,
+      "step": 18140
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3907611707218779e-05,
+      "loss": 0.4378,
+      "step": 18141
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3907275871912415e-05,
+      "loss": 0.4039,
+      "step": 18142
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3906940036606048e-05,
+      "loss": 0.2279,
+      "step": 18143
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3906604201299683e-05,
+      "loss": 0.3493,
+      "step": 18144
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3906268365993318e-05,
+      "loss": 0.4278,
+      "step": 18145
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3905932530686952e-05,
+      "loss": 0.265,
+      "step": 18146
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3905596695380585e-05,
+      "loss": 0.2765,
+      "step": 18147
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3905260860074222e-05,
+      "loss": 0.2718,
+      "step": 18148
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3904925024767855e-05,
+      "loss": 0.2166,
+      "step": 18149
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3904589189461488e-05,
+      "loss": 0.1407,
+      "step": 18150
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3904253354155124e-05,
+      "loss": 0.2805,
+      "step": 18151
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3903917518848757e-05,
+      "loss": 0.3553,
+      "step": 18152
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3903581683542392e-05,
+      "loss": 0.3972,
+      "step": 18153
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3903245848236026e-05,
+      "loss": 0.3853,
+      "step": 18154
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.390291001292966e-05,
+      "loss": 0.3125,
+      "step": 18155
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3902574177623294e-05,
+      "loss": 0.3806,
+      "step": 18156
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.390223834231693e-05,
+      "loss": 0.236,
+      "step": 18157
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3901902507010563e-05,
+      "loss": 0.3852,
+      "step": 18158
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3901566671704196e-05,
+      "loss": 0.3728,
+      "step": 18159
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3901230836397832e-05,
+      "loss": 0.2837,
+      "step": 18160
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3900895001091465e-05,
+      "loss": 0.328,
+      "step": 18161
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.39005591657851e-05,
+      "loss": 0.3633,
+      "step": 18162
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.3900223330478735e-05,
+      "loss": 0.3279,
+      "step": 18163
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3899887495172368e-05,
+      "loss": 0.2525,
+      "step": 18164
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3899551659866002e-05,
+      "loss": 0.3586,
+      "step": 18165
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3899215824559637e-05,
+      "loss": 0.29,
+      "step": 18166
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3898879989253272e-05,
+      "loss": 0.2172,
+      "step": 18167
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3898544153946904e-05,
+      "loss": 0.2863,
+      "step": 18168
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.389820831864054e-05,
+      "loss": 0.3958,
+      "step": 18169
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3897872483334174e-05,
+      "loss": 0.3494,
+      "step": 18170
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3897536648027807e-05,
+      "loss": 0.3877,
+      "step": 18171
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3897200812721443e-05,
+      "loss": 0.3082,
+      "step": 18172
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3896864977415076e-05,
+      "loss": 0.271,
+      "step": 18173
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.389652914210871e-05,
+      "loss": 0.4383,
+      "step": 18174
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3896193306802345e-05,
+      "loss": 0.439,
+      "step": 18175
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.389585747149598e-05,
+      "loss": 0.4506,
+      "step": 18176
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3895521636189613e-05,
+      "loss": 0.4414,
+      "step": 18177
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.389518580088325e-05,
+      "loss": 0.3608,
+      "step": 18178
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3894849965576882e-05,
+      "loss": 0.3704,
+      "step": 18179
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3894514130270515e-05,
+      "loss": 0.4398,
+      "step": 18180
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3894178294964151e-05,
+      "loss": 0.3169,
+      "step": 18181
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3893842459657784e-05,
+      "loss": 0.3938,
+      "step": 18182
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3893506624351419e-05,
+      "loss": 0.4018,
+      "step": 18183
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3893170789045054e-05,
+      "loss": 0.2707,
+      "step": 18184
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3892834953738687e-05,
+      "loss": 0.4296,
+      "step": 18185
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3892499118432321e-05,
+      "loss": 0.419,
+      "step": 18186
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3892163283125956e-05,
+      "loss": 0.3072,
+      "step": 18187
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.389182744781959e-05,
+      "loss": 0.4127,
+      "step": 18188
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3891491612513224e-05,
+      "loss": 0.26,
+      "step": 18189
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.389115577720686e-05,
+      "loss": 0.3806,
+      "step": 18190
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3890819941900493e-05,
+      "loss": 0.1793,
+      "step": 18191
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3890484106594126e-05,
+      "loss": 0.2012,
+      "step": 18192
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3890148271287762e-05,
+      "loss": 0.2647,
+      "step": 18193
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3889812435981395e-05,
+      "loss": 0.3124,
+      "step": 18194
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.388947660067503e-05,
+      "loss": 0.3925,
+      "step": 18195
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3889140765368664e-05,
+      "loss": 0.2608,
+      "step": 18196
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3888804930062299e-05,
+      "loss": 0.2785,
+      "step": 18197
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3888469094755932e-05,
+      "loss": 0.2984,
+      "step": 18198
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3888133259449568e-05,
+      "loss": 0.3514,
+      "step": 18199
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3887797424143201e-05,
+      "loss": 0.1758,
+      "step": 18200
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3887461588836834e-05,
+      "loss": 0.3296,
+      "step": 18201
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.388712575353047e-05,
+      "loss": 0.3106,
+      "step": 18202
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3886789918224104e-05,
+      "loss": 0.4134,
+      "step": 18203
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3886454082917738e-05,
+      "loss": 0.2682,
+      "step": 18204
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3886118247611373e-05,
+      "loss": 0.4536,
+      "step": 18205
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3885782412305008e-05,
+      "loss": 0.2231,
+      "step": 18206
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.388544657699864e-05,
+      "loss": 0.4132,
+      "step": 18207
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3885110741692275e-05,
+      "loss": 0.3697,
+      "step": 18208
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.388477490638591e-05,
+      "loss": 0.2208,
+      "step": 18209
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3884439071079543e-05,
+      "loss": 0.5343,
+      "step": 18210
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3884103235773179e-05,
+      "loss": 0.4473,
+      "step": 18211
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3883767400466812e-05,
+      "loss": 0.2706,
+      "step": 18212
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3883431565160445e-05,
+      "loss": 0.3879,
+      "step": 18213
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3883095729854081e-05,
+      "loss": 0.3315,
+      "step": 18214
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3882759894547714e-05,
+      "loss": 0.2221,
+      "step": 18215
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3882424059241349e-05,
+      "loss": 0.2601,
+      "step": 18216
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3882088223934984e-05,
+      "loss": 0.3226,
+      "step": 18217
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3881752388628618e-05,
+      "loss": 0.2626,
+      "step": 18218
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3881416553322251e-05,
+      "loss": 0.4264,
+      "step": 18219
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3881080718015887e-05,
+      "loss": 0.3464,
+      "step": 18220
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.388074488270952e-05,
+      "loss": 0.1842,
+      "step": 18221
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3880409047403153e-05,
+      "loss": 0.3322,
+      "step": 18222
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.388007321209679e-05,
+      "loss": 0.3124,
+      "step": 18223
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3879737376790423e-05,
+      "loss": 0.2214,
+      "step": 18224
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3879401541484057e-05,
+      "loss": 0.1559,
+      "step": 18225
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3879065706177692e-05,
+      "loss": 0.4921,
+      "step": 18226
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3878729870871327e-05,
+      "loss": 0.3782,
+      "step": 18227
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.387839403556496e-05,
+      "loss": 0.1996,
+      "step": 18228
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3878058200258594e-05,
+      "loss": 0.1391,
+      "step": 18229
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3877722364952229e-05,
+      "loss": 0.2879,
+      "step": 18230
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3877386529645862e-05,
+      "loss": 0.3128,
+      "step": 18231
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3877050694339498e-05,
+      "loss": 0.3845,
+      "step": 18232
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3876714859033131e-05,
+      "loss": 0.2234,
+      "step": 18233
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3876379023726764e-05,
+      "loss": 0.5782,
+      "step": 18234
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.38760431884204e-05,
+      "loss": 0.2802,
+      "step": 18235
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3875707353114033e-05,
+      "loss": 0.2417,
+      "step": 18236
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3875371517807668e-05,
+      "loss": 0.3025,
+      "step": 18237
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3875035682501303e-05,
+      "loss": 0.4165,
+      "step": 18238
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3874699847194937e-05,
+      "loss": 0.3491,
+      "step": 18239
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.387436401188857e-05,
+      "loss": 0.169,
+      "step": 18240
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3874028176582207e-05,
+      "loss": 0.4125,
+      "step": 18241
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.387369234127584e-05,
+      "loss": 0.188,
+      "step": 18242
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3873356505969473e-05,
+      "loss": 0.3751,
+      "step": 18243
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3873020670663109e-05,
+      "loss": 0.3658,
+      "step": 18244
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3872684835356742e-05,
+      "loss": 0.3303,
+      "step": 18245
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3872349000050376e-05,
+      "loss": 0.3466,
+      "step": 18246
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3872013164744011e-05,
+      "loss": 0.2217,
+      "step": 18247
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3871677329437646e-05,
+      "loss": 0.2352,
+      "step": 18248
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3871341494131279e-05,
+      "loss": 0.2871,
+      "step": 18249
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3871005658824915e-05,
+      "loss": 0.2696,
+      "step": 18250
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3870669823518548e-05,
+      "loss": 0.5065,
+      "step": 18251
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3870333988212181e-05,
+      "loss": 0.2514,
+      "step": 18252
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3869998152905817e-05,
+      "loss": 0.105,
+      "step": 18253
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.386966231759945e-05,
+      "loss": 0.3423,
+      "step": 18254
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3869326482293083e-05,
+      "loss": 0.2732,
+      "step": 18255
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.386899064698672e-05,
+      "loss": 0.2093,
+      "step": 18256
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3868654811680352e-05,
+      "loss": 0.5931,
+      "step": 18257
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3868318976373987e-05,
+      "loss": 0.1647,
+      "step": 18258
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3867983141067622e-05,
+      "loss": 0.3268,
+      "step": 18259
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3867647305761256e-05,
+      "loss": 0.2477,
+      "step": 18260
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.386731147045489e-05,
+      "loss": 0.4165,
+      "step": 18261
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3866975635148526e-05,
+      "loss": 0.211,
+      "step": 18262
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3866639799842159e-05,
+      "loss": 0.2591,
+      "step": 18263
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3866303964535792e-05,
+      "loss": 0.4229,
+      "step": 18264
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3865968129229428e-05,
+      "loss": 0.3739,
+      "step": 18265
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3865632293923061e-05,
+      "loss": 0.1666,
+      "step": 18266
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3865296458616696e-05,
+      "loss": 0.2908,
+      "step": 18267
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.386496062331033e-05,
+      "loss": 0.2473,
+      "step": 18268
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3864624788003965e-05,
+      "loss": 0.4228,
+      "step": 18269
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3864288952697598e-05,
+      "loss": 0.4308,
+      "step": 18270
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3863953117391234e-05,
+      "loss": 0.2006,
+      "step": 18271
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3863617282084867e-05,
+      "loss": 0.4404,
+      "step": 18272
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.38632814467785e-05,
+      "loss": 0.3768,
+      "step": 18273
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3862945611472136e-05,
+      "loss": 0.311,
+      "step": 18274
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.386260977616577e-05,
+      "loss": 0.2325,
+      "step": 18275
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3862273940859402e-05,
+      "loss": 0.216,
+      "step": 18276
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3861938105553039e-05,
+      "loss": 0.2975,
+      "step": 18277
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3861602270246672e-05,
+      "loss": 0.4707,
+      "step": 18278
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3861266434940306e-05,
+      "loss": 0.3156,
+      "step": 18279
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3860930599633941e-05,
+      "loss": 0.2933,
+      "step": 18280
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3860594764327576e-05,
+      "loss": 0.309,
+      "step": 18281
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3860258929021209e-05,
+      "loss": 0.2412,
+      "step": 18282
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3859923093714845e-05,
+      "loss": 0.3395,
+      "step": 18283
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3859587258408478e-05,
+      "loss": 0.3533,
+      "step": 18284
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.385925142310211e-05,
+      "loss": 0.2293,
+      "step": 18285
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3858915587795747e-05,
+      "loss": 0.2244,
+      "step": 18286
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.385857975248938e-05,
+      "loss": 0.2016,
+      "step": 18287
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3858243917183015e-05,
+      "loss": 0.3939,
+      "step": 18288
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.385790808187665e-05,
+      "loss": 0.2489,
+      "step": 18289
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3857572246570284e-05,
+      "loss": 0.2616,
+      "step": 18290
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3857236411263917e-05,
+      "loss": 0.3557,
+      "step": 18291
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3856900575957553e-05,
+      "loss": 0.2484,
+      "step": 18292
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3856564740651186e-05,
+      "loss": 0.2813,
+      "step": 18293
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.385622890534482e-05,
+      "loss": 0.4342,
+      "step": 18294
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3855893070038456e-05,
+      "loss": 0.3189,
+      "step": 18295
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3855557234732088e-05,
+      "loss": 0.515,
+      "step": 18296
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3855221399425723e-05,
+      "loss": 0.3004,
+      "step": 18297
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3854885564119358e-05,
+      "loss": 0.3035,
+      "step": 18298
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.385454972881299e-05,
+      "loss": 0.3944,
+      "step": 18299
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3854213893506625e-05,
+      "loss": 0.2191,
+      "step": 18300
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.385387805820026e-05,
+      "loss": 0.2964,
+      "step": 18301
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3853542222893895e-05,
+      "loss": 0.134,
+      "step": 18302
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3853206387587528e-05,
+      "loss": 0.1577,
+      "step": 18303
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3852870552281164e-05,
+      "loss": 0.3319,
+      "step": 18304
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3852534716974797e-05,
+      "loss": 0.4918,
+      "step": 18305
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.385219888166843e-05,
+      "loss": 0.2503,
+      "step": 18306
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3851863046362066e-05,
+      "loss": 0.5461,
+      "step": 18307
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.38515272110557e-05,
+      "loss": 0.2918,
+      "step": 18308
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3851191375749334e-05,
+      "loss": 0.3571,
+      "step": 18309
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3850855540442968e-05,
+      "loss": 0.5715,
+      "step": 18310
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3850519705136603e-05,
+      "loss": 0.2881,
+      "step": 18311
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3850183869830236e-05,
+      "loss": 0.4188,
+      "step": 18312
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3849848034523872e-05,
+      "loss": 0.414,
+      "step": 18313
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3849512199217505e-05,
+      "loss": 0.3453,
+      "step": 18314
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3849176363911138e-05,
+      "loss": 0.1574,
+      "step": 18315
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3848840528604775e-05,
+      "loss": 0.1802,
+      "step": 18316
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3848504693298408e-05,
+      "loss": 0.3004,
+      "step": 18317
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3848168857992042e-05,
+      "loss": 0.4886,
+      "step": 18318
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3847833022685675e-05,
+      "loss": 0.3151,
+      "step": 18319
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.384749718737931e-05,
+      "loss": 0.2772,
+      "step": 18320
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3847161352072945e-05,
+      "loss": 0.4115,
+      "step": 18321
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3846825516766577e-05,
+      "loss": 0.4202,
+      "step": 18322
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3846489681460214e-05,
+      "loss": 0.2647,
+      "step": 18323
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3846153846153847e-05,
+      "loss": 0.3788,
+      "step": 18324
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.384581801084748e-05,
+      "loss": 0.2262,
+      "step": 18325
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3845482175541116e-05,
+      "loss": 0.3492,
+      "step": 18326
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3845146340234749e-05,
+      "loss": 0.2837,
+      "step": 18327
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3844810504928384e-05,
+      "loss": 0.2865,
+      "step": 18328
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3844474669622018e-05,
+      "loss": 0.4045,
+      "step": 18329
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3844138834315653e-05,
+      "loss": 0.2444,
+      "step": 18330
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3843802999009286e-05,
+      "loss": 0.311,
+      "step": 18331
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3843467163702922e-05,
+      "loss": 0.2428,
+      "step": 18332
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3843131328396555e-05,
+      "loss": 0.3249,
+      "step": 18333
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3842795493090188e-05,
+      "loss": 0.3925,
+      "step": 18334
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3842459657783824e-05,
+      "loss": 0.4229,
+      "step": 18335
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3842123822477457e-05,
+      "loss": 0.4239,
+      "step": 18336
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3841787987171092e-05,
+      "loss": 0.27,
+      "step": 18337
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3841452151864727e-05,
+      "loss": 0.5139,
+      "step": 18338
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3841116316558361e-05,
+      "loss": 0.1898,
+      "step": 18339
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3840780481251994e-05,
+      "loss": 0.3194,
+      "step": 18340
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.384044464594563e-05,
+      "loss": 0.3763,
+      "step": 18341
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3840108810639264e-05,
+      "loss": 0.2835,
+      "step": 18342
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3839772975332897e-05,
+      "loss": 0.3644,
+      "step": 18343
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3839437140026533e-05,
+      "loss": 0.4072,
+      "step": 18344
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3839101304720166e-05,
+      "loss": 0.2753,
+      "step": 18345
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3838765469413799e-05,
+      "loss": 0.3534,
+      "step": 18346
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3838429634107435e-05,
+      "loss": 0.159,
+      "step": 18347
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3838093798801068e-05,
+      "loss": 0.2815,
+      "step": 18348
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3837757963494703e-05,
+      "loss": 0.3411,
+      "step": 18349
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3837422128188337e-05,
+      "loss": 0.2944,
+      "step": 18350
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3837086292881972e-05,
+      "loss": 0.2846,
+      "step": 18351
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3836750457575605e-05,
+      "loss": 0.3189,
+      "step": 18352
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3836414622269241e-05,
+      "loss": 0.3491,
+      "step": 18353
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3836078786962874e-05,
+      "loss": 0.289,
+      "step": 18354
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3835742951656507e-05,
+      "loss": 0.1631,
+      "step": 18355
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3835407116350144e-05,
+      "loss": 0.2955,
+      "step": 18356
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3835071281043777e-05,
+      "loss": 0.297,
+      "step": 18357
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3834735445737411e-05,
+      "loss": 0.2902,
+      "step": 18358
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3834399610431046e-05,
+      "loss": 0.3992,
+      "step": 18359
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.383406377512468e-05,
+      "loss": 0.3282,
+      "step": 18360
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.3833727939818313e-05,
+      "loss": 0.4974,
+      "step": 18361
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.383339210451195e-05,
+      "loss": 0.1619,
+      "step": 18362
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3833056269205583e-05,
+      "loss": 0.2561,
+      "step": 18363
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3832720433899216e-05,
+      "loss": 0.3124,
+      "step": 18364
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3832384598592852e-05,
+      "loss": 0.3029,
+      "step": 18365
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3832048763286485e-05,
+      "loss": 0.3217,
+      "step": 18366
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3831712927980118e-05,
+      "loss": 0.3382,
+      "step": 18367
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3831377092673754e-05,
+      "loss": 0.341,
+      "step": 18368
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3831041257367387e-05,
+      "loss": 0.3378,
+      "step": 18369
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3830705422061022e-05,
+      "loss": 0.2818,
+      "step": 18370
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3830369586754657e-05,
+      "loss": 0.3112,
+      "step": 18371
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3830033751448291e-05,
+      "loss": 0.3634,
+      "step": 18372
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3829697916141924e-05,
+      "loss": 0.4075,
+      "step": 18373
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.382936208083556e-05,
+      "loss": 0.274,
+      "step": 18374
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3829026245529193e-05,
+      "loss": 0.3504,
+      "step": 18375
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3828690410222826e-05,
+      "loss": 0.3472,
+      "step": 18376
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3828354574916463e-05,
+      "loss": 0.2996,
+      "step": 18377
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3828018739610096e-05,
+      "loss": 0.1682,
+      "step": 18378
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.382768290430373e-05,
+      "loss": 0.374,
+      "step": 18379
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3827347068997365e-05,
+      "loss": 0.3013,
+      "step": 18380
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3827011233691e-05,
+      "loss": 0.2095,
+      "step": 18381
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3826675398384633e-05,
+      "loss": 0.3143,
+      "step": 18382
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3826339563078269e-05,
+      "loss": 0.3259,
+      "step": 18383
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3826003727771902e-05,
+      "loss": 0.3286,
+      "step": 18384
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3825667892465535e-05,
+      "loss": 0.3828,
+      "step": 18385
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3825332057159171e-05,
+      "loss": 0.4404,
+      "step": 18386
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3824996221852804e-05,
+      "loss": 0.1765,
+      "step": 18387
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3824660386546437e-05,
+      "loss": 0.3523,
+      "step": 18388
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3824324551240073e-05,
+      "loss": 0.2791,
+      "step": 18389
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3823988715933706e-05,
+      "loss": 0.2698,
+      "step": 18390
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3823652880627341e-05,
+      "loss": 0.3312,
+      "step": 18391
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3823317045320976e-05,
+      "loss": 0.1804,
+      "step": 18392
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.382298121001461e-05,
+      "loss": 0.269,
+      "step": 18393
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3822645374708243e-05,
+      "loss": 0.4358,
+      "step": 18394
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.382230953940188e-05,
+      "loss": 0.4303,
+      "step": 18395
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3821973704095513e-05,
+      "loss": 0.2321,
+      "step": 18396
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3821637868789146e-05,
+      "loss": 0.2107,
+      "step": 18397
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3821302033482782e-05,
+      "loss": 0.4074,
+      "step": 18398
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3820966198176415e-05,
+      "loss": 0.2546,
+      "step": 18399
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.382063036287005e-05,
+      "loss": 0.1817,
+      "step": 18400
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3820294527563684e-05,
+      "loss": 0.367,
+      "step": 18401
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3819958692257319e-05,
+      "loss": 0.4452,
+      "step": 18402
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3819622856950952e-05,
+      "loss": 0.5185,
+      "step": 18403
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3819287021644588e-05,
+      "loss": 0.4203,
+      "step": 18404
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3818951186338221e-05,
+      "loss": 0.3485,
+      "step": 18405
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3818615351031854e-05,
+      "loss": 0.4031,
+      "step": 18406
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.381827951572549e-05,
+      "loss": 0.2072,
+      "step": 18407
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3817943680419123e-05,
+      "loss": 0.4291,
+      "step": 18408
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3817607845112758e-05,
+      "loss": 0.1765,
+      "step": 18409
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3817272009806393e-05,
+      "loss": 0.4923,
+      "step": 18410
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3816936174500026e-05,
+      "loss": 0.3175,
+      "step": 18411
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.381660033919366e-05,
+      "loss": 0.2779,
+      "step": 18412
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3816264503887295e-05,
+      "loss": 0.4227,
+      "step": 18413
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.381592866858093e-05,
+      "loss": 0.2283,
+      "step": 18414
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3815592833274562e-05,
+      "loss": 0.4536,
+      "step": 18415
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3815256997968199e-05,
+      "loss": 0.3944,
+      "step": 18416
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3814921162661832e-05,
+      "loss": 0.4456,
+      "step": 18417
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3814585327355465e-05,
+      "loss": 0.2575,
+      "step": 18418
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3814249492049101e-05,
+      "loss": 0.375,
+      "step": 18419
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3813913656742734e-05,
+      "loss": 0.3026,
+      "step": 18420
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3813577821436369e-05,
+      "loss": 0.3144,
+      "step": 18421
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3813241986130003e-05,
+      "loss": 0.3149,
+      "step": 18422
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3812906150823638e-05,
+      "loss": 0.2356,
+      "step": 18423
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.381257031551727e-05,
+      "loss": 0.4357,
+      "step": 18424
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3812234480210907e-05,
+      "loss": 0.3334,
+      "step": 18425
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.381189864490454e-05,
+      "loss": 0.2564,
+      "step": 18426
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3811562809598173e-05,
+      "loss": 0.2795,
+      "step": 18427
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.381122697429181e-05,
+      "loss": 0.2885,
+      "step": 18428
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3810891138985442e-05,
+      "loss": 0.4352,
+      "step": 18429
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3810555303679077e-05,
+      "loss": 0.3044,
+      "step": 18430
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3810219468372712e-05,
+      "loss": 0.2512,
+      "step": 18431
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3809883633066346e-05,
+      "loss": 0.4846,
+      "step": 18432
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.380954779775998e-05,
+      "loss": 0.254,
+      "step": 18433
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3809211962453614e-05,
+      "loss": 0.3651,
+      "step": 18434
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3808876127147249e-05,
+      "loss": 0.3607,
+      "step": 18435
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3808540291840882e-05,
+      "loss": 0.2406,
+      "step": 18436
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3808204456534518e-05,
+      "loss": 0.3733,
+      "step": 18437
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.380786862122815e-05,
+      "loss": 0.3898,
+      "step": 18438
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3807532785921784e-05,
+      "loss": 0.181,
+      "step": 18439
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.380719695061542e-05,
+      "loss": 0.4126,
+      "step": 18440
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3806861115309053e-05,
+      "loss": 0.3918,
+      "step": 18441
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3806525280002688e-05,
+      "loss": 0.2065,
+      "step": 18442
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3806189444696322e-05,
+      "loss": 0.6269,
+      "step": 18443
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3805853609389957e-05,
+      "loss": 0.4171,
+      "step": 18444
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.380551777408359e-05,
+      "loss": 0.2756,
+      "step": 18445
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3805181938777226e-05,
+      "loss": 0.4043,
+      "step": 18446
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.380484610347086e-05,
+      "loss": 0.6942,
+      "step": 18447
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3804510268164492e-05,
+      "loss": 0.3982,
+      "step": 18448
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3804174432858129e-05,
+      "loss": 0.193,
+      "step": 18449
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3803838597551762e-05,
+      "loss": 0.274,
+      "step": 18450
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3803502762245396e-05,
+      "loss": 0.2903,
+      "step": 18451
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.380316692693903e-05,
+      "loss": 0.3334,
+      "step": 18452
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3802831091632665e-05,
+      "loss": 0.2778,
+      "step": 18453
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3802495256326298e-05,
+      "loss": 0.3283,
+      "step": 18454
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3802159421019933e-05,
+      "loss": 0.3413,
+      "step": 18455
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3801823585713568e-05,
+      "loss": 0.3262,
+      "step": 18456
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.38014877504072e-05,
+      "loss": 0.328,
+      "step": 18457
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3801151915100837e-05,
+      "loss": 0.346,
+      "step": 18458
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.380081607979447e-05,
+      "loss": 0.5515,
+      "step": 18459
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3800480244488103e-05,
+      "loss": 0.1807,
+      "step": 18460
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.380014440918174e-05,
+      "loss": 0.3988,
+      "step": 18461
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3799808573875372e-05,
+      "loss": 0.286,
+      "step": 18462
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3799472738569007e-05,
+      "loss": 0.3377,
+      "step": 18463
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3799136903262641e-05,
+      "loss": 0.3025,
+      "step": 18464
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3798801067956276e-05,
+      "loss": 0.1971,
+      "step": 18465
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3798465232649909e-05,
+      "loss": 0.3157,
+      "step": 18466
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3798129397343545e-05,
+      "loss": 0.5646,
+      "step": 18467
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3797793562037178e-05,
+      "loss": 0.2499,
+      "step": 18468
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3797457726730811e-05,
+      "loss": 0.4251,
+      "step": 18469
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3797121891424448e-05,
+      "loss": 0.209,
+      "step": 18470
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.379678605611808e-05,
+      "loss": 0.2607,
+      "step": 18471
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3796450220811715e-05,
+      "loss": 0.5197,
+      "step": 18472
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.379611438550535e-05,
+      "loss": 0.3393,
+      "step": 18473
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3795778550198985e-05,
+      "loss": 0.3137,
+      "step": 18474
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3795442714892618e-05,
+      "loss": 0.3072,
+      "step": 18475
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3795106879586254e-05,
+      "loss": 0.3506,
+      "step": 18476
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3794771044279887e-05,
+      "loss": 0.2778,
+      "step": 18477
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.379443520897352e-05,
+      "loss": 0.2233,
+      "step": 18478
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3794099373667156e-05,
+      "loss": 0.363,
+      "step": 18479
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3793763538360789e-05,
+      "loss": 0.428,
+      "step": 18480
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3793427703054422e-05,
+      "loss": 0.4065,
+      "step": 18481
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3793091867748058e-05,
+      "loss": 0.2609,
+      "step": 18482
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3792756032441691e-05,
+      "loss": 0.2914,
+      "step": 18483
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3792420197135326e-05,
+      "loss": 0.3003,
+      "step": 18484
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.379208436182896e-05,
+      "loss": 0.4142,
+      "step": 18485
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3791748526522595e-05,
+      "loss": 0.2231,
+      "step": 18486
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3791412691216228e-05,
+      "loss": 0.3581,
+      "step": 18487
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3791076855909865e-05,
+      "loss": 0.5719,
+      "step": 18488
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3790741020603498e-05,
+      "loss": 0.2379,
+      "step": 18489
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.379040518529713e-05,
+      "loss": 0.3046,
+      "step": 18490
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3790069349990767e-05,
+      "loss": 0.3369,
+      "step": 18491
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.37897335146844e-05,
+      "loss": 0.2596,
+      "step": 18492
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3789397679378034e-05,
+      "loss": 0.2501,
+      "step": 18493
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3789061844071669e-05,
+      "loss": 0.3767,
+      "step": 18494
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3788726008765304e-05,
+      "loss": 0.502,
+      "step": 18495
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3788390173458937e-05,
+      "loss": 0.5093,
+      "step": 18496
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3788054338152573e-05,
+      "loss": 0.192,
+      "step": 18497
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3787718502846206e-05,
+      "loss": 0.2627,
+      "step": 18498
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3787382667539839e-05,
+      "loss": 0.2273,
+      "step": 18499
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3787046832233474e-05,
+      "loss": 0.4678,
+      "step": 18500
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3786710996927108e-05,
+      "loss": 0.3144,
+      "step": 18501
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3786375161620741e-05,
+      "loss": 0.3076,
+      "step": 18502
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3786039326314376e-05,
+      "loss": 0.3033,
+      "step": 18503
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.378570349100801e-05,
+      "loss": 0.319,
+      "step": 18504
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3785367655701645e-05,
+      "loss": 0.2374,
+      "step": 18505
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3785031820395278e-05,
+      "loss": 0.1788,
+      "step": 18506
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3784695985088914e-05,
+      "loss": 0.2483,
+      "step": 18507
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3784360149782547e-05,
+      "loss": 0.3564,
+      "step": 18508
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.378402431447618e-05,
+      "loss": 0.2347,
+      "step": 18509
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3783688479169817e-05,
+      "loss": 0.434,
+      "step": 18510
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.378335264386345e-05,
+      "loss": 0.3415,
+      "step": 18511
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3783016808557084e-05,
+      "loss": 0.2579,
+      "step": 18512
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3782680973250719e-05,
+      "loss": 0.3234,
+      "step": 18513
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3782345137944354e-05,
+      "loss": 0.3413,
+      "step": 18514
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3782009302637986e-05,
+      "loss": 0.3,
+      "step": 18515
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3781673467331623e-05,
+      "loss": 0.3601,
+      "step": 18516
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3781337632025256e-05,
+      "loss": 0.2201,
+      "step": 18517
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3781001796718889e-05,
+      "loss": 0.226,
+      "step": 18518
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3780665961412525e-05,
+      "loss": 0.355,
+      "step": 18519
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3780330126106158e-05,
+      "loss": 0.3016,
+      "step": 18520
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3779994290799793e-05,
+      "loss": 0.3564,
+      "step": 18521
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3779658455493427e-05,
+      "loss": 0.2911,
+      "step": 18522
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3779322620187062e-05,
+      "loss": 0.3187,
+      "step": 18523
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3778986784880695e-05,
+      "loss": 0.271,
+      "step": 18524
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.377865094957433e-05,
+      "loss": 0.1485,
+      "step": 18525
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3778315114267964e-05,
+      "loss": 0.2831,
+      "step": 18526
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3777979278961597e-05,
+      "loss": 0.2852,
+      "step": 18527
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3777643443655234e-05,
+      "loss": 0.3963,
+      "step": 18528
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3777307608348866e-05,
+      "loss": 0.4358,
+      "step": 18529
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.37769717730425e-05,
+      "loss": 0.2728,
+      "step": 18530
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3776635937736136e-05,
+      "loss": 0.2254,
+      "step": 18531
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3776300102429769e-05,
+      "loss": 0.3692,
+      "step": 18532
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3775964267123403e-05,
+      "loss": 0.2612,
+      "step": 18533
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3775628431817038e-05,
+      "loss": 0.464,
+      "step": 18534
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3775292596510673e-05,
+      "loss": 0.3867,
+      "step": 18535
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3774956761204306e-05,
+      "loss": 0.281,
+      "step": 18536
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3774620925897942e-05,
+      "loss": 0.242,
+      "step": 18537
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3774285090591575e-05,
+      "loss": 0.3694,
+      "step": 18538
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3773949255285208e-05,
+      "loss": 0.2199,
+      "step": 18539
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3773613419978844e-05,
+      "loss": 0.3417,
+      "step": 18540
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3773277584672477e-05,
+      "loss": 0.2998,
+      "step": 18541
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3772941749366112e-05,
+      "loss": 0.3544,
+      "step": 18542
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3772605914059746e-05,
+      "loss": 0.2358,
+      "step": 18543
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3772270078753381e-05,
+      "loss": 0.2594,
+      "step": 18544
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3771934243447014e-05,
+      "loss": 0.4483,
+      "step": 18545
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3771598408140649e-05,
+      "loss": 0.4355,
+      "step": 18546
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3771262572834283e-05,
+      "loss": 0.2918,
+      "step": 18547
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3770926737527916e-05,
+      "loss": 0.3406,
+      "step": 18548
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3770590902221553e-05,
+      "loss": 0.3632,
+      "step": 18549
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3770255066915186e-05,
+      "loss": 0.3569,
+      "step": 18550
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3769919231608819e-05,
+      "loss": 0.24,
+      "step": 18551
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3769583396302455e-05,
+      "loss": 0.3055,
+      "step": 18552
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3769247560996088e-05,
+      "loss": 0.3201,
+      "step": 18553
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3768911725689722e-05,
+      "loss": 0.3157,
+      "step": 18554
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3768575890383357e-05,
+      "loss": 0.3261,
+      "step": 18555
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3768240055076992e-05,
+      "loss": 0.3823,
+      "step": 18556
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3767904219770625e-05,
+      "loss": 0.2615,
+      "step": 18557
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3767568384464261e-05,
+      "loss": 0.2706,
+      "step": 18558
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3767232549157894e-05,
+      "loss": 0.3093,
+      "step": 18559
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.3766896713851527e-05,
+      "loss": 0.224,
+      "step": 18560
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3766560878545163e-05,
+      "loss": 0.366,
+      "step": 18561
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3766225043238796e-05,
+      "loss": 0.4218,
+      "step": 18562
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3765889207932431e-05,
+      "loss": 0.367,
+      "step": 18563
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3765553372626066e-05,
+      "loss": 0.2556,
+      "step": 18564
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.37652175373197e-05,
+      "loss": 0.3117,
+      "step": 18565
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3764881702013333e-05,
+      "loss": 0.321,
+      "step": 18566
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.376454586670697e-05,
+      "loss": 0.3705,
+      "step": 18567
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3764210031400602e-05,
+      "loss": 0.4296,
+      "step": 18568
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3763874196094235e-05,
+      "loss": 0.4134,
+      "step": 18569
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3763538360787872e-05,
+      "loss": 0.3258,
+      "step": 18570
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3763202525481505e-05,
+      "loss": 0.4196,
+      "step": 18571
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3762866690175138e-05,
+      "loss": 0.3045,
+      "step": 18572
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3762530854868774e-05,
+      "loss": 0.4752,
+      "step": 18573
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3762195019562407e-05,
+      "loss": 0.2972,
+      "step": 18574
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3761859184256042e-05,
+      "loss": 0.4027,
+      "step": 18575
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3761523348949676e-05,
+      "loss": 0.2823,
+      "step": 18576
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3761187513643311e-05,
+      "loss": 0.2582,
+      "step": 18577
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3760851678336944e-05,
+      "loss": 0.2414,
+      "step": 18578
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.376051584303058e-05,
+      "loss": 0.3643,
+      "step": 18579
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3760180007724213e-05,
+      "loss": 0.2457,
+      "step": 18580
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3759844172417846e-05,
+      "loss": 0.3166,
+      "step": 18581
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3759508337111482e-05,
+      "loss": 0.183,
+      "step": 18582
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3759172501805115e-05,
+      "loss": 0.245,
+      "step": 18583
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.375883666649875e-05,
+      "loss": 0.2917,
+      "step": 18584
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3758500831192385e-05,
+      "loss": 0.2462,
+      "step": 18585
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.375816499588602e-05,
+      "loss": 0.3562,
+      "step": 18586
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3757829160579652e-05,
+      "loss": 0.2471,
+      "step": 18587
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3757493325273289e-05,
+      "loss": 0.1854,
+      "step": 18588
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3757157489966922e-05,
+      "loss": 0.4092,
+      "step": 18589
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3756821654660555e-05,
+      "loss": 0.3137,
+      "step": 18590
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3756485819354191e-05,
+      "loss": 0.3264,
+      "step": 18591
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3756149984047824e-05,
+      "loss": 0.1855,
+      "step": 18592
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3755814148741457e-05,
+      "loss": 0.3728,
+      "step": 18593
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3755478313435093e-05,
+      "loss": 0.2096,
+      "step": 18594
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3755142478128726e-05,
+      "loss": 0.2132,
+      "step": 18595
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.375480664282236e-05,
+      "loss": 0.2831,
+      "step": 18596
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3754470807515995e-05,
+      "loss": 0.3598,
+      "step": 18597
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.375413497220963e-05,
+      "loss": 0.2376,
+      "step": 18598
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3753799136903263e-05,
+      "loss": 0.2998,
+      "step": 18599
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.37534633015969e-05,
+      "loss": 0.182,
+      "step": 18600
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3753127466290532e-05,
+      "loss": 0.3367,
+      "step": 18601
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3752791630984165e-05,
+      "loss": 0.2759,
+      "step": 18602
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3752455795677802e-05,
+      "loss": 0.3005,
+      "step": 18603
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3752119960371435e-05,
+      "loss": 0.4305,
+      "step": 18604
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.375178412506507e-05,
+      "loss": 0.3515,
+      "step": 18605
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3751448289758704e-05,
+      "loss": 0.2719,
+      "step": 18606
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3751112454452338e-05,
+      "loss": 0.2569,
+      "step": 18607
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3750776619145971e-05,
+      "loss": 0.365,
+      "step": 18608
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3750440783839608e-05,
+      "loss": 0.3292,
+      "step": 18609
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.375010494853324e-05,
+      "loss": 0.3175,
+      "step": 18610
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3749769113226874e-05,
+      "loss": 0.5816,
+      "step": 18611
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.374943327792051e-05,
+      "loss": 0.3941,
+      "step": 18612
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3749097442614143e-05,
+      "loss": 0.3098,
+      "step": 18613
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3748761607307776e-05,
+      "loss": 0.2883,
+      "step": 18614
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3748425772001412e-05,
+      "loss": 0.2719,
+      "step": 18615
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3748089936695045e-05,
+      "loss": 0.2868,
+      "step": 18616
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.374775410138868e-05,
+      "loss": 0.2972,
+      "step": 18617
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3747418266082315e-05,
+      "loss": 0.2963,
+      "step": 18618
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.374708243077595e-05,
+      "loss": 0.2705,
+      "step": 18619
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3746746595469582e-05,
+      "loss": 0.2269,
+      "step": 18620
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3746410760163218e-05,
+      "loss": 0.2445,
+      "step": 18621
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3746074924856851e-05,
+      "loss": 0.2819,
+      "step": 18622
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3745739089550484e-05,
+      "loss": 0.413,
+      "step": 18623
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.374540325424412e-05,
+      "loss": 0.2736,
+      "step": 18624
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3745067418937754e-05,
+      "loss": 0.2992,
+      "step": 18625
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3744731583631388e-05,
+      "loss": 0.2683,
+      "step": 18626
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3744395748325023e-05,
+      "loss": 0.4468,
+      "step": 18627
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3744059913018658e-05,
+      "loss": 0.3705,
+      "step": 18628
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.374372407771229e-05,
+      "loss": 0.3222,
+      "step": 18629
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3743388242405927e-05,
+      "loss": 0.4131,
+      "step": 18630
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.374305240709956e-05,
+      "loss": 0.4585,
+      "step": 18631
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3742716571793193e-05,
+      "loss": 0.3523,
+      "step": 18632
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3742380736486829e-05,
+      "loss": 0.2423,
+      "step": 18633
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3742044901180462e-05,
+      "loss": 0.3379,
+      "step": 18634
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3741709065874097e-05,
+      "loss": 0.3501,
+      "step": 18635
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3741373230567731e-05,
+      "loss": 0.4635,
+      "step": 18636
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3741037395261364e-05,
+      "loss": 0.3045,
+      "step": 18637
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3740701559954999e-05,
+      "loss": 0.2534,
+      "step": 18638
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3740365724648634e-05,
+      "loss": 0.3388,
+      "step": 18639
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3740029889342268e-05,
+      "loss": 0.2848,
+      "step": 18640
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3739694054035901e-05,
+      "loss": 0.3674,
+      "step": 18641
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3739358218729538e-05,
+      "loss": 0.326,
+      "step": 18642
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.373902238342317e-05,
+      "loss": 0.2748,
+      "step": 18643
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3738686548116803e-05,
+      "loss": 0.3725,
+      "step": 18644
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.373835071281044e-05,
+      "loss": 0.1283,
+      "step": 18645
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3738014877504073e-05,
+      "loss": 0.1591,
+      "step": 18646
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3737679042197707e-05,
+      "loss": 0.4364,
+      "step": 18647
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3737343206891342e-05,
+      "loss": 0.3607,
+      "step": 18648
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3737007371584977e-05,
+      "loss": 0.3396,
+      "step": 18649
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.373667153627861e-05,
+      "loss": 0.219,
+      "step": 18650
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3736335700972246e-05,
+      "loss": 0.3673,
+      "step": 18651
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3735999865665879e-05,
+      "loss": 0.268,
+      "step": 18652
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3735664030359512e-05,
+      "loss": 0.2588,
+      "step": 18653
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3735328195053148e-05,
+      "loss": 0.2079,
+      "step": 18654
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3734992359746781e-05,
+      "loss": 0.3237,
+      "step": 18655
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3734656524440416e-05,
+      "loss": 0.2772,
+      "step": 18656
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.373432068913405e-05,
+      "loss": 0.3171,
+      "step": 18657
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3733984853827685e-05,
+      "loss": 0.3453,
+      "step": 18658
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3733649018521318e-05,
+      "loss": 0.2974,
+      "step": 18659
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3733313183214953e-05,
+      "loss": 0.4349,
+      "step": 18660
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3732977347908587e-05,
+      "loss": 0.2357,
+      "step": 18661
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.373264151260222e-05,
+      "loss": 0.4496,
+      "step": 18662
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3732305677295857e-05,
+      "loss": 0.3085,
+      "step": 18663
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.373196984198949e-05,
+      "loss": 0.3181,
+      "step": 18664
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3731634006683123e-05,
+      "loss": 0.4012,
+      "step": 18665
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3731298171376759e-05,
+      "loss": 0.3638,
+      "step": 18666
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3730962336070392e-05,
+      "loss": 0.2595,
+      "step": 18667
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3730626500764027e-05,
+      "loss": 0.3471,
+      "step": 18668
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3730290665457661e-05,
+      "loss": 0.5072,
+      "step": 18669
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3729954830151296e-05,
+      "loss": 0.229,
+      "step": 18670
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3729618994844929e-05,
+      "loss": 0.3459,
+      "step": 18671
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3729283159538565e-05,
+      "loss": 0.4006,
+      "step": 18672
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3728947324232198e-05,
+      "loss": 0.2495,
+      "step": 18673
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3728611488925831e-05,
+      "loss": 0.2667,
+      "step": 18674
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3728275653619467e-05,
+      "loss": 0.2242,
+      "step": 18675
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.37279398183131e-05,
+      "loss": 0.3896,
+      "step": 18676
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3727603983006735e-05,
+      "loss": 0.3583,
+      "step": 18677
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3727268147700368e-05,
+      "loss": 0.288,
+      "step": 18678
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3726932312394004e-05,
+      "loss": 0.1332,
+      "step": 18679
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3726596477087637e-05,
+      "loss": 0.5968,
+      "step": 18680
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.372626064178127e-05,
+      "loss": 0.443,
+      "step": 18681
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3725924806474907e-05,
+      "loss": 0.2417,
+      "step": 18682
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.372558897116854e-05,
+      "loss": 0.3781,
+      "step": 18683
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3725253135862172e-05,
+      "loss": 0.3667,
+      "step": 18684
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3724917300555809e-05,
+      "loss": 0.3659,
+      "step": 18685
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3724581465249442e-05,
+      "loss": 0.2366,
+      "step": 18686
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3724245629943076e-05,
+      "loss": 0.366,
+      "step": 18687
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3723909794636711e-05,
+      "loss": 0.3339,
+      "step": 18688
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3723573959330346e-05,
+      "loss": 0.3089,
+      "step": 18689
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3723238124023979e-05,
+      "loss": 0.2799,
+      "step": 18690
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3722902288717615e-05,
+      "loss": 0.2612,
+      "step": 18691
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3722566453411248e-05,
+      "loss": 0.4002,
+      "step": 18692
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3722230618104881e-05,
+      "loss": 0.3904,
+      "step": 18693
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3721894782798517e-05,
+      "loss": 0.2147,
+      "step": 18694
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.372155894749215e-05,
+      "loss": 0.225,
+      "step": 18695
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3721223112185785e-05,
+      "loss": 0.1413,
+      "step": 18696
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.372088727687942e-05,
+      "loss": 0.2189,
+      "step": 18697
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3720551441573054e-05,
+      "loss": 0.2088,
+      "step": 18698
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3720215606266687e-05,
+      "loss": 0.4483,
+      "step": 18699
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3719879770960323e-05,
+      "loss": 0.2633,
+      "step": 18700
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3719543935653956e-05,
+      "loss": 0.2179,
+      "step": 18701
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.371920810034759e-05,
+      "loss": 0.2914,
+      "step": 18702
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3718872265041226e-05,
+      "loss": 0.2542,
+      "step": 18703
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3718536429734859e-05,
+      "loss": 0.2307,
+      "step": 18704
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3718200594428492e-05,
+      "loss": 0.2838,
+      "step": 18705
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3717864759122128e-05,
+      "loss": 0.3816,
+      "step": 18706
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3717528923815761e-05,
+      "loss": 0.2534,
+      "step": 18707
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3717193088509396e-05,
+      "loss": 0.2658,
+      "step": 18708
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.371685725320303e-05,
+      "loss": 0.2659,
+      "step": 18709
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3716521417896665e-05,
+      "loss": 0.1918,
+      "step": 18710
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3716185582590298e-05,
+      "loss": 0.2839,
+      "step": 18711
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3715849747283934e-05,
+      "loss": 0.3745,
+      "step": 18712
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3715513911977567e-05,
+      "loss": 0.3041,
+      "step": 18713
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.37151780766712e-05,
+      "loss": 0.1841,
+      "step": 18714
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3714842241364836e-05,
+      "loss": 0.3718,
+      "step": 18715
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.371450640605847e-05,
+      "loss": 0.4848,
+      "step": 18716
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3714170570752104e-05,
+      "loss": 0.4978,
+      "step": 18717
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3713834735445739e-05,
+      "loss": 0.2753,
+      "step": 18718
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3713498900139373e-05,
+      "loss": 0.214,
+      "step": 18719
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3713163064833006e-05,
+      "loss": 0.406,
+      "step": 18720
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3712827229526643e-05,
+      "loss": 0.2653,
+      "step": 18721
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3712491394220275e-05,
+      "loss": 0.2232,
+      "step": 18722
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3712155558913908e-05,
+      "loss": 0.2595,
+      "step": 18723
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3711819723607545e-05,
+      "loss": 0.2439,
+      "step": 18724
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3711483888301178e-05,
+      "loss": 0.5003,
+      "step": 18725
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3711148052994812e-05,
+      "loss": 0.3313,
+      "step": 18726
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3710812217688447e-05,
+      "loss": 0.274,
+      "step": 18727
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.371047638238208e-05,
+      "loss": 0.3101,
+      "step": 18728
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3710140547075715e-05,
+      "loss": 0.3022,
+      "step": 18729
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.370980471176935e-05,
+      "loss": 0.2341,
+      "step": 18730
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3709468876462984e-05,
+      "loss": 0.3619,
+      "step": 18731
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3709133041156617e-05,
+      "loss": 0.1581,
+      "step": 18732
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3708797205850253e-05,
+      "loss": 0.3267,
+      "step": 18733
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3708461370543886e-05,
+      "loss": 0.3927,
+      "step": 18734
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3708125535237519e-05,
+      "loss": 0.3025,
+      "step": 18735
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3707789699931155e-05,
+      "loss": 0.6589,
+      "step": 18736
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3707453864624788e-05,
+      "loss": 0.3587,
+      "step": 18737
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3707118029318423e-05,
+      "loss": 0.4037,
+      "step": 18738
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3706782194012058e-05,
+      "loss": 0.3035,
+      "step": 18739
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3706446358705692e-05,
+      "loss": 0.3818,
+      "step": 18740
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3706110523399325e-05,
+      "loss": 0.1913,
+      "step": 18741
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3705774688092962e-05,
+      "loss": 0.3514,
+      "step": 18742
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3705438852786595e-05,
+      "loss": 0.1633,
+      "step": 18743
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3705103017480228e-05,
+      "loss": 0.334,
+      "step": 18744
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3704767182173864e-05,
+      "loss": 0.2661,
+      "step": 18745
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3704431346867497e-05,
+      "loss": 0.4737,
+      "step": 18746
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3704095511561132e-05,
+      "loss": 0.349,
+      "step": 18747
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3703759676254766e-05,
+      "loss": 0.4058,
+      "step": 18748
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3703423840948399e-05,
+      "loss": 0.3419,
+      "step": 18749
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3703088005642034e-05,
+      "loss": 0.2225,
+      "step": 18750
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3702752170335668e-05,
+      "loss": 0.1918,
+      "step": 18751
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3702416335029303e-05,
+      "loss": 0.2771,
+      "step": 18752
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3702080499722936e-05,
+      "loss": 0.4474,
+      "step": 18753
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3701744664416572e-05,
+      "loss": 0.5153,
+      "step": 18754
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3701408829110205e-05,
+      "loss": 0.3935,
+      "step": 18755
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3701072993803838e-05,
+      "loss": 0.3391,
+      "step": 18756
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3700737158497475e-05,
+      "loss": 0.3037,
+      "step": 18757
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3700401323191108e-05,
+      "loss": 0.2749,
+      "step": 18758
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.3700065487884742e-05,
+      "loss": 0.2696,
+      "step": 18759
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3699729652578377e-05,
+      "loss": 0.4225,
+      "step": 18760
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3699393817272011e-05,
+      "loss": 0.3694,
+      "step": 18761
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3699057981965644e-05,
+      "loss": 0.3537,
+      "step": 18762
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.369872214665928e-05,
+      "loss": 0.3858,
+      "step": 18763
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3698386311352914e-05,
+      "loss": 0.3721,
+      "step": 18764
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3698050476046547e-05,
+      "loss": 0.3777,
+      "step": 18765
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3697714640740183e-05,
+      "loss": 0.2125,
+      "step": 18766
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3697378805433816e-05,
+      "loss": 0.319,
+      "step": 18767
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.369704297012745e-05,
+      "loss": 0.3567,
+      "step": 18768
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3696707134821085e-05,
+      "loss": 0.2047,
+      "step": 18769
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.369637129951472e-05,
+      "loss": 0.3011,
+      "step": 18770
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3696035464208353e-05,
+      "loss": 0.2962,
+      "step": 18771
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3695699628901988e-05,
+      "loss": 0.2944,
+      "step": 18772
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3695363793595622e-05,
+      "loss": 0.3494,
+      "step": 18773
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3695027958289255e-05,
+      "loss": 0.2546,
+      "step": 18774
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3694692122982891e-05,
+      "loss": 0.4062,
+      "step": 18775
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3694356287676524e-05,
+      "loss": 0.3935,
+      "step": 18776
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3694020452370157e-05,
+      "loss": 0.5137,
+      "step": 18777
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3693684617063794e-05,
+      "loss": 0.4602,
+      "step": 18778
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3693348781757427e-05,
+      "loss": 0.4449,
+      "step": 18779
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3693012946451061e-05,
+      "loss": 0.2151,
+      "step": 18780
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3692677111144696e-05,
+      "loss": 0.4168,
+      "step": 18781
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.369234127583833e-05,
+      "loss": 0.4215,
+      "step": 18782
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3692005440531964e-05,
+      "loss": 0.284,
+      "step": 18783
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.36916696052256e-05,
+      "loss": 0.2302,
+      "step": 18784
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3691333769919233e-05,
+      "loss": 0.188,
+      "step": 18785
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3690997934612866e-05,
+      "loss": 0.3658,
+      "step": 18786
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3690662099306502e-05,
+      "loss": 0.3692,
+      "step": 18787
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3690326264000135e-05,
+      "loss": 0.36,
+      "step": 18788
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.368999042869377e-05,
+      "loss": 0.4625,
+      "step": 18789
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3689654593387404e-05,
+      "loss": 0.4665,
+      "step": 18790
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3689318758081039e-05,
+      "loss": 0.235,
+      "step": 18791
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3688982922774672e-05,
+      "loss": 0.2054,
+      "step": 18792
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3688647087468308e-05,
+      "loss": 0.246,
+      "step": 18793
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3688311252161941e-05,
+      "loss": 0.2729,
+      "step": 18794
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3687975416855574e-05,
+      "loss": 0.2801,
+      "step": 18795
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.368763958154921e-05,
+      "loss": 0.2131,
+      "step": 18796
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3687303746242844e-05,
+      "loss": 0.401,
+      "step": 18797
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3686967910936477e-05,
+      "loss": 0.4961,
+      "step": 18798
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3686632075630113e-05,
+      "loss": 0.3871,
+      "step": 18799
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3686296240323746e-05,
+      "loss": 0.2175,
+      "step": 18800
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.368596040501738e-05,
+      "loss": 0.2096,
+      "step": 18801
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3685624569711015e-05,
+      "loss": 0.3744,
+      "step": 18802
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.368528873440465e-05,
+      "loss": 0.3001,
+      "step": 18803
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3684952899098283e-05,
+      "loss": 0.3524,
+      "step": 18804
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3684617063791919e-05,
+      "loss": 0.3753,
+      "step": 18805
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3684281228485552e-05,
+      "loss": 0.358,
+      "step": 18806
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3683945393179185e-05,
+      "loss": 0.3306,
+      "step": 18807
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3683609557872821e-05,
+      "loss": 0.3489,
+      "step": 18808
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3683273722566454e-05,
+      "loss": 0.3506,
+      "step": 18809
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3682937887260089e-05,
+      "loss": 0.2468,
+      "step": 18810
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3682602051953724e-05,
+      "loss": 0.3101,
+      "step": 18811
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3682266216647358e-05,
+      "loss": 0.351,
+      "step": 18812
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3681930381340991e-05,
+      "loss": 0.3332,
+      "step": 18813
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3681594546034627e-05,
+      "loss": 0.3175,
+      "step": 18814
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.368125871072826e-05,
+      "loss": 0.2062,
+      "step": 18815
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3680922875421893e-05,
+      "loss": 0.3278,
+      "step": 18816
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.368058704011553e-05,
+      "loss": 0.4903,
+      "step": 18817
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3680251204809163e-05,
+      "loss": 0.2724,
+      "step": 18818
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3679915369502796e-05,
+      "loss": 0.3515,
+      "step": 18819
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3679579534196432e-05,
+      "loss": 0.2702,
+      "step": 18820
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3679243698890065e-05,
+      "loss": 0.2884,
+      "step": 18821
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.36789078635837e-05,
+      "loss": 0.2363,
+      "step": 18822
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3678572028277334e-05,
+      "loss": 0.3244,
+      "step": 18823
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3678236192970969e-05,
+      "loss": 0.2735,
+      "step": 18824
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3677900357664602e-05,
+      "loss": 0.3843,
+      "step": 18825
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3677564522358238e-05,
+      "loss": 0.3141,
+      "step": 18826
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3677228687051871e-05,
+      "loss": 0.4596,
+      "step": 18827
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3676892851745504e-05,
+      "loss": 0.2564,
+      "step": 18828
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.367655701643914e-05,
+      "loss": 0.3894,
+      "step": 18829
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3676221181132773e-05,
+      "loss": 0.2987,
+      "step": 18830
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3675885345826408e-05,
+      "loss": 0.283,
+      "step": 18831
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3675549510520043e-05,
+      "loss": 0.4009,
+      "step": 18832
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3675213675213677e-05,
+      "loss": 0.1603,
+      "step": 18833
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.367487783990731e-05,
+      "loss": 0.2415,
+      "step": 18834
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3674542004600947e-05,
+      "loss": 0.3021,
+      "step": 18835
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.367420616929458e-05,
+      "loss": 0.4235,
+      "step": 18836
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3673870333988213e-05,
+      "loss": 0.3375,
+      "step": 18837
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3673534498681849e-05,
+      "loss": 0.3806,
+      "step": 18838
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3673198663375482e-05,
+      "loss": 0.3333,
+      "step": 18839
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3672862828069115e-05,
+      "loss": 0.4261,
+      "step": 18840
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3672526992762751e-05,
+      "loss": 0.1275,
+      "step": 18841
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3672191157456384e-05,
+      "loss": 0.2504,
+      "step": 18842
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3671855322150019e-05,
+      "loss": 0.2997,
+      "step": 18843
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3671519486843653e-05,
+      "loss": 0.2598,
+      "step": 18844
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3671183651537288e-05,
+      "loss": 0.4175,
+      "step": 18845
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3670847816230921e-05,
+      "loss": 0.2934,
+      "step": 18846
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3670511980924557e-05,
+      "loss": 0.3941,
+      "step": 18847
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.367017614561819e-05,
+      "loss": 0.4273,
+      "step": 18848
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3669840310311823e-05,
+      "loss": 0.3333,
+      "step": 18849
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.366950447500546e-05,
+      "loss": 0.236,
+      "step": 18850
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3669168639699092e-05,
+      "loss": 0.3561,
+      "step": 18851
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3668832804392727e-05,
+      "loss": 0.4115,
+      "step": 18852
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3668496969086362e-05,
+      "loss": 0.1997,
+      "step": 18853
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3668161133779996e-05,
+      "loss": 0.2859,
+      "step": 18854
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.366782529847363e-05,
+      "loss": 0.2479,
+      "step": 18855
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3667489463167266e-05,
+      "loss": 0.4007,
+      "step": 18856
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3667153627860899e-05,
+      "loss": 0.494,
+      "step": 18857
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3666817792554532e-05,
+      "loss": 0.32,
+      "step": 18858
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3666481957248166e-05,
+      "loss": 0.2716,
+      "step": 18859
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3666146121941801e-05,
+      "loss": 0.277,
+      "step": 18860
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3665810286635436e-05,
+      "loss": 0.3866,
+      "step": 18861
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3665474451329069e-05,
+      "loss": 0.2238,
+      "step": 18862
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3665138616022703e-05,
+      "loss": 0.3328,
+      "step": 18863
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3664802780716338e-05,
+      "loss": 0.1558,
+      "step": 18864
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.366446694540997e-05,
+      "loss": 0.3774,
+      "step": 18865
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3664131110103607e-05,
+      "loss": 0.1637,
+      "step": 18866
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.366379527479724e-05,
+      "loss": 0.1646,
+      "step": 18867
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3663459439490873e-05,
+      "loss": 0.3305,
+      "step": 18868
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.366312360418451e-05,
+      "loss": 0.3327,
+      "step": 18869
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3662787768878142e-05,
+      "loss": 0.3487,
+      "step": 18870
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3662451933571777e-05,
+      "loss": 0.1511,
+      "step": 18871
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3662116098265412e-05,
+      "loss": 0.1632,
+      "step": 18872
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3661780262959046e-05,
+      "loss": 0.2705,
+      "step": 18873
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.366144442765268e-05,
+      "loss": 0.3579,
+      "step": 18874
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3661108592346316e-05,
+      "loss": 0.3866,
+      "step": 18875
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3660772757039949e-05,
+      "loss": 0.3218,
+      "step": 18876
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3660436921733581e-05,
+      "loss": 0.4135,
+      "step": 18877
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3660101086427218e-05,
+      "loss": 0.2969,
+      "step": 18878
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.365976525112085e-05,
+      "loss": 0.2967,
+      "step": 18879
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3659429415814485e-05,
+      "loss": 0.4072,
+      "step": 18880
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.365909358050812e-05,
+      "loss": 0.4082,
+      "step": 18881
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3658757745201755e-05,
+      "loss": 0.328,
+      "step": 18882
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3658421909895388e-05,
+      "loss": 0.3784,
+      "step": 18883
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3658086074589022e-05,
+      "loss": 0.3125,
+      "step": 18884
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3657750239282657e-05,
+      "loss": 0.3754,
+      "step": 18885
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.365741440397629e-05,
+      "loss": 0.3003,
+      "step": 18886
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3657078568669926e-05,
+      "loss": 0.37,
+      "step": 18887
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.365674273336356e-05,
+      "loss": 0.3838,
+      "step": 18888
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3656406898057192e-05,
+      "loss": 0.2584,
+      "step": 18889
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3656071062750828e-05,
+      "loss": 0.4207,
+      "step": 18890
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3655735227444461e-05,
+      "loss": 0.3176,
+      "step": 18891
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3655399392138096e-05,
+      "loss": 0.173,
+      "step": 18892
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.365506355683173e-05,
+      "loss": 0.4014,
+      "step": 18893
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3654727721525365e-05,
+      "loss": 0.2776,
+      "step": 18894
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3654391886218998e-05,
+      "loss": 0.2941,
+      "step": 18895
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3654056050912635e-05,
+      "loss": 0.3847,
+      "step": 18896
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3653720215606268e-05,
+      "loss": 0.2377,
+      "step": 18897
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.36533843802999e-05,
+      "loss": 0.273,
+      "step": 18898
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3653048544993537e-05,
+      "loss": 0.3526,
+      "step": 18899
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.365271270968717e-05,
+      "loss": 0.4453,
+      "step": 18900
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3652376874380805e-05,
+      "loss": 0.2291,
+      "step": 18901
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.365204103907444e-05,
+      "loss": 0.5198,
+      "step": 18902
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3651705203768074e-05,
+      "loss": 0.3443,
+      "step": 18903
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3651369368461707e-05,
+      "loss": 0.3809,
+      "step": 18904
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3651033533155343e-05,
+      "loss": 0.363,
+      "step": 18905
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3650697697848976e-05,
+      "loss": 0.2412,
+      "step": 18906
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3650361862542609e-05,
+      "loss": 0.279,
+      "step": 18907
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3650026027236245e-05,
+      "loss": 0.2548,
+      "step": 18908
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3649690191929878e-05,
+      "loss": 0.2502,
+      "step": 18909
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3649354356623511e-05,
+      "loss": 0.3454,
+      "step": 18910
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3649018521317148e-05,
+      "loss": 0.4366,
+      "step": 18911
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.364868268601078e-05,
+      "loss": 0.4136,
+      "step": 18912
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3648346850704415e-05,
+      "loss": 0.4245,
+      "step": 18913
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.364801101539805e-05,
+      "loss": 0.2487,
+      "step": 18914
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3647675180091685e-05,
+      "loss": 0.3415,
+      "step": 18915
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3647339344785317e-05,
+      "loss": 0.3136,
+      "step": 18916
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3647003509478954e-05,
+      "loss": 0.605,
+      "step": 18917
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3646667674172587e-05,
+      "loss": 0.1876,
+      "step": 18918
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.364633183886622e-05,
+      "loss": 0.258,
+      "step": 18919
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3645996003559856e-05,
+      "loss": 0.4004,
+      "step": 18920
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3645660168253489e-05,
+      "loss": 0.2864,
+      "step": 18921
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3645324332947124e-05,
+      "loss": 0.5298,
+      "step": 18922
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3644988497640758e-05,
+      "loss": 0.4141,
+      "step": 18923
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3644652662334393e-05,
+      "loss": 0.3179,
+      "step": 18924
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3644316827028026e-05,
+      "loss": 0.3566,
+      "step": 18925
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3643980991721662e-05,
+      "loss": 0.3279,
+      "step": 18926
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3643645156415295e-05,
+      "loss": 0.2777,
+      "step": 18927
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3643309321108928e-05,
+      "loss": 0.3538,
+      "step": 18928
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3642973485802564e-05,
+      "loss": 0.3315,
+      "step": 18929
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3642637650496197e-05,
+      "loss": 0.2394,
+      "step": 18930
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.364230181518983e-05,
+      "loss": 0.4196,
+      "step": 18931
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3641965979883467e-05,
+      "loss": 0.3314,
+      "step": 18932
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.36416301445771e-05,
+      "loss": 0.319,
+      "step": 18933
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3641294309270734e-05,
+      "loss": 0.3738,
+      "step": 18934
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3640958473964369e-05,
+      "loss": 0.2304,
+      "step": 18935
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3640622638658004e-05,
+      "loss": 0.2636,
+      "step": 18936
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3640286803351637e-05,
+      "loss": 0.3998,
+      "step": 18937
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3639950968045273e-05,
+      "loss": 0.4242,
+      "step": 18938
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3639615132738906e-05,
+      "loss": 0.3906,
+      "step": 18939
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3639279297432539e-05,
+      "loss": 0.2208,
+      "step": 18940
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3638943462126175e-05,
+      "loss": 0.1458,
+      "step": 18941
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3638607626819808e-05,
+      "loss": 0.3973,
+      "step": 18942
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3638271791513443e-05,
+      "loss": 0.2364,
+      "step": 18943
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3637935956207077e-05,
+      "loss": 0.2757,
+      "step": 18944
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3637600120900712e-05,
+      "loss": 0.3097,
+      "step": 18945
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3637264285594345e-05,
+      "loss": 0.3217,
+      "step": 18946
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3636928450287981e-05,
+      "loss": 0.2716,
+      "step": 18947
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3636592614981614e-05,
+      "loss": 0.3428,
+      "step": 18948
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3636256779675247e-05,
+      "loss": 0.5497,
+      "step": 18949
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3635920944368884e-05,
+      "loss": 0.3204,
+      "step": 18950
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3635585109062517e-05,
+      "loss": 0.2098,
+      "step": 18951
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3635249273756151e-05,
+      "loss": 0.4777,
+      "step": 18952
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3634913438449786e-05,
+      "loss": 0.2251,
+      "step": 18953
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3634577603143419e-05,
+      "loss": 0.3016,
+      "step": 18954
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3634241767837053e-05,
+      "loss": 0.3146,
+      "step": 18955
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3633905932530688e-05,
+      "loss": 0.4562,
+      "step": 18956
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.3633570097224323e-05,
+      "loss": 0.2122,
+      "step": 18957
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3633234261917956e-05,
+      "loss": 0.4445,
+      "step": 18958
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3632898426611592e-05,
+      "loss": 0.3371,
+      "step": 18959
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3632562591305225e-05,
+      "loss": 0.346,
+      "step": 18960
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3632226755998858e-05,
+      "loss": 0.2161,
+      "step": 18961
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3631890920692494e-05,
+      "loss": 0.2169,
+      "step": 18962
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3631555085386127e-05,
+      "loss": 0.3891,
+      "step": 18963
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3631219250079762e-05,
+      "loss": 0.318,
+      "step": 18964
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3630883414773397e-05,
+      "loss": 0.2358,
+      "step": 18965
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3630547579467031e-05,
+      "loss": 0.4148,
+      "step": 18966
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3630211744160664e-05,
+      "loss": 0.1951,
+      "step": 18967
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.36298759088543e-05,
+      "loss": 0.3096,
+      "step": 18968
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3629540073547933e-05,
+      "loss": 0.2915,
+      "step": 18969
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3629204238241566e-05,
+      "loss": 0.3184,
+      "step": 18970
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3628868402935203e-05,
+      "loss": 0.2946,
+      "step": 18971
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3628532567628836e-05,
+      "loss": 0.3442,
+      "step": 18972
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.362819673232247e-05,
+      "loss": 0.383,
+      "step": 18973
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3627860897016105e-05,
+      "loss": 0.4038,
+      "step": 18974
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3627525061709738e-05,
+      "loss": 0.2764,
+      "step": 18975
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3627189226403373e-05,
+      "loss": 0.3628,
+      "step": 18976
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3626853391097007e-05,
+      "loss": 0.2123,
+      "step": 18977
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3626517555790642e-05,
+      "loss": 0.4103,
+      "step": 18978
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3626181720484275e-05,
+      "loss": 0.2702,
+      "step": 18979
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3625845885177911e-05,
+      "loss": 0.344,
+      "step": 18980
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3625510049871544e-05,
+      "loss": 0.2281,
+      "step": 18981
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3625174214565177e-05,
+      "loss": 0.2677,
+      "step": 18982
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3624838379258813e-05,
+      "loss": 0.2651,
+      "step": 18983
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3624502543952446e-05,
+      "loss": 0.1891,
+      "step": 18984
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3624166708646081e-05,
+      "loss": 0.4586,
+      "step": 18985
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3623830873339716e-05,
+      "loss": 0.3249,
+      "step": 18986
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.362349503803335e-05,
+      "loss": 0.2437,
+      "step": 18987
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3623159202726983e-05,
+      "loss": 0.5262,
+      "step": 18988
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.362282336742062e-05,
+      "loss": 0.2103,
+      "step": 18989
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3622487532114253e-05,
+      "loss": 0.3425,
+      "step": 18990
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3622151696807886e-05,
+      "loss": 0.2429,
+      "step": 18991
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3621815861501522e-05,
+      "loss": 0.4026,
+      "step": 18992
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3621480026195155e-05,
+      "loss": 0.3592,
+      "step": 18993
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.362114419088879e-05,
+      "loss": 0.3543,
+      "step": 18994
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3620808355582424e-05,
+      "loss": 0.2494,
+      "step": 18995
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3620472520276059e-05,
+      "loss": 0.2606,
+      "step": 18996
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3620136684969692e-05,
+      "loss": 0.4012,
+      "step": 18997
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3619800849663326e-05,
+      "loss": 0.3063,
+      "step": 18998
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3619465014356961e-05,
+      "loss": 0.2084,
+      "step": 18999
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3619129179050594e-05,
+      "loss": 0.295,
+      "step": 19000
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.361879334374423e-05,
+      "loss": 0.1936,
+      "step": 19001
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3618457508437863e-05,
+      "loss": 0.4265,
+      "step": 19002
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3618121673131496e-05,
+      "loss": 0.2206,
+      "step": 19003
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3617785837825133e-05,
+      "loss": 0.503,
+      "step": 19004
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3617450002518766e-05,
+      "loss": 0.2387,
+      "step": 19005
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.36171141672124e-05,
+      "loss": 0.5001,
+      "step": 19006
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3616778331906035e-05,
+      "loss": 0.3503,
+      "step": 19007
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.361644249659967e-05,
+      "loss": 0.3864,
+      "step": 19008
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3616106661293302e-05,
+      "loss": 0.367,
+      "step": 19009
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3615770825986939e-05,
+      "loss": 0.167,
+      "step": 19010
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3615434990680572e-05,
+      "loss": 0.2879,
+      "step": 19011
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3615099155374205e-05,
+      "loss": 0.3031,
+      "step": 19012
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3614763320067841e-05,
+      "loss": 0.4454,
+      "step": 19013
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3614427484761474e-05,
+      "loss": 0.3882,
+      "step": 19014
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3614091649455109e-05,
+      "loss": 0.4306,
+      "step": 19015
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3613755814148743e-05,
+      "loss": 0.3526,
+      "step": 19016
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3613419978842378e-05,
+      "loss": 0.2022,
+      "step": 19017
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.361308414353601e-05,
+      "loss": 0.3368,
+      "step": 19018
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3612748308229645e-05,
+      "loss": 0.2112,
+      "step": 19019
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.361241247292328e-05,
+      "loss": 0.2214,
+      "step": 19020
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3612076637616913e-05,
+      "loss": 0.5134,
+      "step": 19021
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.361174080231055e-05,
+      "loss": 0.3344,
+      "step": 19022
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3611404967004182e-05,
+      "loss": 0.5027,
+      "step": 19023
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3611069131697815e-05,
+      "loss": 0.2026,
+      "step": 19024
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3610733296391452e-05,
+      "loss": 0.3513,
+      "step": 19025
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3610397461085085e-05,
+      "loss": 0.1815,
+      "step": 19026
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.361006162577872e-05,
+      "loss": 0.377,
+      "step": 19027
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3609725790472354e-05,
+      "loss": 0.2958,
+      "step": 19028
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3609389955165989e-05,
+      "loss": 0.3172,
+      "step": 19029
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3609054119859622e-05,
+      "loss": 0.1273,
+      "step": 19030
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3608718284553258e-05,
+      "loss": 0.1307,
+      "step": 19031
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.360838244924689e-05,
+      "loss": 0.1884,
+      "step": 19032
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3608046613940524e-05,
+      "loss": 0.5251,
+      "step": 19033
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.360771077863416e-05,
+      "loss": 0.4875,
+      "step": 19034
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3607374943327793e-05,
+      "loss": 0.1617,
+      "step": 19035
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3607039108021428e-05,
+      "loss": 0.3089,
+      "step": 19036
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3606703272715062e-05,
+      "loss": 0.3054,
+      "step": 19037
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3606367437408697e-05,
+      "loss": 0.3906,
+      "step": 19038
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.360603160210233e-05,
+      "loss": 0.5226,
+      "step": 19039
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3605695766795963e-05,
+      "loss": 0.2102,
+      "step": 19040
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.36053599314896e-05,
+      "loss": 0.3483,
+      "step": 19041
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3605024096183232e-05,
+      "loss": 0.4062,
+      "step": 19042
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3604688260876865e-05,
+      "loss": 0.2516,
+      "step": 19043
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3604352425570502e-05,
+      "loss": 0.3031,
+      "step": 19044
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3604016590264134e-05,
+      "loss": 0.1452,
+      "step": 19045
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3603680754957769e-05,
+      "loss": 0.3003,
+      "step": 19046
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3603344919651404e-05,
+      "loss": 0.235,
+      "step": 19047
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3603009084345038e-05,
+      "loss": 0.2845,
+      "step": 19048
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3602673249038671e-05,
+      "loss": 0.3871,
+      "step": 19049
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3602337413732308e-05,
+      "loss": 0.3992,
+      "step": 19050
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.360200157842594e-05,
+      "loss": 0.3189,
+      "step": 19051
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3601665743119574e-05,
+      "loss": 0.2789,
+      "step": 19052
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.360132990781321e-05,
+      "loss": 0.2135,
+      "step": 19053
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3600994072506843e-05,
+      "loss": 0.3382,
+      "step": 19054
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3600658237200478e-05,
+      "loss": 0.2376,
+      "step": 19055
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3600322401894112e-05,
+      "loss": 0.5291,
+      "step": 19056
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3599986566587747e-05,
+      "loss": 0.4512,
+      "step": 19057
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.359965073128138e-05,
+      "loss": 0.2137,
+      "step": 19058
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3599314895975016e-05,
+      "loss": 0.3179,
+      "step": 19059
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3598979060668649e-05,
+      "loss": 0.2311,
+      "step": 19060
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3598643225362282e-05,
+      "loss": 0.3327,
+      "step": 19061
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3598307390055918e-05,
+      "loss": 0.3038,
+      "step": 19062
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3597971554749551e-05,
+      "loss": 0.4044,
+      "step": 19063
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3597635719443186e-05,
+      "loss": 0.2589,
+      "step": 19064
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.359729988413682e-05,
+      "loss": 0.2369,
+      "step": 19065
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3596964048830454e-05,
+      "loss": 0.3815,
+      "step": 19066
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3596628213524088e-05,
+      "loss": 0.2705,
+      "step": 19067
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3596292378217723e-05,
+      "loss": 0.2554,
+      "step": 19068
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3595956542911358e-05,
+      "loss": 0.3994,
+      "step": 19069
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.359562070760499e-05,
+      "loss": 0.34,
+      "step": 19070
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3595284872298627e-05,
+      "loss": 0.2831,
+      "step": 19071
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.359494903699226e-05,
+      "loss": 0.3882,
+      "step": 19072
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3594613201685893e-05,
+      "loss": 0.2726,
+      "step": 19073
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3594277366379529e-05,
+      "loss": 0.2788,
+      "step": 19074
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3593941531073162e-05,
+      "loss": 0.4461,
+      "step": 19075
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3593605695766797e-05,
+      "loss": 0.363,
+      "step": 19076
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3593269860460431e-05,
+      "loss": 0.3903,
+      "step": 19077
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3592934025154066e-05,
+      "loss": 0.3533,
+      "step": 19078
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3592598189847699e-05,
+      "loss": 0.1975,
+      "step": 19079
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3592262354541335e-05,
+      "loss": 0.3474,
+      "step": 19080
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3591926519234968e-05,
+      "loss": 0.2733,
+      "step": 19081
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3591590683928601e-05,
+      "loss": 0.2543,
+      "step": 19082
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3591254848622238e-05,
+      "loss": 0.3151,
+      "step": 19083
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.359091901331587e-05,
+      "loss": 0.3531,
+      "step": 19084
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3590583178009505e-05,
+      "loss": 0.2682,
+      "step": 19085
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.359024734270314e-05,
+      "loss": 0.3677,
+      "step": 19086
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3589911507396774e-05,
+      "loss": 0.3506,
+      "step": 19087
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3589575672090407e-05,
+      "loss": 0.3333,
+      "step": 19088
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3589239836784042e-05,
+      "loss": 0.1879,
+      "step": 19089
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3588904001477677e-05,
+      "loss": 0.293,
+      "step": 19090
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.358856816617131e-05,
+      "loss": 0.3138,
+      "step": 19091
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3588232330864946e-05,
+      "loss": 0.34,
+      "step": 19092
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3587896495558579e-05,
+      "loss": 0.3487,
+      "step": 19093
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3587560660252212e-05,
+      "loss": 0.475,
+      "step": 19094
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3587224824945848e-05,
+      "loss": 0.448,
+      "step": 19095
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3586888989639481e-05,
+      "loss": 0.2813,
+      "step": 19096
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3586553154333116e-05,
+      "loss": 0.228,
+      "step": 19097
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.358621731902675e-05,
+      "loss": 0.1737,
+      "step": 19098
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3585881483720385e-05,
+      "loss": 0.3281,
+      "step": 19099
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3585545648414018e-05,
+      "loss": 0.207,
+      "step": 19100
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3585209813107654e-05,
+      "loss": 0.3169,
+      "step": 19101
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3584873977801287e-05,
+      "loss": 0.2602,
+      "step": 19102
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.358453814249492e-05,
+      "loss": 0.2274,
+      "step": 19103
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3584202307188557e-05,
+      "loss": 0.3267,
+      "step": 19104
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.358386647188219e-05,
+      "loss": 0.3321,
+      "step": 19105
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3583530636575824e-05,
+      "loss": 0.365,
+      "step": 19106
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3583194801269459e-05,
+      "loss": 0.2257,
+      "step": 19107
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3582858965963094e-05,
+      "loss": 0.2641,
+      "step": 19108
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3582523130656726e-05,
+      "loss": 0.3473,
+      "step": 19109
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3582187295350361e-05,
+      "loss": 0.3347,
+      "step": 19110
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3581851460043996e-05,
+      "loss": 0.3298,
+      "step": 19111
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3581515624737629e-05,
+      "loss": 0.2886,
+      "step": 19112
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3581179789431265e-05,
+      "loss": 0.1931,
+      "step": 19113
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3580843954124898e-05,
+      "loss": 0.4149,
+      "step": 19114
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3580508118818531e-05,
+      "loss": 0.2716,
+      "step": 19115
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3580172283512167e-05,
+      "loss": 0.15,
+      "step": 19116
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.35798364482058e-05,
+      "loss": 0.2559,
+      "step": 19117
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3579500612899435e-05,
+      "loss": 0.2876,
+      "step": 19118
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.357916477759307e-05,
+      "loss": 0.2797,
+      "step": 19119
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3578828942286704e-05,
+      "loss": 0.2094,
+      "step": 19120
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3578493106980337e-05,
+      "loss": 0.3087,
+      "step": 19121
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3578157271673974e-05,
+      "loss": 0.3188,
+      "step": 19122
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3577821436367606e-05,
+      "loss": 0.3807,
+      "step": 19123
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.357748560106124e-05,
+      "loss": 0.1941,
+      "step": 19124
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3577149765754876e-05,
+      "loss": 0.4069,
+      "step": 19125
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3576813930448509e-05,
+      "loss": 0.5019,
+      "step": 19126
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3576478095142143e-05,
+      "loss": 0.2727,
+      "step": 19127
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3576142259835778e-05,
+      "loss": 0.2853,
+      "step": 19128
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3575806424529413e-05,
+      "loss": 0.2132,
+      "step": 19129
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3575470589223046e-05,
+      "loss": 0.297,
+      "step": 19130
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3575134753916682e-05,
+      "loss": 0.2353,
+      "step": 19131
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3574798918610315e-05,
+      "loss": 0.3315,
+      "step": 19132
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3574463083303948e-05,
+      "loss": 0.1715,
+      "step": 19133
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3574127247997584e-05,
+      "loss": 0.4252,
+      "step": 19134
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3573791412691217e-05,
+      "loss": 0.2607,
+      "step": 19135
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.357345557738485e-05,
+      "loss": 0.2244,
+      "step": 19136
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3573119742078486e-05,
+      "loss": 0.2691,
+      "step": 19137
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.357278390677212e-05,
+      "loss": 0.3832,
+      "step": 19138
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3572448071465754e-05,
+      "loss": 0.4761,
+      "step": 19139
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3572112236159389e-05,
+      "loss": 0.2358,
+      "step": 19140
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3571776400853023e-05,
+      "loss": 0.2502,
+      "step": 19141
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3571440565546656e-05,
+      "loss": 0.448,
+      "step": 19142
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3571104730240293e-05,
+      "loss": 0.3054,
+      "step": 19143
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3570768894933926e-05,
+      "loss": 0.4795,
+      "step": 19144
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3570433059627559e-05,
+      "loss": 0.1947,
+      "step": 19145
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3570097224321195e-05,
+      "loss": 0.2383,
+      "step": 19146
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3569761389014828e-05,
+      "loss": 0.37,
+      "step": 19147
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3569425553708462e-05,
+      "loss": 0.1826,
+      "step": 19148
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3569089718402097e-05,
+      "loss": 0.2588,
+      "step": 19149
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3568753883095732e-05,
+      "loss": 0.3886,
+      "step": 19150
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3568418047789365e-05,
+      "loss": 0.3752,
+      "step": 19151
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3568082212483001e-05,
+      "loss": 0.0825,
+      "step": 19152
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3567746377176634e-05,
+      "loss": 0.1989,
+      "step": 19153
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3567410541870267e-05,
+      "loss": 0.2841,
+      "step": 19154
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3567074706563903e-05,
+      "loss": 0.2656,
+      "step": 19155
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.3566738871257536e-05,
+      "loss": 0.5406,
+      "step": 19156
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.356640303595117e-05,
+      "loss": 0.4665,
+      "step": 19157
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3566067200644806e-05,
+      "loss": 0.3175,
+      "step": 19158
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3565731365338439e-05,
+      "loss": 0.2115,
+      "step": 19159
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3565395530032073e-05,
+      "loss": 0.374,
+      "step": 19160
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3565059694725708e-05,
+      "loss": 0.4822,
+      "step": 19161
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3564723859419342e-05,
+      "loss": 0.2259,
+      "step": 19162
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3564388024112975e-05,
+      "loss": 0.4809,
+      "step": 19163
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3564052188806612e-05,
+      "loss": 0.3165,
+      "step": 19164
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3563716353500245e-05,
+      "loss": 0.2512,
+      "step": 19165
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3563380518193878e-05,
+      "loss": 0.4931,
+      "step": 19166
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3563044682887514e-05,
+      "loss": 0.419,
+      "step": 19167
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3562708847581147e-05,
+      "loss": 0.2093,
+      "step": 19168
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3562373012274782e-05,
+      "loss": 0.2065,
+      "step": 19169
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3562037176968416e-05,
+      "loss": 0.2653,
+      "step": 19170
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3561701341662051e-05,
+      "loss": 0.1682,
+      "step": 19171
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3561365506355684e-05,
+      "loss": 0.3576,
+      "step": 19172
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.356102967104932e-05,
+      "loss": 0.1299,
+      "step": 19173
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3560693835742953e-05,
+      "loss": 0.3183,
+      "step": 19174
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3560358000436586e-05,
+      "loss": 0.3932,
+      "step": 19175
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3560022165130222e-05,
+      "loss": 0.4919,
+      "step": 19176
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3559686329823855e-05,
+      "loss": 0.1843,
+      "step": 19177
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3559350494517488e-05,
+      "loss": 0.2854,
+      "step": 19178
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3559014659211125e-05,
+      "loss": 0.3214,
+      "step": 19179
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3558678823904758e-05,
+      "loss": 0.3909,
+      "step": 19180
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3558342988598392e-05,
+      "loss": 0.3149,
+      "step": 19181
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3558007153292027e-05,
+      "loss": 0.2446,
+      "step": 19182
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3557671317985662e-05,
+      "loss": 0.5525,
+      "step": 19183
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3557335482679295e-05,
+      "loss": 0.1924,
+      "step": 19184
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3556999647372931e-05,
+      "loss": 0.3829,
+      "step": 19185
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3556663812066564e-05,
+      "loss": 0.3133,
+      "step": 19186
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3556327976760197e-05,
+      "loss": 0.2144,
+      "step": 19187
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3555992141453833e-05,
+      "loss": 0.3525,
+      "step": 19188
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3555656306147466e-05,
+      "loss": 0.2747,
+      "step": 19189
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.35553204708411e-05,
+      "loss": 0.2363,
+      "step": 19190
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3554984635534735e-05,
+      "loss": 0.2491,
+      "step": 19191
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.355464880022837e-05,
+      "loss": 0.3357,
+      "step": 19192
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3554312964922003e-05,
+      "loss": 0.411,
+      "step": 19193
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.355397712961564e-05,
+      "loss": 0.3219,
+      "step": 19194
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3553641294309272e-05,
+      "loss": 0.2205,
+      "step": 19195
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3553305459002905e-05,
+      "loss": 0.2122,
+      "step": 19196
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3552969623696542e-05,
+      "loss": 0.1205,
+      "step": 19197
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3552633788390175e-05,
+      "loss": 0.2767,
+      "step": 19198
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.355229795308381e-05,
+      "loss": 0.2107,
+      "step": 19199
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3551962117777444e-05,
+      "loss": 0.3009,
+      "step": 19200
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3551626282471077e-05,
+      "loss": 0.1542,
+      "step": 19201
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3551290447164711e-05,
+      "loss": 0.1592,
+      "step": 19202
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3550954611858346e-05,
+      "loss": 0.3072,
+      "step": 19203
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.355061877655198e-05,
+      "loss": 0.2113,
+      "step": 19204
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3550282941245614e-05,
+      "loss": 0.3891,
+      "step": 19205
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.354994710593925e-05,
+      "loss": 0.3516,
+      "step": 19206
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3549611270632883e-05,
+      "loss": 0.2309,
+      "step": 19207
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3549275435326516e-05,
+      "loss": 0.2537,
+      "step": 19208
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3548939600020152e-05,
+      "loss": 0.5299,
+      "step": 19209
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3548603764713785e-05,
+      "loss": 0.3563,
+      "step": 19210
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.354826792940742e-05,
+      "loss": 0.3231,
+      "step": 19211
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3547932094101055e-05,
+      "loss": 0.1774,
+      "step": 19212
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3547596258794689e-05,
+      "loss": 0.3573,
+      "step": 19213
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3547260423488322e-05,
+      "loss": 0.5736,
+      "step": 19214
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3546924588181958e-05,
+      "loss": 0.2483,
+      "step": 19215
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3546588752875591e-05,
+      "loss": 0.2361,
+      "step": 19216
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3546252917569224e-05,
+      "loss": 0.2683,
+      "step": 19217
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.354591708226286e-05,
+      "loss": 0.3453,
+      "step": 19218
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3545581246956494e-05,
+      "loss": 0.184,
+      "step": 19219
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3545245411650128e-05,
+      "loss": 0.3052,
+      "step": 19220
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3544909576343761e-05,
+      "loss": 0.3359,
+      "step": 19221
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3544573741037398e-05,
+      "loss": 0.3682,
+      "step": 19222
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.354423790573103e-05,
+      "loss": 0.5494,
+      "step": 19223
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3543902070424663e-05,
+      "loss": 0.3628,
+      "step": 19224
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.35435662351183e-05,
+      "loss": 0.2998,
+      "step": 19225
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3543230399811933e-05,
+      "loss": 0.3012,
+      "step": 19226
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3542894564505566e-05,
+      "loss": 0.3998,
+      "step": 19227
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3542558729199202e-05,
+      "loss": 0.2918,
+      "step": 19228
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3542222893892835e-05,
+      "loss": 0.336,
+      "step": 19229
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.354188705858647e-05,
+      "loss": 0.297,
+      "step": 19230
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3541551223280104e-05,
+      "loss": 0.4088,
+      "step": 19231
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3541215387973739e-05,
+      "loss": 0.4281,
+      "step": 19232
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3540879552667372e-05,
+      "loss": 0.3227,
+      "step": 19233
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3540543717361008e-05,
+      "loss": 0.2128,
+      "step": 19234
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3540207882054641e-05,
+      "loss": 0.2652,
+      "step": 19235
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3539872046748274e-05,
+      "loss": 0.3592,
+      "step": 19236
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.353953621144191e-05,
+      "loss": 0.2578,
+      "step": 19237
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3539200376135543e-05,
+      "loss": 0.2618,
+      "step": 19238
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3538864540829178e-05,
+      "loss": 0.2586,
+      "step": 19239
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3538528705522813e-05,
+      "loss": 0.2803,
+      "step": 19240
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3538192870216447e-05,
+      "loss": 0.4137,
+      "step": 19241
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.353785703491008e-05,
+      "loss": 0.308,
+      "step": 19242
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3537521199603717e-05,
+      "loss": 0.2694,
+      "step": 19243
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.353718536429735e-05,
+      "loss": 0.2645,
+      "step": 19244
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3536849528990983e-05,
+      "loss": 0.2914,
+      "step": 19245
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3536513693684619e-05,
+      "loss": 0.3953,
+      "step": 19246
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3536177858378252e-05,
+      "loss": 0.2884,
+      "step": 19247
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3535842023071885e-05,
+      "loss": 0.3883,
+      "step": 19248
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3535506187765521e-05,
+      "loss": 0.3573,
+      "step": 19249
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3535170352459154e-05,
+      "loss": 0.2488,
+      "step": 19250
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3534834517152789e-05,
+      "loss": 0.4914,
+      "step": 19251
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3534498681846423e-05,
+      "loss": 0.2476,
+      "step": 19252
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3534162846540058e-05,
+      "loss": 0.6841,
+      "step": 19253
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3533827011233691e-05,
+      "loss": 0.2171,
+      "step": 19254
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3533491175927327e-05,
+      "loss": 0.3673,
+      "step": 19255
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.353315534062096e-05,
+      "loss": 0.4281,
+      "step": 19256
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3532819505314593e-05,
+      "loss": 0.331,
+      "step": 19257
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.353248367000823e-05,
+      "loss": 0.4354,
+      "step": 19258
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3532147834701863e-05,
+      "loss": 0.178,
+      "step": 19259
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3531811999395497e-05,
+      "loss": 0.289,
+      "step": 19260
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3531476164089132e-05,
+      "loss": 0.4232,
+      "step": 19261
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3531140328782767e-05,
+      "loss": 0.4967,
+      "step": 19262
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.35308044934764e-05,
+      "loss": 0.3772,
+      "step": 19263
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3530468658170036e-05,
+      "loss": 0.2746,
+      "step": 19264
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3530132822863669e-05,
+      "loss": 0.3061,
+      "step": 19265
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3529796987557302e-05,
+      "loss": 0.303,
+      "step": 19266
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3529461152250938e-05,
+      "loss": 0.3317,
+      "step": 19267
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3529125316944571e-05,
+      "loss": 0.3909,
+      "step": 19268
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3528789481638204e-05,
+      "loss": 0.2288,
+      "step": 19269
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.352845364633184e-05,
+      "loss": 0.2593,
+      "step": 19270
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3528117811025473e-05,
+      "loss": 0.2654,
+      "step": 19271
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3527781975719108e-05,
+      "loss": 0.3471,
+      "step": 19272
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3527446140412743e-05,
+      "loss": 0.279,
+      "step": 19273
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3527110305106377e-05,
+      "loss": 0.4405,
+      "step": 19274
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.352677446980001e-05,
+      "loss": 0.3289,
+      "step": 19275
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3526438634493647e-05,
+      "loss": 0.3483,
+      "step": 19276
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.352610279918728e-05,
+      "loss": 0.3147,
+      "step": 19277
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3525766963880912e-05,
+      "loss": 0.3459,
+      "step": 19278
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3525431128574549e-05,
+      "loss": 0.1995,
+      "step": 19279
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3525095293268182e-05,
+      "loss": 0.3512,
+      "step": 19280
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3524759457961816e-05,
+      "loss": 0.2721,
+      "step": 19281
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3524423622655451e-05,
+      "loss": 0.2372,
+      "step": 19282
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3524087787349086e-05,
+      "loss": 0.4429,
+      "step": 19283
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3523751952042719e-05,
+      "loss": 0.3682,
+      "step": 19284
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3523416116736355e-05,
+      "loss": 0.3133,
+      "step": 19285
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3523080281429988e-05,
+      "loss": 0.393,
+      "step": 19286
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3522744446123621e-05,
+      "loss": 0.2763,
+      "step": 19287
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3522408610817257e-05,
+      "loss": 0.4044,
+      "step": 19288
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.352207277551089e-05,
+      "loss": 0.2576,
+      "step": 19289
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3521736940204525e-05,
+      "loss": 0.3228,
+      "step": 19290
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.352140110489816e-05,
+      "loss": 0.2861,
+      "step": 19291
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3521065269591792e-05,
+      "loss": 0.2113,
+      "step": 19292
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3520729434285427e-05,
+      "loss": 0.3821,
+      "step": 19293
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3520393598979062e-05,
+      "loss": 0.3531,
+      "step": 19294
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3520057763672696e-05,
+      "loss": 0.2688,
+      "step": 19295
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.351972192836633e-05,
+      "loss": 0.3488,
+      "step": 19296
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3519386093059966e-05,
+      "loss": 0.4628,
+      "step": 19297
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3519050257753599e-05,
+      "loss": 0.3648,
+      "step": 19298
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3518714422447232e-05,
+      "loss": 0.2329,
+      "step": 19299
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3518378587140868e-05,
+      "loss": 0.2298,
+      "step": 19300
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3518042751834501e-05,
+      "loss": 0.3158,
+      "step": 19301
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3517706916528135e-05,
+      "loss": 0.2034,
+      "step": 19302
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.351737108122177e-05,
+      "loss": 0.2477,
+      "step": 19303
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3517035245915405e-05,
+      "loss": 0.2185,
+      "step": 19304
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3516699410609038e-05,
+      "loss": 0.4053,
+      "step": 19305
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3516363575302674e-05,
+      "loss": 0.4369,
+      "step": 19306
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3516027739996307e-05,
+      "loss": 0.3825,
+      "step": 19307
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.351569190468994e-05,
+      "loss": 0.214,
+      "step": 19308
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3515356069383576e-05,
+      "loss": 0.3906,
+      "step": 19309
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.351502023407721e-05,
+      "loss": 0.1041,
+      "step": 19310
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3514684398770844e-05,
+      "loss": 0.2967,
+      "step": 19311
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3514348563464479e-05,
+      "loss": 0.1992,
+      "step": 19312
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3514012728158112e-05,
+      "loss": 0.3127,
+      "step": 19313
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3513676892851746e-05,
+      "loss": 0.2692,
+      "step": 19314
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.351334105754538e-05,
+      "loss": 0.2437,
+      "step": 19315
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3513005222239015e-05,
+      "loss": 0.2426,
+      "step": 19316
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3512669386932648e-05,
+      "loss": 0.2331,
+      "step": 19317
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3512333551626285e-05,
+      "loss": 0.3559,
+      "step": 19318
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3511997716319918e-05,
+      "loss": 0.3218,
+      "step": 19319
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.351166188101355e-05,
+      "loss": 0.322,
+      "step": 19320
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3511326045707187e-05,
+      "loss": 0.2441,
+      "step": 19321
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.351099021040082e-05,
+      "loss": 0.1882,
+      "step": 19322
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3510654375094455e-05,
+      "loss": 0.4414,
+      "step": 19323
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.351031853978809e-05,
+      "loss": 0.2348,
+      "step": 19324
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3509982704481724e-05,
+      "loss": 0.3248,
+      "step": 19325
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3509646869175357e-05,
+      "loss": 0.3276,
+      "step": 19326
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3509311033868993e-05,
+      "loss": 0.2776,
+      "step": 19327
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3508975198562626e-05,
+      "loss": 0.341,
+      "step": 19328
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3508639363256259e-05,
+      "loss": 0.4787,
+      "step": 19329
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3508303527949895e-05,
+      "loss": 0.2993,
+      "step": 19330
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3507967692643528e-05,
+      "loss": 0.2859,
+      "step": 19331
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3507631857337163e-05,
+      "loss": 0.3229,
+      "step": 19332
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3507296022030798e-05,
+      "loss": 0.3872,
+      "step": 19333
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3506960186724432e-05,
+      "loss": 0.1752,
+      "step": 19334
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3506624351418065e-05,
+      "loss": 0.2443,
+      "step": 19335
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.35062885161117e-05,
+      "loss": 0.252,
+      "step": 19336
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3505952680805335e-05,
+      "loss": 0.2619,
+      "step": 19337
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3505616845498968e-05,
+      "loss": 0.1663,
+      "step": 19338
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3505281010192604e-05,
+      "loss": 0.6762,
+      "step": 19339
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3504945174886237e-05,
+      "loss": 0.4014,
+      "step": 19340
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.350460933957987e-05,
+      "loss": 0.4413,
+      "step": 19341
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3504273504273506e-05,
+      "loss": 0.5213,
+      "step": 19342
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3503937668967139e-05,
+      "loss": 0.2077,
+      "step": 19343
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3503601833660774e-05,
+      "loss": 0.3301,
+      "step": 19344
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3503265998354408e-05,
+      "loss": 0.3854,
+      "step": 19345
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3502930163048043e-05,
+      "loss": 0.2918,
+      "step": 19346
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3502594327741676e-05,
+      "loss": 0.3073,
+      "step": 19347
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3502258492435312e-05,
+      "loss": 0.3472,
+      "step": 19348
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3501922657128945e-05,
+      "loss": 0.3446,
+      "step": 19349
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3501586821822578e-05,
+      "loss": 0.3584,
+      "step": 19350
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3501250986516215e-05,
+      "loss": 0.2234,
+      "step": 19351
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3500915151209848e-05,
+      "loss": 0.3314,
+      "step": 19352
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3500579315903482e-05,
+      "loss": 0.231,
+      "step": 19353
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.3500243480597117e-05,
+      "loss": 0.2163,
+      "step": 19354
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3499907645290751e-05,
+      "loss": 0.3596,
+      "step": 19355
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3499571809984384e-05,
+      "loss": 0.1524,
+      "step": 19356
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.349923597467802e-05,
+      "loss": 0.3766,
+      "step": 19357
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3498900139371654e-05,
+      "loss": 0.2031,
+      "step": 19358
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3498564304065287e-05,
+      "loss": 0.3607,
+      "step": 19359
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3498228468758923e-05,
+      "loss": 0.3097,
+      "step": 19360
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3497892633452556e-05,
+      "loss": 0.218,
+      "step": 19361
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3497556798146189e-05,
+      "loss": 0.2727,
+      "step": 19362
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3497220962839825e-05,
+      "loss": 0.3246,
+      "step": 19363
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3496885127533458e-05,
+      "loss": 0.4188,
+      "step": 19364
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3496549292227093e-05,
+      "loss": 0.2999,
+      "step": 19365
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3496213456920728e-05,
+      "loss": 0.3455,
+      "step": 19366
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3495877621614362e-05,
+      "loss": 0.1938,
+      "step": 19367
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3495541786307995e-05,
+      "loss": 0.31,
+      "step": 19368
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3495205951001631e-05,
+      "loss": 0.3171,
+      "step": 19369
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3494870115695264e-05,
+      "loss": 0.2002,
+      "step": 19370
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3494534280388897e-05,
+      "loss": 0.2515,
+      "step": 19371
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3494198445082534e-05,
+      "loss": 0.4014,
+      "step": 19372
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3493862609776167e-05,
+      "loss": 0.2701,
+      "step": 19373
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3493526774469801e-05,
+      "loss": 0.292,
+      "step": 19374
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3493190939163436e-05,
+      "loss": 0.2068,
+      "step": 19375
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.349285510385707e-05,
+      "loss": 0.2825,
+      "step": 19376
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3492519268550704e-05,
+      "loss": 0.2397,
+      "step": 19377
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.349218343324434e-05,
+      "loss": 0.3793,
+      "step": 19378
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3491847597937973e-05,
+      "loss": 0.3446,
+      "step": 19379
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3491511762631606e-05,
+      "loss": 0.3656,
+      "step": 19380
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3491175927325242e-05,
+      "loss": 0.3478,
+      "step": 19381
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3490840092018875e-05,
+      "loss": 0.2165,
+      "step": 19382
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3490504256712508e-05,
+      "loss": 0.4864,
+      "step": 19383
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3490168421406144e-05,
+      "loss": 0.483,
+      "step": 19384
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3489832586099777e-05,
+      "loss": 0.4333,
+      "step": 19385
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3489496750793412e-05,
+      "loss": 0.4743,
+      "step": 19386
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3489160915487047e-05,
+      "loss": 0.4254,
+      "step": 19387
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3488825080180681e-05,
+      "loss": 0.2369,
+      "step": 19388
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3488489244874314e-05,
+      "loss": 0.3173,
+      "step": 19389
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.348815340956795e-05,
+      "loss": 0.2614,
+      "step": 19390
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3487817574261584e-05,
+      "loss": 0.4046,
+      "step": 19391
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3487481738955216e-05,
+      "loss": 0.3098,
+      "step": 19392
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3487145903648853e-05,
+      "loss": 0.4718,
+      "step": 19393
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3486810068342486e-05,
+      "loss": 0.3208,
+      "step": 19394
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.348647423303612e-05,
+      "loss": 0.5083,
+      "step": 19395
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3486138397729755e-05,
+      "loss": 0.4932,
+      "step": 19396
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.348580256242339e-05,
+      "loss": 0.272,
+      "step": 19397
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3485466727117023e-05,
+      "loss": 0.2044,
+      "step": 19398
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3485130891810659e-05,
+      "loss": 0.2493,
+      "step": 19399
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3484795056504292e-05,
+      "loss": 0.3201,
+      "step": 19400
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3484459221197925e-05,
+      "loss": 0.2797,
+      "step": 19401
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.348412338589156e-05,
+      "loss": 0.1689,
+      "step": 19402
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3483787550585194e-05,
+      "loss": 0.2741,
+      "step": 19403
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3483451715278827e-05,
+      "loss": 0.2206,
+      "step": 19404
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3483115879972462e-05,
+      "loss": 0.2579,
+      "step": 19405
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3482780044666096e-05,
+      "loss": 0.262,
+      "step": 19406
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3482444209359731e-05,
+      "loss": 0.3056,
+      "step": 19407
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3482108374053364e-05,
+      "loss": 0.3699,
+      "step": 19408
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3481772538747e-05,
+      "loss": 0.1603,
+      "step": 19409
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3481436703440633e-05,
+      "loss": 0.4939,
+      "step": 19410
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3481100868134266e-05,
+      "loss": 0.5071,
+      "step": 19411
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3480765032827903e-05,
+      "loss": 0.1779,
+      "step": 19412
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3480429197521536e-05,
+      "loss": 0.2843,
+      "step": 19413
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.348009336221517e-05,
+      "loss": 0.138,
+      "step": 19414
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3479757526908805e-05,
+      "loss": 0.2839,
+      "step": 19415
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.347942169160244e-05,
+      "loss": 0.4385,
+      "step": 19416
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3479085856296073e-05,
+      "loss": 0.2791,
+      "step": 19417
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3478750020989709e-05,
+      "loss": 0.3414,
+      "step": 19418
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3478414185683342e-05,
+      "loss": 0.4197,
+      "step": 19419
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3478078350376975e-05,
+      "loss": 0.4558,
+      "step": 19420
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3477742515070611e-05,
+      "loss": 0.2512,
+      "step": 19421
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3477406679764244e-05,
+      "loss": 0.2898,
+      "step": 19422
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3477070844457879e-05,
+      "loss": 0.3551,
+      "step": 19423
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3476735009151513e-05,
+      "loss": 0.383,
+      "step": 19424
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3476399173845148e-05,
+      "loss": 0.2729,
+      "step": 19425
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3476063338538781e-05,
+      "loss": 0.2478,
+      "step": 19426
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3475727503232416e-05,
+      "loss": 0.3768,
+      "step": 19427
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.347539166792605e-05,
+      "loss": 0.2869,
+      "step": 19428
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3475055832619683e-05,
+      "loss": 0.3596,
+      "step": 19429
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.347471999731332e-05,
+      "loss": 0.2904,
+      "step": 19430
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3474384162006952e-05,
+      "loss": 0.2592,
+      "step": 19431
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3474048326700585e-05,
+      "loss": 0.3327,
+      "step": 19432
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3473712491394222e-05,
+      "loss": 0.2147,
+      "step": 19433
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3473376656087855e-05,
+      "loss": 0.4362,
+      "step": 19434
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.347304082078149e-05,
+      "loss": 0.2274,
+      "step": 19435
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3472704985475124e-05,
+      "loss": 0.3881,
+      "step": 19436
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3472369150168759e-05,
+      "loss": 0.2256,
+      "step": 19437
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3472033314862392e-05,
+      "loss": 0.3707,
+      "step": 19438
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3471697479556028e-05,
+      "loss": 0.4072,
+      "step": 19439
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3471361644249661e-05,
+      "loss": 0.2991,
+      "step": 19440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3471025808943294e-05,
+      "loss": 0.1633,
+      "step": 19441
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.347068997363693e-05,
+      "loss": 0.4584,
+      "step": 19442
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3470354138330563e-05,
+      "loss": 0.3157,
+      "step": 19443
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3470018303024198e-05,
+      "loss": 0.3333,
+      "step": 19444
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3469682467717832e-05,
+      "loss": 0.1911,
+      "step": 19445
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3469346632411467e-05,
+      "loss": 0.3567,
+      "step": 19446
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.34690107971051e-05,
+      "loss": 0.3721,
+      "step": 19447
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3468674961798736e-05,
+      "loss": 0.429,
+      "step": 19448
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.346833912649237e-05,
+      "loss": 0.2136,
+      "step": 19449
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3468003291186002e-05,
+      "loss": 0.3239,
+      "step": 19450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3467667455879639e-05,
+      "loss": 0.1662,
+      "step": 19451
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3467331620573272e-05,
+      "loss": 0.2346,
+      "step": 19452
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3466995785266905e-05,
+      "loss": 0.229,
+      "step": 19453
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3466659949960541e-05,
+      "loss": 0.3631,
+      "step": 19454
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3466324114654174e-05,
+      "loss": 0.3295,
+      "step": 19455
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3465988279347809e-05,
+      "loss": 0.4217,
+      "step": 19456
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3465652444041443e-05,
+      "loss": 0.2486,
+      "step": 19457
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3465316608735078e-05,
+      "loss": 0.3413,
+      "step": 19458
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.346498077342871e-05,
+      "loss": 0.2242,
+      "step": 19459
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3464644938122347e-05,
+      "loss": 0.4891,
+      "step": 19460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.346430910281598e-05,
+      "loss": 0.2526,
+      "step": 19461
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3463973267509613e-05,
+      "loss": 0.3364,
+      "step": 19462
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.346363743220325e-05,
+      "loss": 0.3244,
+      "step": 19463
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3463301596896882e-05,
+      "loss": 0.518,
+      "step": 19464
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3462965761590517e-05,
+      "loss": 0.2895,
+      "step": 19465
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3462629926284152e-05,
+      "loss": 0.4367,
+      "step": 19466
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3462294090977786e-05,
+      "loss": 0.4017,
+      "step": 19467
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.346195825567142e-05,
+      "loss": 0.3153,
+      "step": 19468
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3461622420365056e-05,
+      "loss": 0.382,
+      "step": 19469
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3461286585058688e-05,
+      "loss": 0.3124,
+      "step": 19470
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3460950749752321e-05,
+      "loss": 0.2533,
+      "step": 19471
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3460614914445958e-05,
+      "loss": 0.295,
+      "step": 19472
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.346027907913959e-05,
+      "loss": 0.3529,
+      "step": 19473
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3459943243833224e-05,
+      "loss": 0.4715,
+      "step": 19474
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.345960740852686e-05,
+      "loss": 0.279,
+      "step": 19475
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3459271573220493e-05,
+      "loss": 0.3712,
+      "step": 19476
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3458935737914128e-05,
+      "loss": 0.3865,
+      "step": 19477
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3458599902607762e-05,
+      "loss": 0.2616,
+      "step": 19478
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3458264067301397e-05,
+      "loss": 0.4104,
+      "step": 19479
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.345792823199503e-05,
+      "loss": 0.2706,
+      "step": 19480
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3457592396688666e-05,
+      "loss": 0.3913,
+      "step": 19481
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.34572565613823e-05,
+      "loss": 0.4154,
+      "step": 19482
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3456920726075932e-05,
+      "loss": 0.3346,
+      "step": 19483
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3456584890769568e-05,
+      "loss": 0.3847,
+      "step": 19484
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3456249055463201e-05,
+      "loss": 0.4134,
+      "step": 19485
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3455913220156836e-05,
+      "loss": 0.1648,
+      "step": 19486
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.345557738485047e-05,
+      "loss": 0.2912,
+      "step": 19487
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3455241549544105e-05,
+      "loss": 0.2479,
+      "step": 19488
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3454905714237738e-05,
+      "loss": 0.348,
+      "step": 19489
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3454569878931375e-05,
+      "loss": 0.0917,
+      "step": 19490
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3454234043625008e-05,
+      "loss": 0.2131,
+      "step": 19491
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.345389820831864e-05,
+      "loss": 0.3779,
+      "step": 19492
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3453562373012277e-05,
+      "loss": 0.2756,
+      "step": 19493
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.345322653770591e-05,
+      "loss": 0.3321,
+      "step": 19494
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3452890702399543e-05,
+      "loss": 0.296,
+      "step": 19495
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.345255486709318e-05,
+      "loss": 0.3692,
+      "step": 19496
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3452219031786812e-05,
+      "loss": 0.4594,
+      "step": 19497
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3451883196480447e-05,
+      "loss": 0.3141,
+      "step": 19498
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3451547361174081e-05,
+      "loss": 0.3605,
+      "step": 19499
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3451211525867716e-05,
+      "loss": 0.3508,
+      "step": 19500
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3450875690561349e-05,
+      "loss": 0.2816,
+      "step": 19501
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3450539855254985e-05,
+      "loss": 0.34,
+      "step": 19502
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3450204019948618e-05,
+      "loss": 0.2908,
+      "step": 19503
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3449868184642251e-05,
+      "loss": 0.2712,
+      "step": 19504
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3449532349335888e-05,
+      "loss": 0.2721,
+      "step": 19505
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.344919651402952e-05,
+      "loss": 0.4796,
+      "step": 19506
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3448860678723155e-05,
+      "loss": 0.3388,
+      "step": 19507
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.344852484341679e-05,
+      "loss": 0.5228,
+      "step": 19508
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3448189008110424e-05,
+      "loss": 0.2783,
+      "step": 19509
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3447853172804057e-05,
+      "loss": 0.244,
+      "step": 19510
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3447517337497694e-05,
+      "loss": 0.1776,
+      "step": 19511
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3447181502191327e-05,
+      "loss": 0.2786,
+      "step": 19512
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.344684566688496e-05,
+      "loss": 0.2274,
+      "step": 19513
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3446509831578596e-05,
+      "loss": 0.2691,
+      "step": 19514
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3446173996272229e-05,
+      "loss": 0.3759,
+      "step": 19515
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3445838160965864e-05,
+      "loss": 0.3405,
+      "step": 19516
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3445502325659498e-05,
+      "loss": 0.3833,
+      "step": 19517
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3445166490353131e-05,
+      "loss": 0.3461,
+      "step": 19518
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3444830655046766e-05,
+      "loss": 0.2032,
+      "step": 19519
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.34444948197404e-05,
+      "loss": 0.4627,
+      "step": 19520
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3444158984434035e-05,
+      "loss": 0.2576,
+      "step": 19521
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3443823149127668e-05,
+      "loss": 0.5229,
+      "step": 19522
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3443487313821304e-05,
+      "loss": 0.1733,
+      "step": 19523
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3443151478514937e-05,
+      "loss": 0.2832,
+      "step": 19524
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.344281564320857e-05,
+      "loss": 0.3018,
+      "step": 19525
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3442479807902207e-05,
+      "loss": 0.2097,
+      "step": 19526
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.344214397259584e-05,
+      "loss": 0.3804,
+      "step": 19527
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3441808137289474e-05,
+      "loss": 0.3604,
+      "step": 19528
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3441472301983109e-05,
+      "loss": 0.4308,
+      "step": 19529
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3441136466676744e-05,
+      "loss": 0.3189,
+      "step": 19530
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3440800631370377e-05,
+      "loss": 0.3694,
+      "step": 19531
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3440464796064013e-05,
+      "loss": 0.3119,
+      "step": 19532
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3440128960757646e-05,
+      "loss": 0.2613,
+      "step": 19533
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3439793125451279e-05,
+      "loss": 0.3575,
+      "step": 19534
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3439457290144915e-05,
+      "loss": 0.3483,
+      "step": 19535
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3439121454838548e-05,
+      "loss": 0.3854,
+      "step": 19536
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3438785619532183e-05,
+      "loss": 0.354,
+      "step": 19537
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3438449784225817e-05,
+      "loss": 0.2467,
+      "step": 19538
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.343811394891945e-05,
+      "loss": 0.3485,
+      "step": 19539
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3437778113613085e-05,
+      "loss": 0.2147,
+      "step": 19540
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.343744227830672e-05,
+      "loss": 0.3953,
+      "step": 19541
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3437106443000354e-05,
+      "loss": 0.5051,
+      "step": 19542
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3436770607693987e-05,
+      "loss": 0.2656,
+      "step": 19543
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3436434772387624e-05,
+      "loss": 0.2991,
+      "step": 19544
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3436098937081257e-05,
+      "loss": 0.3752,
+      "step": 19545
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.343576310177489e-05,
+      "loss": 0.3834,
+      "step": 19546
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3435427266468526e-05,
+      "loss": 0.1543,
+      "step": 19547
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3435091431162159e-05,
+      "loss": 0.3054,
+      "step": 19548
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3434755595855793e-05,
+      "loss": 0.2552,
+      "step": 19549
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3434419760549428e-05,
+      "loss": 0.3026,
+      "step": 19550
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3434083925243063e-05,
+      "loss": 0.2579,
+      "step": 19551
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3433748089936696e-05,
+      "loss": 0.3897,
+      "step": 19552
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.3433412254630332e-05,
+      "loss": 0.4143,
+      "step": 19553
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3433076419323965e-05,
+      "loss": 0.1989,
+      "step": 19554
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3432740584017598e-05,
+      "loss": 0.3004,
+      "step": 19555
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3432404748711234e-05,
+      "loss": 0.3954,
+      "step": 19556
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3432068913404867e-05,
+      "loss": 0.258,
+      "step": 19557
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3431733078098502e-05,
+      "loss": 0.3844,
+      "step": 19558
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3431397242792137e-05,
+      "loss": 0.2268,
+      "step": 19559
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3431061407485771e-05,
+      "loss": 0.4044,
+      "step": 19560
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3430725572179404e-05,
+      "loss": 0.4266,
+      "step": 19561
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3430389736873039e-05,
+      "loss": 0.3355,
+      "step": 19562
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3430053901566673e-05,
+      "loss": 0.4247,
+      "step": 19563
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3429718066260306e-05,
+      "loss": 0.4694,
+      "step": 19564
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3429382230953943e-05,
+      "loss": 0.1674,
+      "step": 19565
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3429046395647576e-05,
+      "loss": 0.378,
+      "step": 19566
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3428710560341209e-05,
+      "loss": 0.2252,
+      "step": 19567
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3428374725034845e-05,
+      "loss": 0.3412,
+      "step": 19568
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3428038889728478e-05,
+      "loss": 0.2218,
+      "step": 19569
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3427703054422113e-05,
+      "loss": 0.354,
+      "step": 19570
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3427367219115747e-05,
+      "loss": 0.3248,
+      "step": 19571
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3427031383809382e-05,
+      "loss": 0.4789,
+      "step": 19572
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3426695548503015e-05,
+      "loss": 0.3368,
+      "step": 19573
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3426359713196651e-05,
+      "loss": 0.4001,
+      "step": 19574
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3426023877890284e-05,
+      "loss": 0.4089,
+      "step": 19575
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3425688042583917e-05,
+      "loss": 0.2369,
+      "step": 19576
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3425352207277553e-05,
+      "loss": 0.1759,
+      "step": 19577
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3425016371971186e-05,
+      "loss": 0.1896,
+      "step": 19578
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3424680536664821e-05,
+      "loss": 0.3691,
+      "step": 19579
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3424344701358456e-05,
+      "loss": 0.3225,
+      "step": 19580
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.342400886605209e-05,
+      "loss": 0.2965,
+      "step": 19581
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3423673030745723e-05,
+      "loss": 0.347,
+      "step": 19582
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3423337195439356e-05,
+      "loss": 0.1839,
+      "step": 19583
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3423001360132993e-05,
+      "loss": 0.225,
+      "step": 19584
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3422665524826626e-05,
+      "loss": 0.292,
+      "step": 19585
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3422329689520258e-05,
+      "loss": 0.2223,
+      "step": 19586
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3421993854213895e-05,
+      "loss": 0.4623,
+      "step": 19587
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3421658018907528e-05,
+      "loss": 0.3033,
+      "step": 19588
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3421322183601162e-05,
+      "loss": 0.386,
+      "step": 19589
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3420986348294797e-05,
+      "loss": 0.3506,
+      "step": 19590
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3420650512988432e-05,
+      "loss": 0.2704,
+      "step": 19591
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3420314677682065e-05,
+      "loss": 0.1875,
+      "step": 19592
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3419978842375701e-05,
+      "loss": 0.3175,
+      "step": 19593
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3419643007069334e-05,
+      "loss": 0.3539,
+      "step": 19594
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3419307171762967e-05,
+      "loss": 0.4581,
+      "step": 19595
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3418971336456603e-05,
+      "loss": 0.4297,
+      "step": 19596
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3418635501150236e-05,
+      "loss": 0.3883,
+      "step": 19597
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.341829966584387e-05,
+      "loss": 0.3829,
+      "step": 19598
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3417963830537505e-05,
+      "loss": 0.2293,
+      "step": 19599
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.341762799523114e-05,
+      "loss": 0.1787,
+      "step": 19600
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3417292159924773e-05,
+      "loss": 0.1792,
+      "step": 19601
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.341695632461841e-05,
+      "loss": 0.4376,
+      "step": 19602
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3416620489312042e-05,
+      "loss": 0.4933,
+      "step": 19603
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3416284654005675e-05,
+      "loss": 0.3555,
+      "step": 19604
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3415948818699312e-05,
+      "loss": 0.3784,
+      "step": 19605
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3415612983392945e-05,
+      "loss": 0.324,
+      "step": 19606
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.341527714808658e-05,
+      "loss": 0.3626,
+      "step": 19607
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3414941312780214e-05,
+      "loss": 0.2343,
+      "step": 19608
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3414605477473847e-05,
+      "loss": 0.4406,
+      "step": 19609
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3414269642167482e-05,
+      "loss": 0.1777,
+      "step": 19610
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3413933806861116e-05,
+      "loss": 0.3108,
+      "step": 19611
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.341359797155475e-05,
+      "loss": 0.3314,
+      "step": 19612
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3413262136248384e-05,
+      "loss": 0.1689,
+      "step": 19613
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.341292630094202e-05,
+      "loss": 0.2353,
+      "step": 19614
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3412590465635653e-05,
+      "loss": 0.4563,
+      "step": 19615
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3412254630329286e-05,
+      "loss": 0.4465,
+      "step": 19616
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3411918795022922e-05,
+      "loss": 0.4255,
+      "step": 19617
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3411582959716555e-05,
+      "loss": 0.3561,
+      "step": 19618
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.341124712441019e-05,
+      "loss": 0.2427,
+      "step": 19619
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3410911289103825e-05,
+      "loss": 0.2757,
+      "step": 19620
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.341057545379746e-05,
+      "loss": 0.2825,
+      "step": 19621
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3410239618491092e-05,
+      "loss": 0.3466,
+      "step": 19622
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3409903783184729e-05,
+      "loss": 0.2845,
+      "step": 19623
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3409567947878362e-05,
+      "loss": 0.431,
+      "step": 19624
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3409232112571994e-05,
+      "loss": 0.4734,
+      "step": 19625
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.340889627726563e-05,
+      "loss": 0.3822,
+      "step": 19626
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3408560441959264e-05,
+      "loss": 0.3011,
+      "step": 19627
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3408224606652898e-05,
+      "loss": 0.3058,
+      "step": 19628
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3407888771346533e-05,
+      "loss": 0.2495,
+      "step": 19629
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3407552936040166e-05,
+      "loss": 0.3024,
+      "step": 19630
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.34072171007338e-05,
+      "loss": 0.4301,
+      "step": 19631
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3406881265427435e-05,
+      "loss": 0.2366,
+      "step": 19632
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.340654543012107e-05,
+      "loss": 0.1867,
+      "step": 19633
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3406209594814703e-05,
+      "loss": 0.4048,
+      "step": 19634
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.340587375950834e-05,
+      "loss": 0.4538,
+      "step": 19635
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3405537924201972e-05,
+      "loss": 0.2652,
+      "step": 19636
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3405202088895605e-05,
+      "loss": 0.1748,
+      "step": 19637
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3404866253589241e-05,
+      "loss": 0.2609,
+      "step": 19638
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3404530418282874e-05,
+      "loss": 0.2753,
+      "step": 19639
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3404194582976509e-05,
+      "loss": 0.275,
+      "step": 19640
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3403858747670144e-05,
+      "loss": 0.3956,
+      "step": 19641
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3403522912363778e-05,
+      "loss": 0.3303,
+      "step": 19642
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3403187077057411e-05,
+      "loss": 0.1642,
+      "step": 19643
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3402851241751048e-05,
+      "loss": 0.2027,
+      "step": 19644
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.340251540644468e-05,
+      "loss": 0.3522,
+      "step": 19645
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3402179571138314e-05,
+      "loss": 0.3305,
+      "step": 19646
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.340184373583195e-05,
+      "loss": 0.2074,
+      "step": 19647
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3401507900525583e-05,
+      "loss": 0.3332,
+      "step": 19648
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3401172065219218e-05,
+      "loss": 0.6448,
+      "step": 19649
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3400836229912852e-05,
+      "loss": 0.4397,
+      "step": 19650
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3400500394606487e-05,
+      "loss": 0.2773,
+      "step": 19651
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.340016455930012e-05,
+      "loss": 0.2639,
+      "step": 19652
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3399828723993754e-05,
+      "loss": 0.3325,
+      "step": 19653
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3399492888687389e-05,
+      "loss": 0.2405,
+      "step": 19654
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3399157053381022e-05,
+      "loss": 0.244,
+      "step": 19655
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3398821218074658e-05,
+      "loss": 0.3416,
+      "step": 19656
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3398485382768291e-05,
+      "loss": 0.2152,
+      "step": 19657
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3398149547461924e-05,
+      "loss": 0.3792,
+      "step": 19658
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.339781371215556e-05,
+      "loss": 0.177,
+      "step": 19659
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3397477876849194e-05,
+      "loss": 0.4142,
+      "step": 19660
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3397142041542828e-05,
+      "loss": 0.2542,
+      "step": 19661
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3396806206236463e-05,
+      "loss": 0.3706,
+      "step": 19662
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3396470370930098e-05,
+      "loss": 0.2781,
+      "step": 19663
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.339613453562373e-05,
+      "loss": 0.5732,
+      "step": 19664
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3395798700317367e-05,
+      "loss": 0.4271,
+      "step": 19665
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3395462865011e-05,
+      "loss": 0.3518,
+      "step": 19666
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3395127029704633e-05,
+      "loss": 0.3316,
+      "step": 19667
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3394791194398269e-05,
+      "loss": 0.3402,
+      "step": 19668
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3394455359091902e-05,
+      "loss": 0.2725,
+      "step": 19669
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3394119523785537e-05,
+      "loss": 0.197,
+      "step": 19670
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3393783688479171e-05,
+      "loss": 0.3577,
+      "step": 19671
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3393447853172806e-05,
+      "loss": 0.3461,
+      "step": 19672
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3393112017866439e-05,
+      "loss": 0.4301,
+      "step": 19673
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3392776182560074e-05,
+      "loss": 0.3049,
+      "step": 19674
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3392440347253708e-05,
+      "loss": 0.2972,
+      "step": 19675
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3392104511947341e-05,
+      "loss": 0.2513,
+      "step": 19676
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3391768676640977e-05,
+      "loss": 0.4635,
+      "step": 19677
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.339143284133461e-05,
+      "loss": 0.2802,
+      "step": 19678
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3391097006028243e-05,
+      "loss": 0.2392,
+      "step": 19679
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.339076117072188e-05,
+      "loss": 0.3036,
+      "step": 19680
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3390425335415513e-05,
+      "loss": 0.3252,
+      "step": 19681
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3390089500109147e-05,
+      "loss": 0.5584,
+      "step": 19682
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3389753664802782e-05,
+      "loss": 0.5229,
+      "step": 19683
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3389417829496417e-05,
+      "loss": 0.4358,
+      "step": 19684
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.338908199419005e-05,
+      "loss": 0.4609,
+      "step": 19685
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3388746158883686e-05,
+      "loss": 0.3435,
+      "step": 19686
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3388410323577319e-05,
+      "loss": 0.2972,
+      "step": 19687
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3388074488270952e-05,
+      "loss": 0.2184,
+      "step": 19688
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3387738652964588e-05,
+      "loss": 0.2346,
+      "step": 19689
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3387402817658221e-05,
+      "loss": 0.1803,
+      "step": 19690
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3387066982351856e-05,
+      "loss": 0.365,
+      "step": 19691
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.338673114704549e-05,
+      "loss": 0.4672,
+      "step": 19692
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3386395311739125e-05,
+      "loss": 0.3509,
+      "step": 19693
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3386059476432758e-05,
+      "loss": 0.1956,
+      "step": 19694
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3385723641126394e-05,
+      "loss": 0.325,
+      "step": 19695
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3385387805820027e-05,
+      "loss": 0.3304,
+      "step": 19696
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.338505197051366e-05,
+      "loss": 0.3946,
+      "step": 19697
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3384716135207297e-05,
+      "loss": 0.2205,
+      "step": 19698
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.338438029990093e-05,
+      "loss": 0.3631,
+      "step": 19699
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3384044464594563e-05,
+      "loss": 0.3494,
+      "step": 19700
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3383708629288199e-05,
+      "loss": 0.2872,
+      "step": 19701
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3383372793981832e-05,
+      "loss": 0.3452,
+      "step": 19702
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3383036958675466e-05,
+      "loss": 0.3145,
+      "step": 19703
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3382701123369101e-05,
+      "loss": 0.2788,
+      "step": 19704
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3382365288062736e-05,
+      "loss": 0.3672,
+      "step": 19705
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3382029452756369e-05,
+      "loss": 0.2866,
+      "step": 19706
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3381693617450005e-05,
+      "loss": 0.431,
+      "step": 19707
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3381357782143638e-05,
+      "loss": 0.2188,
+      "step": 19708
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3381021946837271e-05,
+      "loss": 0.569,
+      "step": 19709
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3380686111530907e-05,
+      "loss": 0.3303,
+      "step": 19710
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.338035027622454e-05,
+      "loss": 0.2442,
+      "step": 19711
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3380014440918175e-05,
+      "loss": 0.3417,
+      "step": 19712
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.337967860561181e-05,
+      "loss": 0.2731,
+      "step": 19713
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3379342770305444e-05,
+      "loss": 0.3769,
+      "step": 19714
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3379006934999077e-05,
+      "loss": 0.1604,
+      "step": 19715
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3378671099692713e-05,
+      "loss": 0.2588,
+      "step": 19716
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3378335264386346e-05,
+      "loss": 0.5137,
+      "step": 19717
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.337799942907998e-05,
+      "loss": 0.3241,
+      "step": 19718
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3377663593773616e-05,
+      "loss": 0.2214,
+      "step": 19719
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3377327758467249e-05,
+      "loss": 0.3658,
+      "step": 19720
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3376991923160882e-05,
+      "loss": 0.309,
+      "step": 19721
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3376656087854518e-05,
+      "loss": 0.3914,
+      "step": 19722
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3376320252548151e-05,
+      "loss": 0.4085,
+      "step": 19723
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3375984417241786e-05,
+      "loss": 0.3081,
+      "step": 19724
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.337564858193542e-05,
+      "loss": 0.4208,
+      "step": 19725
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3375312746629055e-05,
+      "loss": 0.2898,
+      "step": 19726
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3374976911322688e-05,
+      "loss": 0.3653,
+      "step": 19727
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3374641076016324e-05,
+      "loss": 0.4457,
+      "step": 19728
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3374305240709957e-05,
+      "loss": 0.2115,
+      "step": 19729
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.337396940540359e-05,
+      "loss": 0.2386,
+      "step": 19730
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3373633570097226e-05,
+      "loss": 0.3605,
+      "step": 19731
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.337329773479086e-05,
+      "loss": 0.2476,
+      "step": 19732
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3372961899484494e-05,
+      "loss": 0.2195,
+      "step": 19733
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3372626064178129e-05,
+      "loss": 0.4327,
+      "step": 19734
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3372290228871763e-05,
+      "loss": 0.2833,
+      "step": 19735
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3371954393565396e-05,
+      "loss": 0.3793,
+      "step": 19736
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3371618558259033e-05,
+      "loss": 0.2941,
+      "step": 19737
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3371282722952666e-05,
+      "loss": 0.2663,
+      "step": 19738
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3370946887646299e-05,
+      "loss": 0.4188,
+      "step": 19739
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3370611052339935e-05,
+      "loss": 0.5095,
+      "step": 19740
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3370275217033568e-05,
+      "loss": 0.4165,
+      "step": 19741
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3369939381727202e-05,
+      "loss": 0.245,
+      "step": 19742
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3369603546420837e-05,
+      "loss": 0.4224,
+      "step": 19743
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.336926771111447e-05,
+      "loss": 0.3341,
+      "step": 19744
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3368931875808105e-05,
+      "loss": 0.3184,
+      "step": 19745
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.336859604050174e-05,
+      "loss": 0.2069,
+      "step": 19746
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3368260205195374e-05,
+      "loss": 0.439,
+      "step": 19747
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3367924369889007e-05,
+      "loss": 0.4005,
+      "step": 19748
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3367588534582643e-05,
+      "loss": 0.2596,
+      "step": 19749
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.3367252699276276e-05,
+      "loss": 0.3398,
+      "step": 19750
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.336691686396991e-05,
+      "loss": 0.2938,
+      "step": 19751
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3366581028663546e-05,
+      "loss": 0.1899,
+      "step": 19752
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3366245193357179e-05,
+      "loss": 0.2728,
+      "step": 19753
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3365909358050813e-05,
+      "loss": 0.2288,
+      "step": 19754
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3365573522744448e-05,
+      "loss": 0.3099,
+      "step": 19755
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3365237687438082e-05,
+      "loss": 0.2314,
+      "step": 19756
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3364901852131715e-05,
+      "loss": 0.2917,
+      "step": 19757
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3364566016825352e-05,
+      "loss": 0.6171,
+      "step": 19758
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3364230181518985e-05,
+      "loss": 0.2327,
+      "step": 19759
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3363894346212618e-05,
+      "loss": 0.2864,
+      "step": 19760
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3363558510906252e-05,
+      "loss": 0.2365,
+      "step": 19761
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3363222675599887e-05,
+      "loss": 0.3973,
+      "step": 19762
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3362886840293522e-05,
+      "loss": 0.3331,
+      "step": 19763
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3362551004987155e-05,
+      "loss": 0.2306,
+      "step": 19764
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.336221516968079e-05,
+      "loss": 0.2407,
+      "step": 19765
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3361879334374424e-05,
+      "loss": 0.3063,
+      "step": 19766
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3361543499068057e-05,
+      "loss": 0.4452,
+      "step": 19767
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3361207663761693e-05,
+      "loss": 0.2501,
+      "step": 19768
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3360871828455326e-05,
+      "loss": 0.3706,
+      "step": 19769
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3360535993148959e-05,
+      "loss": 0.3058,
+      "step": 19770
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3360200157842595e-05,
+      "loss": 0.3698,
+      "step": 19771
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3359864322536228e-05,
+      "loss": 0.2773,
+      "step": 19772
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3359528487229863e-05,
+      "loss": 0.2305,
+      "step": 19773
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3359192651923498e-05,
+      "loss": 0.39,
+      "step": 19774
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3358856816617132e-05,
+      "loss": 0.2989,
+      "step": 19775
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3358520981310765e-05,
+      "loss": 0.2966,
+      "step": 19776
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3358185146004402e-05,
+      "loss": 0.3746,
+      "step": 19777
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3357849310698035e-05,
+      "loss": 0.3385,
+      "step": 19778
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3357513475391667e-05,
+      "loss": 0.1851,
+      "step": 19779
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3357177640085304e-05,
+      "loss": 0.2908,
+      "step": 19780
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3356841804778937e-05,
+      "loss": 0.5218,
+      "step": 19781
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3356505969472571e-05,
+      "loss": 0.2887,
+      "step": 19782
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3356170134166206e-05,
+      "loss": 0.2656,
+      "step": 19783
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.335583429885984e-05,
+      "loss": 0.2094,
+      "step": 19784
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3355498463553474e-05,
+      "loss": 0.4167,
+      "step": 19785
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.335516262824711e-05,
+      "loss": 0.2812,
+      "step": 19786
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3354826792940743e-05,
+      "loss": 0.455,
+      "step": 19787
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3354490957634376e-05,
+      "loss": 0.3116,
+      "step": 19788
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3354155122328012e-05,
+      "loss": 0.2627,
+      "step": 19789
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3353819287021645e-05,
+      "loss": 0.3074,
+      "step": 19790
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3353483451715278e-05,
+      "loss": 0.3496,
+      "step": 19791
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3353147616408915e-05,
+      "loss": 0.2099,
+      "step": 19792
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3352811781102547e-05,
+      "loss": 0.2264,
+      "step": 19793
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3352475945796182e-05,
+      "loss": 0.3059,
+      "step": 19794
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3352140110489817e-05,
+      "loss": 0.2175,
+      "step": 19795
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3351804275183451e-05,
+      "loss": 0.3445,
+      "step": 19796
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3351468439877084e-05,
+      "loss": 0.3902,
+      "step": 19797
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.335113260457072e-05,
+      "loss": 0.2612,
+      "step": 19798
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3350796769264354e-05,
+      "loss": 0.4819,
+      "step": 19799
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3350460933957987e-05,
+      "loss": 0.4175,
+      "step": 19800
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3350125098651623e-05,
+      "loss": 0.2335,
+      "step": 19801
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3349789263345256e-05,
+      "loss": 0.2347,
+      "step": 19802
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.334945342803889e-05,
+      "loss": 0.3832,
+      "step": 19803
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3349117592732525e-05,
+      "loss": 0.2092,
+      "step": 19804
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.334878175742616e-05,
+      "loss": 0.3479,
+      "step": 19805
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3348445922119793e-05,
+      "loss": 0.223,
+      "step": 19806
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3348110086813429e-05,
+      "loss": 0.2859,
+      "step": 19807
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3347774251507062e-05,
+      "loss": 0.4523,
+      "step": 19808
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3347438416200695e-05,
+      "loss": 0.2304,
+      "step": 19809
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3347102580894331e-05,
+      "loss": 0.2906,
+      "step": 19810
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3346766745587964e-05,
+      "loss": 0.2499,
+      "step": 19811
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3346430910281597e-05,
+      "loss": 0.2369,
+      "step": 19812
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3346095074975234e-05,
+      "loss": 0.3117,
+      "step": 19813
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3345759239668867e-05,
+      "loss": 0.4823,
+      "step": 19814
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3345423404362501e-05,
+      "loss": 0.419,
+      "step": 19815
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3345087569056136e-05,
+      "loss": 0.2751,
+      "step": 19816
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.334475173374977e-05,
+      "loss": 0.1282,
+      "step": 19817
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3344415898443403e-05,
+      "loss": 0.3333,
+      "step": 19818
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.334408006313704e-05,
+      "loss": 0.2206,
+      "step": 19819
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3343744227830673e-05,
+      "loss": 0.2366,
+      "step": 19820
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3343408392524306e-05,
+      "loss": 0.2063,
+      "step": 19821
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3343072557217942e-05,
+      "loss": 0.4063,
+      "step": 19822
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3342736721911575e-05,
+      "loss": 0.336,
+      "step": 19823
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.334240088660521e-05,
+      "loss": 0.4573,
+      "step": 19824
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3342065051298844e-05,
+      "loss": 0.1862,
+      "step": 19825
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3341729215992479e-05,
+      "loss": 0.2154,
+      "step": 19826
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3341393380686112e-05,
+      "loss": 0.3031,
+      "step": 19827
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3341057545379748e-05,
+      "loss": 0.7176,
+      "step": 19828
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3340721710073381e-05,
+      "loss": 0.2363,
+      "step": 19829
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3340385874767014e-05,
+      "loss": 0.1988,
+      "step": 19830
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.334005003946065e-05,
+      "loss": 0.5198,
+      "step": 19831
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3339714204154283e-05,
+      "loss": 0.4383,
+      "step": 19832
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3339378368847916e-05,
+      "loss": 0.2019,
+      "step": 19833
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3339042533541553e-05,
+      "loss": 0.3292,
+      "step": 19834
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3338706698235186e-05,
+      "loss": 0.1975,
+      "step": 19835
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.333837086292882e-05,
+      "loss": 0.4543,
+      "step": 19836
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3338035027622455e-05,
+      "loss": 0.3106,
+      "step": 19837
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.333769919231609e-05,
+      "loss": 0.5497,
+      "step": 19838
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3337363357009723e-05,
+      "loss": 0.4937,
+      "step": 19839
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3337027521703359e-05,
+      "loss": 0.2952,
+      "step": 19840
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3336691686396992e-05,
+      "loss": 0.3509,
+      "step": 19841
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3336355851090625e-05,
+      "loss": 0.2897,
+      "step": 19842
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3336020015784261e-05,
+      "loss": 0.2753,
+      "step": 19843
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3335684180477894e-05,
+      "loss": 0.4705,
+      "step": 19844
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3335348345171529e-05,
+      "loss": 0.2165,
+      "step": 19845
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3335012509865163e-05,
+      "loss": 0.6213,
+      "step": 19846
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3334676674558798e-05,
+      "loss": 0.3186,
+      "step": 19847
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3334340839252431e-05,
+      "loss": 0.3207,
+      "step": 19848
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3334005003946067e-05,
+      "loss": 0.3535,
+      "step": 19849
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.33336691686397e-05,
+      "loss": 0.4922,
+      "step": 19850
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 0.2125,
+      "step": 19851
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.333299749802697e-05,
+      "loss": 0.2748,
+      "step": 19852
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3332661662720603e-05,
+      "loss": 0.48,
+      "step": 19853
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3332325827414237e-05,
+      "loss": 0.2662,
+      "step": 19854
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3331989992107872e-05,
+      "loss": 0.2128,
+      "step": 19855
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3331654156801505e-05,
+      "loss": 0.4705,
+      "step": 19856
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.333131832149514e-05,
+      "loss": 0.4023,
+      "step": 19857
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3330982486188774e-05,
+      "loss": 0.2945,
+      "step": 19858
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3330646650882409e-05,
+      "loss": 0.2991,
+      "step": 19859
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3330310815576042e-05,
+      "loss": 0.2499,
+      "step": 19860
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3329974980269678e-05,
+      "loss": 0.365,
+      "step": 19861
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3329639144963311e-05,
+      "loss": 0.2007,
+      "step": 19862
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3329303309656944e-05,
+      "loss": 0.2868,
+      "step": 19863
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.332896747435058e-05,
+      "loss": 0.3042,
+      "step": 19864
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3328631639044213e-05,
+      "loss": 0.3422,
+      "step": 19865
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3328295803737848e-05,
+      "loss": 0.3144,
+      "step": 19866
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3327959968431483e-05,
+      "loss": 0.362,
+      "step": 19867
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3327624133125117e-05,
+      "loss": 0.2481,
+      "step": 19868
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.332728829781875e-05,
+      "loss": 0.3005,
+      "step": 19869
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3326952462512387e-05,
+      "loss": 0.1605,
+      "step": 19870
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.332661662720602e-05,
+      "loss": 0.2277,
+      "step": 19871
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3326280791899652e-05,
+      "loss": 0.27,
+      "step": 19872
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3325944956593289e-05,
+      "loss": 0.3893,
+      "step": 19873
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3325609121286922e-05,
+      "loss": 0.2115,
+      "step": 19874
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3325273285980556e-05,
+      "loss": 0.3289,
+      "step": 19875
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3324937450674191e-05,
+      "loss": 0.404,
+      "step": 19876
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3324601615367826e-05,
+      "loss": 0.2769,
+      "step": 19877
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3324265780061459e-05,
+      "loss": 0.4004,
+      "step": 19878
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3323929944755093e-05,
+      "loss": 0.2856,
+      "step": 19879
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3323594109448728e-05,
+      "loss": 0.2608,
+      "step": 19880
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3323258274142361e-05,
+      "loss": 0.2377,
+      "step": 19881
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3322922438835997e-05,
+      "loss": 0.3178,
+      "step": 19882
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.332258660352963e-05,
+      "loss": 0.2762,
+      "step": 19883
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3322250768223263e-05,
+      "loss": 0.188,
+      "step": 19884
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.33219149329169e-05,
+      "loss": 0.2532,
+      "step": 19885
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3321579097610532e-05,
+      "loss": 0.3108,
+      "step": 19886
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3321243262304167e-05,
+      "loss": 0.285,
+      "step": 19887
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3320907426997802e-05,
+      "loss": 0.1637,
+      "step": 19888
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3320571591691436e-05,
+      "loss": 0.3033,
+      "step": 19889
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.332023575638507e-05,
+      "loss": 0.2199,
+      "step": 19890
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3319899921078706e-05,
+      "loss": 0.2622,
+      "step": 19891
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3319564085772339e-05,
+      "loss": 0.4013,
+      "step": 19892
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3319228250465972e-05,
+      "loss": 0.2505,
+      "step": 19893
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3318892415159608e-05,
+      "loss": 0.4585,
+      "step": 19894
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.331855657985324e-05,
+      "loss": 0.2368,
+      "step": 19895
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3318220744546875e-05,
+      "loss": 0.2633,
+      "step": 19896
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.331788490924051e-05,
+      "loss": 0.2738,
+      "step": 19897
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3317549073934145e-05,
+      "loss": 0.2001,
+      "step": 19898
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3317213238627778e-05,
+      "loss": 0.1718,
+      "step": 19899
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3316877403321412e-05,
+      "loss": 0.4354,
+      "step": 19900
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3316541568015047e-05,
+      "loss": 0.3141,
+      "step": 19901
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.331620573270868e-05,
+      "loss": 0.4323,
+      "step": 19902
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3315869897402316e-05,
+      "loss": 0.2989,
+      "step": 19903
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.331553406209595e-05,
+      "loss": 0.3077,
+      "step": 19904
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3315198226789582e-05,
+      "loss": 0.2936,
+      "step": 19905
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3314862391483219e-05,
+      "loss": 0.1914,
+      "step": 19906
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3314526556176852e-05,
+      "loss": 0.3362,
+      "step": 19907
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3314190720870486e-05,
+      "loss": 0.2534,
+      "step": 19908
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.331385488556412e-05,
+      "loss": 0.3495,
+      "step": 19909
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3313519050257755e-05,
+      "loss": 0.2964,
+      "step": 19910
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3313183214951388e-05,
+      "loss": 0.3415,
+      "step": 19911
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3312847379645025e-05,
+      "loss": 0.4148,
+      "step": 19912
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3312511544338658e-05,
+      "loss": 0.203,
+      "step": 19913
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.331217570903229e-05,
+      "loss": 0.2333,
+      "step": 19914
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3311839873725927e-05,
+      "loss": 0.2656,
+      "step": 19915
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.331150403841956e-05,
+      "loss": 0.2713,
+      "step": 19916
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3311168203113195e-05,
+      "loss": 0.3052,
+      "step": 19917
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.331083236780683e-05,
+      "loss": 0.2122,
+      "step": 19918
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3310496532500464e-05,
+      "loss": 0.3259,
+      "step": 19919
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3310160697194097e-05,
+      "loss": 0.2615,
+      "step": 19920
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3309824861887733e-05,
+      "loss": 0.5266,
+      "step": 19921
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3309489026581366e-05,
+      "loss": 0.2943,
+      "step": 19922
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3309153191274999e-05,
+      "loss": 0.1616,
+      "step": 19923
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3308817355968635e-05,
+      "loss": 0.2519,
+      "step": 19924
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3308481520662268e-05,
+      "loss": 0.3008,
+      "step": 19925
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3308145685355901e-05,
+      "loss": 0.2467,
+      "step": 19926
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3307809850049538e-05,
+      "loss": 0.3211,
+      "step": 19927
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.330747401474317e-05,
+      "loss": 0.3337,
+      "step": 19928
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3307138179436805e-05,
+      "loss": 0.2717,
+      "step": 19929
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.330680234413044e-05,
+      "loss": 0.1124,
+      "step": 19930
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3306466508824075e-05,
+      "loss": 0.2787,
+      "step": 19931
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3306130673517708e-05,
+      "loss": 0.2814,
+      "step": 19932
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3305794838211344e-05,
+      "loss": 0.3753,
+      "step": 19933
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3305459002904977e-05,
+      "loss": 0.4,
+      "step": 19934
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.330512316759861e-05,
+      "loss": 0.1762,
+      "step": 19935
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3304787332292246e-05,
+      "loss": 0.3536,
+      "step": 19936
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3304451496985879e-05,
+      "loss": 0.4889,
+      "step": 19937
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3304115661679514e-05,
+      "loss": 0.2087,
+      "step": 19938
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3303779826373148e-05,
+      "loss": 0.6042,
+      "step": 19939
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3303443991066783e-05,
+      "loss": 0.3376,
+      "step": 19940
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3303108155760416e-05,
+      "loss": 0.4602,
+      "step": 19941
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3302772320454049e-05,
+      "loss": 0.3482,
+      "step": 19942
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3302436485147685e-05,
+      "loss": 0.2989,
+      "step": 19943
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3302100649841318e-05,
+      "loss": 0.2635,
+      "step": 19944
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3301764814534953e-05,
+      "loss": 0.4093,
+      "step": 19945
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3301428979228588e-05,
+      "loss": 0.4617,
+      "step": 19946
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.330109314392222e-05,
+      "loss": 0.341,
+      "step": 19947
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3300757308615855e-05,
+      "loss": 0.3799,
+      "step": 19948
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.330042147330949e-05,
+      "loss": 0.3577,
+      "step": 19949
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.3300085638003124e-05,
+      "loss": 0.3721,
+      "step": 19950
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3299749802696757e-05,
+      "loss": 0.3519,
+      "step": 19951
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3299413967390394e-05,
+      "loss": 0.3457,
+      "step": 19952
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3299078132084027e-05,
+      "loss": 0.2461,
+      "step": 19953
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.329874229677766e-05,
+      "loss": 0.2302,
+      "step": 19954
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3298406461471296e-05,
+      "loss": 0.3694,
+      "step": 19955
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3298070626164929e-05,
+      "loss": 0.3884,
+      "step": 19956
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3297734790858564e-05,
+      "loss": 0.3564,
+      "step": 19957
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3297398955552198e-05,
+      "loss": 0.1694,
+      "step": 19958
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3297063120245833e-05,
+      "loss": 0.2849,
+      "step": 19959
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3296727284939466e-05,
+      "loss": 0.3908,
+      "step": 19960
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3296391449633102e-05,
+      "loss": 0.2167,
+      "step": 19961
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3296055614326735e-05,
+      "loss": 0.4365,
+      "step": 19962
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3295719779020368e-05,
+      "loss": 0.2545,
+      "step": 19963
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3295383943714004e-05,
+      "loss": 0.5332,
+      "step": 19964
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3295048108407637e-05,
+      "loss": 0.2105,
+      "step": 19965
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3294712273101272e-05,
+      "loss": 0.2372,
+      "step": 19966
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3294376437794907e-05,
+      "loss": 0.4309,
+      "step": 19967
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.329404060248854e-05,
+      "loss": 0.2086,
+      "step": 19968
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3293704767182174e-05,
+      "loss": 0.3029,
+      "step": 19969
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3293368931875809e-05,
+      "loss": 0.1581,
+      "step": 19970
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3293033096569444e-05,
+      "loss": 0.3181,
+      "step": 19971
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3292697261263077e-05,
+      "loss": 0.3824,
+      "step": 19972
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3292361425956713e-05,
+      "loss": 0.3306,
+      "step": 19973
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3292025590650346e-05,
+      "loss": 0.238,
+      "step": 19974
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3291689755343979e-05,
+      "loss": 0.2888,
+      "step": 19975
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3291353920037615e-05,
+      "loss": 0.2141,
+      "step": 19976
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3291018084731248e-05,
+      "loss": 0.282,
+      "step": 19977
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3290682249424883e-05,
+      "loss": 0.2424,
+      "step": 19978
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3290346414118517e-05,
+      "loss": 0.2192,
+      "step": 19979
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3290010578812152e-05,
+      "loss": 0.4411,
+      "step": 19980
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3289674743505785e-05,
+      "loss": 0.2225,
+      "step": 19981
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3289338908199421e-05,
+      "loss": 0.2202,
+      "step": 19982
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3289003072893054e-05,
+      "loss": 0.2722,
+      "step": 19983
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3288667237586687e-05,
+      "loss": 0.2509,
+      "step": 19984
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3288331402280324e-05,
+      "loss": 0.2435,
+      "step": 19985
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3287995566973956e-05,
+      "loss": 0.1573,
+      "step": 19986
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3287659731667591e-05,
+      "loss": 0.2828,
+      "step": 19987
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3287323896361226e-05,
+      "loss": 0.3653,
+      "step": 19988
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.328698806105486e-05,
+      "loss": 0.3117,
+      "step": 19989
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3286652225748493e-05,
+      "loss": 0.269,
+      "step": 19990
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3286316390442128e-05,
+      "loss": 0.2919,
+      "step": 19991
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3285980555135763e-05,
+      "loss": 0.3917,
+      "step": 19992
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3285644719829396e-05,
+      "loss": 0.2749,
+      "step": 19993
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3285308884523032e-05,
+      "loss": 0.2909,
+      "step": 19994
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3284973049216665e-05,
+      "loss": 0.3172,
+      "step": 19995
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3284637213910298e-05,
+      "loss": 0.1699,
+      "step": 19996
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3284301378603934e-05,
+      "loss": 0.3497,
+      "step": 19997
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3283965543297567e-05,
+      "loss": 0.1982,
+      "step": 19998
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3283629707991202e-05,
+      "loss": 0.2328,
+      "step": 19999
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3283293872684836e-05,
+      "loss": 0.2418,
+      "step": 20000
+    },
+    {
+      "epoch": 1.01,
+      "eval_accuracy": 0.8816403294176419,
+      "eval_f1_microaverage": 0.8816403294176419,
+      "eval_loss": 0.32114630937576294,
+      "eval_precision_macroaverage": 0.7439352759741779,
+      "eval_precision_microaverage": 0.8816403294176419,
+      "eval_recall_macroaverage": 0.6743822635668855,
+      "eval_recall_microaverage": 0.8816403294176419,
+      "eval_runtime": 1173.6874,
+      "eval_samples_per_second": 231.951,
+      "eval_steps_per_second": 3.624,
+      "step": 20000
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3282958037378471e-05,
+      "loss": 0.1905,
+      "step": 20001
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3282622202072104e-05,
+      "loss": 0.3189,
+      "step": 20002
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.328228636676574e-05,
+      "loss": 0.2365,
+      "step": 20003
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3281950531459373e-05,
+      "loss": 0.316,
+      "step": 20004
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3281614696153006e-05,
+      "loss": 0.2225,
+      "step": 20005
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3281278860846643e-05,
+      "loss": 0.2299,
+      "step": 20006
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3280943025540276e-05,
+      "loss": 0.3338,
+      "step": 20007
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.328060719023391e-05,
+      "loss": 0.3268,
+      "step": 20008
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3280271354927545e-05,
+      "loss": 0.1255,
+      "step": 20009
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.327993551962118e-05,
+      "loss": 0.3679,
+      "step": 20010
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3279599684314813e-05,
+      "loss": 0.2223,
+      "step": 20011
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3279263849008449e-05,
+      "loss": 0.2564,
+      "step": 20012
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3278928013702082e-05,
+      "loss": 0.3363,
+      "step": 20013
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3278592178395715e-05,
+      "loss": 0.3507,
+      "step": 20014
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3278256343089351e-05,
+      "loss": 0.2719,
+      "step": 20015
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3277920507782984e-05,
+      "loss": 0.3011,
+      "step": 20016
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3277584672476617e-05,
+      "loss": 0.4017,
+      "step": 20017
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3277248837170253e-05,
+      "loss": 0.3628,
+      "step": 20018
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3276913001863886e-05,
+      "loss": 0.1281,
+      "step": 20019
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3276577166557521e-05,
+      "loss": 0.2682,
+      "step": 20020
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3276241331251156e-05,
+      "loss": 0.236,
+      "step": 20021
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.327590549594479e-05,
+      "loss": 0.3739,
+      "step": 20022
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3275569660638423e-05,
+      "loss": 0.3369,
+      "step": 20023
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.327523382533206e-05,
+      "loss": 0.2481,
+      "step": 20024
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3274897990025692e-05,
+      "loss": 0.3053,
+      "step": 20025
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3274562154719325e-05,
+      "loss": 0.1922,
+      "step": 20026
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3274226319412962e-05,
+      "loss": 0.302,
+      "step": 20027
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3273890484106595e-05,
+      "loss": 0.2547,
+      "step": 20028
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.327355464880023e-05,
+      "loss": 0.2878,
+      "step": 20029
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3273218813493864e-05,
+      "loss": 0.3009,
+      "step": 20030
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3272882978187499e-05,
+      "loss": 0.2925,
+      "step": 20031
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3272547142881132e-05,
+      "loss": 0.3154,
+      "step": 20032
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3272211307574768e-05,
+      "loss": 0.5421,
+      "step": 20033
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3271875472268401e-05,
+      "loss": 0.5482,
+      "step": 20034
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3271539636962034e-05,
+      "loss": 0.1064,
+      "step": 20035
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.327120380165567e-05,
+      "loss": 0.2834,
+      "step": 20036
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3270867966349303e-05,
+      "loss": 0.459,
+      "step": 20037
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3270532131042936e-05,
+      "loss": 0.3189,
+      "step": 20038
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3270196295736572e-05,
+      "loss": 0.345,
+      "step": 20039
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3269860460430205e-05,
+      "loss": 0.4047,
+      "step": 20040
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.326952462512384e-05,
+      "loss": 0.2867,
+      "step": 20041
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3269188789817475e-05,
+      "loss": 0.3781,
+      "step": 20042
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.326885295451111e-05,
+      "loss": 0.2332,
+      "step": 20043
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3268517119204742e-05,
+      "loss": 0.5068,
+      "step": 20044
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3268181283898379e-05,
+      "loss": 0.2299,
+      "step": 20045
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3267845448592012e-05,
+      "loss": 0.2296,
+      "step": 20046
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3267509613285645e-05,
+      "loss": 0.3685,
+      "step": 20047
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3267173777979281e-05,
+      "loss": 0.2302,
+      "step": 20048
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3266837942672914e-05,
+      "loss": 0.3479,
+      "step": 20049
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3266502107366549e-05,
+      "loss": 0.468,
+      "step": 20050
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3266166272060183e-05,
+      "loss": 0.2849,
+      "step": 20051
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3265830436753818e-05,
+      "loss": 0.2064,
+      "step": 20052
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.326549460144745e-05,
+      "loss": 0.2678,
+      "step": 20053
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3265158766141087e-05,
+      "loss": 0.1958,
+      "step": 20054
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.326482293083472e-05,
+      "loss": 0.5429,
+      "step": 20055
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3264487095528353e-05,
+      "loss": 0.274,
+      "step": 20056
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.326415126022199e-05,
+      "loss": 0.501,
+      "step": 20057
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3263815424915622e-05,
+      "loss": 0.2503,
+      "step": 20058
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3263479589609255e-05,
+      "loss": 0.3323,
+      "step": 20059
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3263143754302892e-05,
+      "loss": 0.2177,
+      "step": 20060
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3262807918996525e-05,
+      "loss": 0.3144,
+      "step": 20061
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.326247208369016e-05,
+      "loss": 0.4281,
+      "step": 20062
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3262136248383794e-05,
+      "loss": 0.2297,
+      "step": 20063
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3261800413077428e-05,
+      "loss": 0.3641,
+      "step": 20064
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3261464577771061e-05,
+      "loss": 0.1411,
+      "step": 20065
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3261128742464698e-05,
+      "loss": 0.3456,
+      "step": 20066
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.326079290715833e-05,
+      "loss": 0.3383,
+      "step": 20067
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3260457071851964e-05,
+      "loss": 0.2272,
+      "step": 20068
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.32601212365456e-05,
+      "loss": 0.2978,
+      "step": 20069
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3259785401239233e-05,
+      "loss": 0.2981,
+      "step": 20070
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3259449565932868e-05,
+      "loss": 0.3692,
+      "step": 20071
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3259113730626502e-05,
+      "loss": 0.2283,
+      "step": 20072
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3258777895320137e-05,
+      "loss": 0.3343,
+      "step": 20073
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.325844206001377e-05,
+      "loss": 0.351,
+      "step": 20074
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3258106224707406e-05,
+      "loss": 0.2674,
+      "step": 20075
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.325777038940104e-05,
+      "loss": 0.2267,
+      "step": 20076
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3257434554094672e-05,
+      "loss": 0.3806,
+      "step": 20077
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3257098718788308e-05,
+      "loss": 0.3935,
+      "step": 20078
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3256762883481941e-05,
+      "loss": 0.6118,
+      "step": 20079
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3256427048175576e-05,
+      "loss": 0.2601,
+      "step": 20080
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.325609121286921e-05,
+      "loss": 0.4301,
+      "step": 20081
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3255755377562844e-05,
+      "loss": 0.2596,
+      "step": 20082
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3255419542256478e-05,
+      "loss": 0.3057,
+      "step": 20083
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3255083706950113e-05,
+      "loss": 0.2772,
+      "step": 20084
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3254747871643748e-05,
+      "loss": 0.4154,
+      "step": 20085
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.325441203633738e-05,
+      "loss": 0.3348,
+      "step": 20086
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3254076201031017e-05,
+      "loss": 0.2951,
+      "step": 20087
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.325374036572465e-05,
+      "loss": 0.2836,
+      "step": 20088
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3253404530418283e-05,
+      "loss": 0.3543,
+      "step": 20089
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3253068695111919e-05,
+      "loss": 0.2867,
+      "step": 20090
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3252732859805552e-05,
+      "loss": 0.3514,
+      "step": 20091
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3252397024499187e-05,
+      "loss": 0.2763,
+      "step": 20092
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3252061189192821e-05,
+      "loss": 0.3312,
+      "step": 20093
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3251725353886456e-05,
+      "loss": 0.3231,
+      "step": 20094
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3251389518580089e-05,
+      "loss": 0.4399,
+      "step": 20095
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3251053683273725e-05,
+      "loss": 0.1933,
+      "step": 20096
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3250717847967358e-05,
+      "loss": 0.2361,
+      "step": 20097
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3250382012660991e-05,
+      "loss": 0.1906,
+      "step": 20098
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3250046177354628e-05,
+      "loss": 0.4358,
+      "step": 20099
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.324971034204826e-05,
+      "loss": 0.3275,
+      "step": 20100
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3249374506741895e-05,
+      "loss": 0.4598,
+      "step": 20101
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.324903867143553e-05,
+      "loss": 0.3423,
+      "step": 20102
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3248702836129163e-05,
+      "loss": 0.241,
+      "step": 20103
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3248367000822797e-05,
+      "loss": 0.2406,
+      "step": 20104
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3248031165516432e-05,
+      "loss": 0.2815,
+      "step": 20105
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3247695330210067e-05,
+      "loss": 0.2978,
+      "step": 20106
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.32473594949037e-05,
+      "loss": 0.2385,
+      "step": 20107
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3247023659597336e-05,
+      "loss": 0.2706,
+      "step": 20108
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3246687824290969e-05,
+      "loss": 0.408,
+      "step": 20109
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3246351988984602e-05,
+      "loss": 0.3212,
+      "step": 20110
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3246016153678238e-05,
+      "loss": 0.2369,
+      "step": 20111
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3245680318371871e-05,
+      "loss": 0.2498,
+      "step": 20112
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3245344483065506e-05,
+      "loss": 0.2375,
+      "step": 20113
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.324500864775914e-05,
+      "loss": 0.2088,
+      "step": 20114
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3244672812452775e-05,
+      "loss": 0.4843,
+      "step": 20115
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3244336977146408e-05,
+      "loss": 0.4214,
+      "step": 20116
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3244001141840044e-05,
+      "loss": 0.2576,
+      "step": 20117
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3243665306533677e-05,
+      "loss": 0.2166,
+      "step": 20118
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.324332947122731e-05,
+      "loss": 0.5557,
+      "step": 20119
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3242993635920947e-05,
+      "loss": 0.2738,
+      "step": 20120
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.324265780061458e-05,
+      "loss": 0.3919,
+      "step": 20121
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3242321965308214e-05,
+      "loss": 0.2734,
+      "step": 20122
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3241986130001847e-05,
+      "loss": 0.247,
+      "step": 20123
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3241650294695484e-05,
+      "loss": 0.3232,
+      "step": 20124
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3241314459389117e-05,
+      "loss": 0.4289,
+      "step": 20125
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.324097862408275e-05,
+      "loss": 0.3981,
+      "step": 20126
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3240642788776386e-05,
+      "loss": 0.3547,
+      "step": 20127
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3240306953470019e-05,
+      "loss": 0.3704,
+      "step": 20128
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3239971118163652e-05,
+      "loss": 0.2438,
+      "step": 20129
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3239635282857288e-05,
+      "loss": 0.1849,
+      "step": 20130
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3239299447550921e-05,
+      "loss": 0.2437,
+      "step": 20131
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3238963612244556e-05,
+      "loss": 0.3828,
+      "step": 20132
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.323862777693819e-05,
+      "loss": 0.2393,
+      "step": 20133
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3238291941631825e-05,
+      "loss": 0.2256,
+      "step": 20134
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3237956106325458e-05,
+      "loss": 0.2436,
+      "step": 20135
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3237620271019094e-05,
+      "loss": 0.1969,
+      "step": 20136
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3237284435712727e-05,
+      "loss": 0.4981,
+      "step": 20137
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.323694860040636e-05,
+      "loss": 0.3783,
+      "step": 20138
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3236612765099997e-05,
+      "loss": 0.5327,
+      "step": 20139
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.323627692979363e-05,
+      "loss": 0.4026,
+      "step": 20140
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3235941094487264e-05,
+      "loss": 0.2269,
+      "step": 20141
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3235605259180899e-05,
+      "loss": 0.2204,
+      "step": 20142
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3235269423874533e-05,
+      "loss": 0.1507,
+      "step": 20143
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3234933588568166e-05,
+      "loss": 0.3871,
+      "step": 20144
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3234597753261803e-05,
+      "loss": 0.2807,
+      "step": 20145
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3234261917955436e-05,
+      "loss": 0.4045,
+      "step": 20146
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3233926082649069e-05,
+      "loss": 0.3117,
+      "step": 20147
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.3233590247342705e-05,
+      "loss": 0.2695,
+      "step": 20148
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3233254412036338e-05,
+      "loss": 0.1971,
+      "step": 20149
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3232918576729971e-05,
+      "loss": 0.3003,
+      "step": 20150
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3232582741423607e-05,
+      "loss": 0.5245,
+      "step": 20151
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.323224690611724e-05,
+      "loss": 0.3358,
+      "step": 20152
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3231911070810875e-05,
+      "loss": 0.2367,
+      "step": 20153
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.323157523550451e-05,
+      "loss": 0.2422,
+      "step": 20154
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3231239400198144e-05,
+      "loss": 0.2237,
+      "step": 20155
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3230903564891777e-05,
+      "loss": 0.2665,
+      "step": 20156
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3230567729585413e-05,
+      "loss": 0.2667,
+      "step": 20157
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3230231894279046e-05,
+      "loss": 0.4885,
+      "step": 20158
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.322989605897268e-05,
+      "loss": 0.1702,
+      "step": 20159
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3229560223666316e-05,
+      "loss": 0.1902,
+      "step": 20160
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3229224388359949e-05,
+      "loss": 0.1893,
+      "step": 20161
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3228888553053583e-05,
+      "loss": 0.4267,
+      "step": 20162
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3228552717747218e-05,
+      "loss": 0.2275,
+      "step": 20163
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3228216882440853e-05,
+      "loss": 0.4195,
+      "step": 20164
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3227881047134486e-05,
+      "loss": 0.323,
+      "step": 20165
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3227545211828122e-05,
+      "loss": 0.2737,
+      "step": 20166
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3227209376521755e-05,
+      "loss": 0.2529,
+      "step": 20167
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3226873541215388e-05,
+      "loss": 0.1242,
+      "step": 20168
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3226537705909024e-05,
+      "loss": 0.322,
+      "step": 20169
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3226201870602657e-05,
+      "loss": 0.2782,
+      "step": 20170
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3225866035296292e-05,
+      "loss": 0.7001,
+      "step": 20171
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3225530199989926e-05,
+      "loss": 0.3127,
+      "step": 20172
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.322519436468356e-05,
+      "loss": 0.2508,
+      "step": 20173
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3224858529377194e-05,
+      "loss": 0.2014,
+      "step": 20174
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3224522694070829e-05,
+      "loss": 0.3755,
+      "step": 20175
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3224186858764463e-05,
+      "loss": 0.3651,
+      "step": 20176
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3223851023458096e-05,
+      "loss": 0.3802,
+      "step": 20177
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3223515188151733e-05,
+      "loss": 0.2697,
+      "step": 20178
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3223179352845365e-05,
+      "loss": 0.4147,
+      "step": 20179
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3222843517538998e-05,
+      "loss": 0.3479,
+      "step": 20180
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3222507682232635e-05,
+      "loss": 0.3009,
+      "step": 20181
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3222171846926268e-05,
+      "loss": 0.2886,
+      "step": 20182
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3221836011619902e-05,
+      "loss": 0.2217,
+      "step": 20183
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3221500176313537e-05,
+      "loss": 0.1597,
+      "step": 20184
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3221164341007172e-05,
+      "loss": 0.2443,
+      "step": 20185
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3220828505700805e-05,
+      "loss": 0.454,
+      "step": 20186
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3220492670394441e-05,
+      "loss": 0.2925,
+      "step": 20187
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3220156835088074e-05,
+      "loss": 0.3006,
+      "step": 20188
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3219820999781707e-05,
+      "loss": 0.4384,
+      "step": 20189
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3219485164475343e-05,
+      "loss": 0.2629,
+      "step": 20190
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3219149329168976e-05,
+      "loss": 0.4891,
+      "step": 20191
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.321881349386261e-05,
+      "loss": 0.3186,
+      "step": 20192
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3218477658556245e-05,
+      "loss": 0.1579,
+      "step": 20193
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3218141823249878e-05,
+      "loss": 0.4431,
+      "step": 20194
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3217805987943513e-05,
+      "loss": 0.4117,
+      "step": 20195
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3217470152637148e-05,
+      "loss": 0.2362,
+      "step": 20196
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3217134317330782e-05,
+      "loss": 0.2264,
+      "step": 20197
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3216798482024415e-05,
+      "loss": 0.1701,
+      "step": 20198
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3216462646718052e-05,
+      "loss": 0.3396,
+      "step": 20199
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3216126811411685e-05,
+      "loss": 0.2803,
+      "step": 20200
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3215790976105318e-05,
+      "loss": 0.1001,
+      "step": 20201
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3215455140798954e-05,
+      "loss": 0.2936,
+      "step": 20202
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3215119305492587e-05,
+      "loss": 0.3109,
+      "step": 20203
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3214783470186222e-05,
+      "loss": 0.2759,
+      "step": 20204
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3214447634879856e-05,
+      "loss": 0.2505,
+      "step": 20205
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.321411179957349e-05,
+      "loss": 0.3142,
+      "step": 20206
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3213775964267124e-05,
+      "loss": 0.281,
+      "step": 20207
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.321344012896076e-05,
+      "loss": 0.2742,
+      "step": 20208
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3213104293654393e-05,
+      "loss": 0.4494,
+      "step": 20209
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3212768458348026e-05,
+      "loss": 0.3652,
+      "step": 20210
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3212432623041662e-05,
+      "loss": 0.2978,
+      "step": 20211
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3212096787735295e-05,
+      "loss": 0.3517,
+      "step": 20212
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.321176095242893e-05,
+      "loss": 0.2999,
+      "step": 20213
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3211425117122565e-05,
+      "loss": 0.1872,
+      "step": 20214
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.32110892818162e-05,
+      "loss": 0.2857,
+      "step": 20215
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3210753446509832e-05,
+      "loss": 0.347,
+      "step": 20216
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3210417611203467e-05,
+      "loss": 0.1857,
+      "step": 20217
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3210081775897102e-05,
+      "loss": 0.1898,
+      "step": 20218
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3209745940590734e-05,
+      "loss": 0.3951,
+      "step": 20219
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.320941010528437e-05,
+      "loss": 0.3306,
+      "step": 20220
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3209074269978004e-05,
+      "loss": 0.3213,
+      "step": 20221
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3208738434671637e-05,
+      "loss": 0.2713,
+      "step": 20222
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3208402599365273e-05,
+      "loss": 0.3363,
+      "step": 20223
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3208066764058906e-05,
+      "loss": 0.3393,
+      "step": 20224
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.320773092875254e-05,
+      "loss": 0.1812,
+      "step": 20225
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3207395093446175e-05,
+      "loss": 0.2971,
+      "step": 20226
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.320705925813981e-05,
+      "loss": 0.3591,
+      "step": 20227
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3206723422833443e-05,
+      "loss": 0.2404,
+      "step": 20228
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.320638758752708e-05,
+      "loss": 0.3224,
+      "step": 20229
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3206051752220712e-05,
+      "loss": 0.2482,
+      "step": 20230
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3205715916914345e-05,
+      "loss": 0.2467,
+      "step": 20231
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3205380081607981e-05,
+      "loss": 0.3052,
+      "step": 20232
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3205044246301614e-05,
+      "loss": 0.188,
+      "step": 20233
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3204708410995249e-05,
+      "loss": 0.2128,
+      "step": 20234
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3204372575688884e-05,
+      "loss": 0.2941,
+      "step": 20235
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3204036740382518e-05,
+      "loss": 0.4961,
+      "step": 20236
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3203700905076151e-05,
+      "loss": 0.1528,
+      "step": 20237
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3203365069769786e-05,
+      "loss": 0.1493,
+      "step": 20238
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.320302923446342e-05,
+      "loss": 0.172,
+      "step": 20239
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3202693399157054e-05,
+      "loss": 0.4004,
+      "step": 20240
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.320235756385069e-05,
+      "loss": 0.2282,
+      "step": 20241
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3202021728544323e-05,
+      "loss": 0.2144,
+      "step": 20242
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3201685893237956e-05,
+      "loss": 0.2996,
+      "step": 20243
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3201350057931592e-05,
+      "loss": 0.2185,
+      "step": 20244
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3201014222625225e-05,
+      "loss": 0.4579,
+      "step": 20245
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.320067838731886e-05,
+      "loss": 0.0901,
+      "step": 20246
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3200342552012494e-05,
+      "loss": 0.2965,
+      "step": 20247
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3200006716706129e-05,
+      "loss": 0.2268,
+      "step": 20248
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3199670881399762e-05,
+      "loss": 0.2902,
+      "step": 20249
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3199335046093398e-05,
+      "loss": 0.5505,
+      "step": 20250
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3198999210787031e-05,
+      "loss": 0.282,
+      "step": 20251
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3198663375480664e-05,
+      "loss": 0.286,
+      "step": 20252
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.31983275401743e-05,
+      "loss": 0.2179,
+      "step": 20253
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3197991704867934e-05,
+      "loss": 0.205,
+      "step": 20254
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3197655869561568e-05,
+      "loss": 0.3236,
+      "step": 20255
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3197320034255203e-05,
+      "loss": 0.1949,
+      "step": 20256
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3196984198948838e-05,
+      "loss": 0.2583,
+      "step": 20257
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.319664836364247e-05,
+      "loss": 0.2262,
+      "step": 20258
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3196312528336107e-05,
+      "loss": 0.2529,
+      "step": 20259
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.319597669302974e-05,
+      "loss": 0.335,
+      "step": 20260
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3195640857723373e-05,
+      "loss": 0.2833,
+      "step": 20261
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3195305022417009e-05,
+      "loss": 0.2676,
+      "step": 20262
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3194969187110642e-05,
+      "loss": 0.5359,
+      "step": 20263
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3194633351804275e-05,
+      "loss": 0.2199,
+      "step": 20264
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3194297516497911e-05,
+      "loss": 0.4052,
+      "step": 20265
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3193961681191544e-05,
+      "loss": 0.229,
+      "step": 20266
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3193625845885179e-05,
+      "loss": 0.2813,
+      "step": 20267
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3193290010578814e-05,
+      "loss": 0.1901,
+      "step": 20268
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3192954175272448e-05,
+      "loss": 0.3676,
+      "step": 20269
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3192618339966081e-05,
+      "loss": 0.3711,
+      "step": 20270
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3192282504659717e-05,
+      "loss": 0.2611,
+      "step": 20271
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.319194666935335e-05,
+      "loss": 0.4091,
+      "step": 20272
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3191610834046983e-05,
+      "loss": 0.284,
+      "step": 20273
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.319127499874062e-05,
+      "loss": 0.1801,
+      "step": 20274
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3190939163434253e-05,
+      "loss": 0.2842,
+      "step": 20275
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3190603328127887e-05,
+      "loss": 0.3259,
+      "step": 20276
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3190267492821522e-05,
+      "loss": 0.4086,
+      "step": 20277
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3189931657515157e-05,
+      "loss": 0.2934,
+      "step": 20278
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.318959582220879e-05,
+      "loss": 0.3619,
+      "step": 20279
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3189259986902426e-05,
+      "loss": 0.2706,
+      "step": 20280
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3188924151596059e-05,
+      "loss": 0.3286,
+      "step": 20281
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3188588316289692e-05,
+      "loss": 0.1536,
+      "step": 20282
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3188252480983328e-05,
+      "loss": 0.2404,
+      "step": 20283
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3187916645676961e-05,
+      "loss": 0.3125,
+      "step": 20284
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3187580810370594e-05,
+      "loss": 0.2481,
+      "step": 20285
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.318724497506423e-05,
+      "loss": 0.3968,
+      "step": 20286
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3186909139757863e-05,
+      "loss": 0.2818,
+      "step": 20287
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3186573304451498e-05,
+      "loss": 0.3605,
+      "step": 20288
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3186237469145133e-05,
+      "loss": 0.2522,
+      "step": 20289
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3185901633838767e-05,
+      "loss": 0.4156,
+      "step": 20290
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.31855657985324e-05,
+      "loss": 0.2441,
+      "step": 20291
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3185229963226037e-05,
+      "loss": 0.1819,
+      "step": 20292
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.318489412791967e-05,
+      "loss": 0.1652,
+      "step": 20293
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3184558292613303e-05,
+      "loss": 0.2266,
+      "step": 20294
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3184222457306939e-05,
+      "loss": 0.3994,
+      "step": 20295
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3183886622000572e-05,
+      "loss": 0.3871,
+      "step": 20296
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3183550786694206e-05,
+      "loss": 0.2057,
+      "step": 20297
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3183214951387841e-05,
+      "loss": 0.2716,
+      "step": 20298
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3182879116081476e-05,
+      "loss": 0.1801,
+      "step": 20299
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3182543280775109e-05,
+      "loss": 0.465,
+      "step": 20300
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3182207445468745e-05,
+      "loss": 0.2238,
+      "step": 20301
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3181871610162378e-05,
+      "loss": 0.2576,
+      "step": 20302
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3181535774856011e-05,
+      "loss": 0.4397,
+      "step": 20303
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3181199939549646e-05,
+      "loss": 0.2008,
+      "step": 20304
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.318086410424328e-05,
+      "loss": 0.2364,
+      "step": 20305
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3180528268936915e-05,
+      "loss": 0.2139,
+      "step": 20306
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3180192433630548e-05,
+      "loss": 0.2618,
+      "step": 20307
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3179856598324182e-05,
+      "loss": 0.3599,
+      "step": 20308
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3179520763017817e-05,
+      "loss": 0.2565,
+      "step": 20309
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.317918492771145e-05,
+      "loss": 0.2647,
+      "step": 20310
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3178849092405086e-05,
+      "loss": 0.3534,
+      "step": 20311
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.317851325709872e-05,
+      "loss": 0.3525,
+      "step": 20312
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3178177421792352e-05,
+      "loss": 0.1267,
+      "step": 20313
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3177841586485989e-05,
+      "loss": 0.3933,
+      "step": 20314
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3177505751179622e-05,
+      "loss": 0.2701,
+      "step": 20315
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3177169915873256e-05,
+      "loss": 0.2786,
+      "step": 20316
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3176834080566891e-05,
+      "loss": 0.3545,
+      "step": 20317
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3176498245260526e-05,
+      "loss": 0.4304,
+      "step": 20318
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3176162409954159e-05,
+      "loss": 0.4002,
+      "step": 20319
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3175826574647795e-05,
+      "loss": 0.335,
+      "step": 20320
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3175490739341428e-05,
+      "loss": 0.2494,
+      "step": 20321
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.317515490403506e-05,
+      "loss": 0.2166,
+      "step": 20322
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3174819068728697e-05,
+      "loss": 0.2411,
+      "step": 20323
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.317448323342233e-05,
+      "loss": 0.2151,
+      "step": 20324
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3174147398115965e-05,
+      "loss": 0.2611,
+      "step": 20325
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.31738115628096e-05,
+      "loss": 0.2447,
+      "step": 20326
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3173475727503234e-05,
+      "loss": 0.3731,
+      "step": 20327
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3173139892196867e-05,
+      "loss": 0.3822,
+      "step": 20328
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3172804056890502e-05,
+      "loss": 0.2654,
+      "step": 20329
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3172468221584136e-05,
+      "loss": 0.4346,
+      "step": 20330
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.317213238627777e-05,
+      "loss": 0.2852,
+      "step": 20331
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3171796550971406e-05,
+      "loss": 0.2053,
+      "step": 20332
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3171460715665039e-05,
+      "loss": 0.2512,
+      "step": 20333
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3171124880358671e-05,
+      "loss": 0.2512,
+      "step": 20334
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3170789045052308e-05,
+      "loss": 0.3916,
+      "step": 20335
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.317045320974594e-05,
+      "loss": 0.2716,
+      "step": 20336
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3170117374439575e-05,
+      "loss": 0.2306,
+      "step": 20337
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.316978153913321e-05,
+      "loss": 0.2114,
+      "step": 20338
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3169445703826845e-05,
+      "loss": 0.3375,
+      "step": 20339
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3169109868520478e-05,
+      "loss": 0.1663,
+      "step": 20340
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3168774033214114e-05,
+      "loss": 0.237,
+      "step": 20341
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3168438197907747e-05,
+      "loss": 0.1642,
+      "step": 20342
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.316810236260138e-05,
+      "loss": 0.3595,
+      "step": 20343
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3167766527295016e-05,
+      "loss": 0.2842,
+      "step": 20344
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.316743069198865e-05,
+      "loss": 0.3748,
+      "step": 20345
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3167094856682284e-05,
+      "loss": 0.3821,
+      "step": 20346
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.3166759021375918e-05,
+      "loss": 0.4435,
+      "step": 20347
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3166423186069553e-05,
+      "loss": 0.2384,
+      "step": 20348
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3166087350763186e-05,
+      "loss": 0.3146,
+      "step": 20349
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3165751515456822e-05,
+      "loss": 0.3603,
+      "step": 20350
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3165415680150455e-05,
+      "loss": 0.1678,
+      "step": 20351
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3165079844844088e-05,
+      "loss": 0.4157,
+      "step": 20352
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3164744009537725e-05,
+      "loss": 0.3424,
+      "step": 20353
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3164408174231358e-05,
+      "loss": 0.3297,
+      "step": 20354
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.316407233892499e-05,
+      "loss": 0.3061,
+      "step": 20355
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3163736503618627e-05,
+      "loss": 0.3584,
+      "step": 20356
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.316340066831226e-05,
+      "loss": 0.3243,
+      "step": 20357
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3163064833005895e-05,
+      "loss": 0.2542,
+      "step": 20358
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.316272899769953e-05,
+      "loss": 0.1772,
+      "step": 20359
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3162393162393164e-05,
+      "loss": 0.4495,
+      "step": 20360
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3162057327086797e-05,
+      "loss": 0.3327,
+      "step": 20361
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3161721491780433e-05,
+      "loss": 0.296,
+      "step": 20362
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3161385656474066e-05,
+      "loss": 0.2134,
+      "step": 20363
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3161049821167699e-05,
+      "loss": 0.3147,
+      "step": 20364
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3160713985861335e-05,
+      "loss": 0.2414,
+      "step": 20365
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3160378150554968e-05,
+      "loss": 0.5148,
+      "step": 20366
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3160042315248603e-05,
+      "loss": 0.3473,
+      "step": 20367
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3159706479942238e-05,
+      "loss": 0.2718,
+      "step": 20368
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3159370644635872e-05,
+      "loss": 0.2373,
+      "step": 20369
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3159034809329505e-05,
+      "loss": 0.2057,
+      "step": 20370
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3158698974023142e-05,
+      "loss": 0.2283,
+      "step": 20371
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3158363138716775e-05,
+      "loss": 0.337,
+      "step": 20372
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3158027303410407e-05,
+      "loss": 0.249,
+      "step": 20373
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3157691468104044e-05,
+      "loss": 0.2942,
+      "step": 20374
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3157355632797677e-05,
+      "loss": 0.5147,
+      "step": 20375
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.315701979749131e-05,
+      "loss": 0.2201,
+      "step": 20376
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3156683962184946e-05,
+      "loss": 0.2446,
+      "step": 20377
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3156348126878579e-05,
+      "loss": 0.2222,
+      "step": 20378
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3156012291572214e-05,
+      "loss": 0.1875,
+      "step": 20379
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3155676456265848e-05,
+      "loss": 0.3751,
+      "step": 20380
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3155340620959483e-05,
+      "loss": 0.2445,
+      "step": 20381
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3155004785653116e-05,
+      "loss": 0.2847,
+      "step": 20382
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3154668950346752e-05,
+      "loss": 0.2711,
+      "step": 20383
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3154333115040385e-05,
+      "loss": 0.2323,
+      "step": 20384
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3153997279734018e-05,
+      "loss": 0.3403,
+      "step": 20385
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3153661444427654e-05,
+      "loss": 0.162,
+      "step": 20386
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3153325609121287e-05,
+      "loss": 0.4051,
+      "step": 20387
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3152989773814922e-05,
+      "loss": 0.2853,
+      "step": 20388
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3152653938508557e-05,
+      "loss": 0.2888,
+      "step": 20389
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3152318103202191e-05,
+      "loss": 0.3192,
+      "step": 20390
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3151982267895824e-05,
+      "loss": 0.1684,
+      "step": 20391
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.315164643258946e-05,
+      "loss": 0.1863,
+      "step": 20392
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3151310597283094e-05,
+      "loss": 0.2865,
+      "step": 20393
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3150974761976727e-05,
+      "loss": 0.2684,
+      "step": 20394
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3150638926670363e-05,
+      "loss": 0.3197,
+      "step": 20395
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3150303091363996e-05,
+      "loss": 0.4499,
+      "step": 20396
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.314996725605763e-05,
+      "loss": 0.3169,
+      "step": 20397
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3149631420751265e-05,
+      "loss": 0.4772,
+      "step": 20398
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3149295585444898e-05,
+      "loss": 0.1754,
+      "step": 20399
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3148959750138533e-05,
+      "loss": 0.2962,
+      "step": 20400
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3148623914832167e-05,
+      "loss": 0.2128,
+      "step": 20401
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3148288079525802e-05,
+      "loss": 0.1801,
+      "step": 20402
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3147952244219435e-05,
+      "loss": 0.2311,
+      "step": 20403
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3147616408913071e-05,
+      "loss": 0.1952,
+      "step": 20404
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3147280573606704e-05,
+      "loss": 0.2097,
+      "step": 20405
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3146944738300337e-05,
+      "loss": 0.3363,
+      "step": 20406
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3146608902993974e-05,
+      "loss": 0.3464,
+      "step": 20407
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3146273067687607e-05,
+      "loss": 0.2946,
+      "step": 20408
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3145937232381241e-05,
+      "loss": 0.4663,
+      "step": 20409
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3145601397074876e-05,
+      "loss": 0.258,
+      "step": 20410
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.314526556176851e-05,
+      "loss": 0.2662,
+      "step": 20411
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3144929726462143e-05,
+      "loss": 0.3713,
+      "step": 20412
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.314459389115578e-05,
+      "loss": 0.3327,
+      "step": 20413
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3144258055849413e-05,
+      "loss": 0.2458,
+      "step": 20414
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3143922220543046e-05,
+      "loss": 0.2913,
+      "step": 20415
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3143586385236682e-05,
+      "loss": 0.1936,
+      "step": 20416
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3143250549930315e-05,
+      "loss": 0.3195,
+      "step": 20417
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.314291471462395e-05,
+      "loss": 0.3468,
+      "step": 20418
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3142578879317584e-05,
+      "loss": 0.47,
+      "step": 20419
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3142243044011217e-05,
+      "loss": 0.354,
+      "step": 20420
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3141907208704852e-05,
+      "loss": 0.4068,
+      "step": 20421
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3141571373398487e-05,
+      "loss": 0.2754,
+      "step": 20422
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3141235538092121e-05,
+      "loss": 0.3203,
+      "step": 20423
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3140899702785754e-05,
+      "loss": 0.2408,
+      "step": 20424
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.314056386747939e-05,
+      "loss": 0.2448,
+      "step": 20425
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3140228032173023e-05,
+      "loss": 0.3406,
+      "step": 20426
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3139892196866656e-05,
+      "loss": 0.2567,
+      "step": 20427
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3139556361560293e-05,
+      "loss": 0.2116,
+      "step": 20428
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3139220526253926e-05,
+      "loss": 0.3657,
+      "step": 20429
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.313888469094756e-05,
+      "loss": 0.4462,
+      "step": 20430
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3138548855641195e-05,
+      "loss": 0.2043,
+      "step": 20431
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.313821302033483e-05,
+      "loss": 0.2921,
+      "step": 20432
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3137877185028463e-05,
+      "loss": 0.3641,
+      "step": 20433
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3137541349722099e-05,
+      "loss": 0.4488,
+      "step": 20434
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3137205514415732e-05,
+      "loss": 0.3133,
+      "step": 20435
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3136869679109365e-05,
+      "loss": 0.2951,
+      "step": 20436
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3136533843803001e-05,
+      "loss": 0.4791,
+      "step": 20437
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3136198008496634e-05,
+      "loss": 0.2852,
+      "step": 20438
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3135862173190269e-05,
+      "loss": 0.2831,
+      "step": 20439
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3135526337883903e-05,
+      "loss": 0.3179,
+      "step": 20440
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3135190502577538e-05,
+      "loss": 0.2392,
+      "step": 20441
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3134854667271171e-05,
+      "loss": 0.1797,
+      "step": 20442
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3134518831964806e-05,
+      "loss": 0.3038,
+      "step": 20443
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.313418299665844e-05,
+      "loss": 0.422,
+      "step": 20444
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3133847161352073e-05,
+      "loss": 0.3127,
+      "step": 20445
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.313351132604571e-05,
+      "loss": 0.1823,
+      "step": 20446
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3133175490739343e-05,
+      "loss": 0.2511,
+      "step": 20447
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3132839655432976e-05,
+      "loss": 0.3115,
+      "step": 20448
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3132503820126612e-05,
+      "loss": 0.429,
+      "step": 20449
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3132167984820245e-05,
+      "loss": 0.2918,
+      "step": 20450
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.313183214951388e-05,
+      "loss": 0.1742,
+      "step": 20451
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3131496314207514e-05,
+      "loss": 0.2356,
+      "step": 20452
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3131160478901149e-05,
+      "loss": 0.2537,
+      "step": 20453
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3130824643594782e-05,
+      "loss": 0.3879,
+      "step": 20454
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3130488808288418e-05,
+      "loss": 0.2439,
+      "step": 20455
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3130152972982051e-05,
+      "loss": 0.3453,
+      "step": 20456
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3129817137675684e-05,
+      "loss": 0.3859,
+      "step": 20457
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.312948130236932e-05,
+      "loss": 0.3114,
+      "step": 20458
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3129145467062953e-05,
+      "loss": 0.2349,
+      "step": 20459
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3128809631756588e-05,
+      "loss": 0.3211,
+      "step": 20460
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3128473796450223e-05,
+      "loss": 0.3362,
+      "step": 20461
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3128137961143857e-05,
+      "loss": 0.1941,
+      "step": 20462
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.312780212583749e-05,
+      "loss": 0.2301,
+      "step": 20463
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3127466290531125e-05,
+      "loss": 0.175,
+      "step": 20464
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.312713045522476e-05,
+      "loss": 0.3379,
+      "step": 20465
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3126794619918392e-05,
+      "loss": 0.2931,
+      "step": 20466
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3126458784612029e-05,
+      "loss": 0.2288,
+      "step": 20467
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3126122949305662e-05,
+      "loss": 0.5347,
+      "step": 20468
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3125787113999295e-05,
+      "loss": 0.3524,
+      "step": 20469
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3125451278692931e-05,
+      "loss": 0.263,
+      "step": 20470
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3125115443386564e-05,
+      "loss": 0.218,
+      "step": 20471
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3124779608080199e-05,
+      "loss": 0.314,
+      "step": 20472
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3124443772773833e-05,
+      "loss": 0.1992,
+      "step": 20473
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3124107937467468e-05,
+      "loss": 0.2663,
+      "step": 20474
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3123772102161101e-05,
+      "loss": 0.3743,
+      "step": 20475
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3123436266854737e-05,
+      "loss": 0.2337,
+      "step": 20476
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.312310043154837e-05,
+      "loss": 0.2619,
+      "step": 20477
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3122764596242003e-05,
+      "loss": 0.3288,
+      "step": 20478
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.312242876093564e-05,
+      "loss": 0.3722,
+      "step": 20479
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3122092925629272e-05,
+      "loss": 0.4467,
+      "step": 20480
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3121757090322907e-05,
+      "loss": 0.3005,
+      "step": 20481
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3121421255016542e-05,
+      "loss": 0.2837,
+      "step": 20482
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3121085419710176e-05,
+      "loss": 0.5185,
+      "step": 20483
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.312074958440381e-05,
+      "loss": 0.3369,
+      "step": 20484
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3120413749097442e-05,
+      "loss": 0.5024,
+      "step": 20485
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3120077913791079e-05,
+      "loss": 0.243,
+      "step": 20486
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3119742078484712e-05,
+      "loss": 0.254,
+      "step": 20487
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3119406243178344e-05,
+      "loss": 0.373,
+      "step": 20488
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.311907040787198e-05,
+      "loss": 0.3377,
+      "step": 20489
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3118734572565614e-05,
+      "loss": 0.2571,
+      "step": 20490
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3118398737259248e-05,
+      "loss": 0.2645,
+      "step": 20491
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3118062901952883e-05,
+      "loss": 0.2196,
+      "step": 20492
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3117727066646518e-05,
+      "loss": 0.2007,
+      "step": 20493
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.311739123134015e-05,
+      "loss": 0.3028,
+      "step": 20494
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3117055396033787e-05,
+      "loss": 0.2781,
+      "step": 20495
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.311671956072742e-05,
+      "loss": 0.2739,
+      "step": 20496
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3116383725421053e-05,
+      "loss": 0.2683,
+      "step": 20497
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.311604789011469e-05,
+      "loss": 0.2381,
+      "step": 20498
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3115712054808322e-05,
+      "loss": 0.3863,
+      "step": 20499
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3115376219501957e-05,
+      "loss": 0.2691,
+      "step": 20500
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3115040384195592e-05,
+      "loss": 0.1885,
+      "step": 20501
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3114704548889226e-05,
+      "loss": 0.2953,
+      "step": 20502
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3114368713582859e-05,
+      "loss": 0.2966,
+      "step": 20503
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3114032878276495e-05,
+      "loss": 0.2895,
+      "step": 20504
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3113697042970128e-05,
+      "loss": 0.2726,
+      "step": 20505
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3113361207663761e-05,
+      "loss": 0.2676,
+      "step": 20506
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3113025372357398e-05,
+      "loss": 0.3348,
+      "step": 20507
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.311268953705103e-05,
+      "loss": 0.4066,
+      "step": 20508
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3112353701744665e-05,
+      "loss": 0.1837,
+      "step": 20509
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.31120178664383e-05,
+      "loss": 0.3371,
+      "step": 20510
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3111682031131933e-05,
+      "loss": 0.4297,
+      "step": 20511
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3111346195825568e-05,
+      "loss": 0.3536,
+      "step": 20512
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3111010360519202e-05,
+      "loss": 0.2414,
+      "step": 20513
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3110674525212837e-05,
+      "loss": 0.2256,
+      "step": 20514
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.311033868990647e-05,
+      "loss": 0.2696,
+      "step": 20515
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3110002854600106e-05,
+      "loss": 0.2766,
+      "step": 20516
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3109667019293739e-05,
+      "loss": 0.2662,
+      "step": 20517
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3109331183987372e-05,
+      "loss": 0.3191,
+      "step": 20518
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3108995348681008e-05,
+      "loss": 0.2802,
+      "step": 20519
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3108659513374641e-05,
+      "loss": 0.2549,
+      "step": 20520
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3108323678068276e-05,
+      "loss": 0.2099,
+      "step": 20521
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.310798784276191e-05,
+      "loss": 0.3475,
+      "step": 20522
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3107652007455545e-05,
+      "loss": 0.2899,
+      "step": 20523
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3107316172149178e-05,
+      "loss": 0.3954,
+      "step": 20524
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3106980336842815e-05,
+      "loss": 0.4645,
+      "step": 20525
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3106644501536448e-05,
+      "loss": 0.3781,
+      "step": 20526
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.310630866623008e-05,
+      "loss": 0.2016,
+      "step": 20527
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3105972830923717e-05,
+      "loss": 0.1935,
+      "step": 20528
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.310563699561735e-05,
+      "loss": 0.3428,
+      "step": 20529
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3105301160310984e-05,
+      "loss": 0.3347,
+      "step": 20530
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3104965325004619e-05,
+      "loss": 0.5448,
+      "step": 20531
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3104629489698254e-05,
+      "loss": 0.5842,
+      "step": 20532
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3104293654391887e-05,
+      "loss": 0.2087,
+      "step": 20533
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3103957819085521e-05,
+      "loss": 0.2235,
+      "step": 20534
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3103621983779156e-05,
+      "loss": 0.1778,
+      "step": 20535
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3103286148472789e-05,
+      "loss": 0.2355,
+      "step": 20536
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3102950313166425e-05,
+      "loss": 0.2329,
+      "step": 20537
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3102614477860058e-05,
+      "loss": 0.2254,
+      "step": 20538
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3102278642553691e-05,
+      "loss": 0.1905,
+      "step": 20539
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3101942807247328e-05,
+      "loss": 0.3966,
+      "step": 20540
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.310160697194096e-05,
+      "loss": 0.329,
+      "step": 20541
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3101271136634595e-05,
+      "loss": 0.3204,
+      "step": 20542
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.310093530132823e-05,
+      "loss": 0.3366,
+      "step": 20543
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3100599466021864e-05,
+      "loss": 0.3228,
+      "step": 20544
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.3100263630715497e-05,
+      "loss": 0.5608,
+      "step": 20545
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3099927795409134e-05,
+      "loss": 0.2054,
+      "step": 20546
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3099591960102767e-05,
+      "loss": 0.3093,
+      "step": 20547
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.30992561247964e-05,
+      "loss": 0.2223,
+      "step": 20548
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3098920289490036e-05,
+      "loss": 0.4371,
+      "step": 20549
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3098584454183669e-05,
+      "loss": 0.1975,
+      "step": 20550
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3098248618877304e-05,
+      "loss": 0.2806,
+      "step": 20551
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3097912783570938e-05,
+      "loss": 0.3163,
+      "step": 20552
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3097576948264573e-05,
+      "loss": 0.3671,
+      "step": 20553
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3097241112958206e-05,
+      "loss": 0.2717,
+      "step": 20554
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.309690527765184e-05,
+      "loss": 0.2916,
+      "step": 20555
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3096569442345475e-05,
+      "loss": 0.2996,
+      "step": 20556
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3096233607039108e-05,
+      "loss": 0.3435,
+      "step": 20557
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3095897771732744e-05,
+      "loss": 0.3369,
+      "step": 20558
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3095561936426377e-05,
+      "loss": 0.3841,
+      "step": 20559
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.309522610112001e-05,
+      "loss": 0.1869,
+      "step": 20560
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3094890265813647e-05,
+      "loss": 0.4875,
+      "step": 20561
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.309455443050728e-05,
+      "loss": 0.4357,
+      "step": 20562
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3094218595200914e-05,
+      "loss": 0.3568,
+      "step": 20563
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3093882759894549e-05,
+      "loss": 0.2945,
+      "step": 20564
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3093546924588184e-05,
+      "loss": 0.21,
+      "step": 20565
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3093211089281816e-05,
+      "loss": 0.2551,
+      "step": 20566
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3092875253975453e-05,
+      "loss": 0.1924,
+      "step": 20567
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3092539418669086e-05,
+      "loss": 0.4828,
+      "step": 20568
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3092203583362719e-05,
+      "loss": 0.2319,
+      "step": 20569
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3091867748056355e-05,
+      "loss": 0.5434,
+      "step": 20570
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3091531912749988e-05,
+      "loss": 0.2734,
+      "step": 20571
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3091196077443623e-05,
+      "loss": 0.326,
+      "step": 20572
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3090860242137257e-05,
+      "loss": 0.1905,
+      "step": 20573
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3090524406830892e-05,
+      "loss": 0.2006,
+      "step": 20574
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3090188571524525e-05,
+      "loss": 0.3114,
+      "step": 20575
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3089852736218161e-05,
+      "loss": 0.311,
+      "step": 20576
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3089516900911794e-05,
+      "loss": 0.3147,
+      "step": 20577
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3089181065605427e-05,
+      "loss": 0.2893,
+      "step": 20578
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3088845230299064e-05,
+      "loss": 0.2953,
+      "step": 20579
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3088509394992696e-05,
+      "loss": 0.3305,
+      "step": 20580
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.308817355968633e-05,
+      "loss": 0.3179,
+      "step": 20581
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3087837724379966e-05,
+      "loss": 0.2988,
+      "step": 20582
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3087501889073599e-05,
+      "loss": 0.3182,
+      "step": 20583
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3087166053767233e-05,
+      "loss": 0.2595,
+      "step": 20584
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3086830218460868e-05,
+      "loss": 0.3741,
+      "step": 20585
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3086494383154503e-05,
+      "loss": 0.3637,
+      "step": 20586
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3086158547848136e-05,
+      "loss": 0.3734,
+      "step": 20587
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3085822712541772e-05,
+      "loss": 0.3014,
+      "step": 20588
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3085486877235405e-05,
+      "loss": 0.2858,
+      "step": 20589
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3085151041929038e-05,
+      "loss": 0.3766,
+      "step": 20590
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3084815206622674e-05,
+      "loss": 0.2871,
+      "step": 20591
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3084479371316307e-05,
+      "loss": 0.1936,
+      "step": 20592
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3084143536009942e-05,
+      "loss": 0.3371,
+      "step": 20593
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3083807700703576e-05,
+      "loss": 0.3282,
+      "step": 20594
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3083471865397211e-05,
+      "loss": 0.3667,
+      "step": 20595
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3083136030090844e-05,
+      "loss": 0.4618,
+      "step": 20596
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.308280019478448e-05,
+      "loss": 0.291,
+      "step": 20597
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3082464359478113e-05,
+      "loss": 0.3578,
+      "step": 20598
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3082128524171746e-05,
+      "loss": 0.3506,
+      "step": 20599
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3081792688865383e-05,
+      "loss": 0.1776,
+      "step": 20600
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3081456853559016e-05,
+      "loss": 0.326,
+      "step": 20601
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3081121018252649e-05,
+      "loss": 0.2998,
+      "step": 20602
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3080785182946285e-05,
+      "loss": 0.3899,
+      "step": 20603
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3080449347639918e-05,
+      "loss": 0.3744,
+      "step": 20604
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3080113512333552e-05,
+      "loss": 0.3412,
+      "step": 20605
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3079777677027187e-05,
+      "loss": 0.1942,
+      "step": 20606
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3079441841720822e-05,
+      "loss": 0.5585,
+      "step": 20607
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3079106006414455e-05,
+      "loss": 0.4008,
+      "step": 20608
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3078770171108091e-05,
+      "loss": 0.2506,
+      "step": 20609
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3078434335801724e-05,
+      "loss": 0.2547,
+      "step": 20610
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3078098500495357e-05,
+      "loss": 0.2614,
+      "step": 20611
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3077762665188993e-05,
+      "loss": 0.2068,
+      "step": 20612
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3077426829882626e-05,
+      "loss": 0.3579,
+      "step": 20613
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3077090994576261e-05,
+      "loss": 0.2068,
+      "step": 20614
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3076755159269896e-05,
+      "loss": 0.349,
+      "step": 20615
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.307641932396353e-05,
+      "loss": 0.2687,
+      "step": 20616
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3076083488657163e-05,
+      "loss": 0.2303,
+      "step": 20617
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.30757476533508e-05,
+      "loss": 0.2712,
+      "step": 20618
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3075411818044432e-05,
+      "loss": 0.3179,
+      "step": 20619
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3075075982738065e-05,
+      "loss": 0.324,
+      "step": 20620
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3074740147431702e-05,
+      "loss": 0.2712,
+      "step": 20621
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3074404312125335e-05,
+      "loss": 0.2797,
+      "step": 20622
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3074068476818968e-05,
+      "loss": 0.3359,
+      "step": 20623
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3073732641512604e-05,
+      "loss": 0.3411,
+      "step": 20624
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3073396806206237e-05,
+      "loss": 0.1168,
+      "step": 20625
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3073060970899872e-05,
+      "loss": 0.4297,
+      "step": 20626
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3072725135593506e-05,
+      "loss": 0.3446,
+      "step": 20627
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3072389300287141e-05,
+      "loss": 0.286,
+      "step": 20628
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3072053464980774e-05,
+      "loss": 0.3082,
+      "step": 20629
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.307171762967441e-05,
+      "loss": 0.1944,
+      "step": 20630
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3071381794368043e-05,
+      "loss": 0.3389,
+      "step": 20631
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3071045959061676e-05,
+      "loss": 0.1431,
+      "step": 20632
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3070710123755312e-05,
+      "loss": 0.1389,
+      "step": 20633
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3070374288448945e-05,
+      "loss": 0.352,
+      "step": 20634
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.307003845314258e-05,
+      "loss": 0.4482,
+      "step": 20635
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3069702617836215e-05,
+      "loss": 0.3554,
+      "step": 20636
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.306936678252985e-05,
+      "loss": 0.2844,
+      "step": 20637
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3069030947223482e-05,
+      "loss": 0.4209,
+      "step": 20638
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3068695111917119e-05,
+      "loss": 0.3275,
+      "step": 20639
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3068359276610752e-05,
+      "loss": 0.4849,
+      "step": 20640
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3068023441304385e-05,
+      "loss": 0.3199,
+      "step": 20641
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3067687605998021e-05,
+      "loss": 0.2827,
+      "step": 20642
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3067351770691654e-05,
+      "loss": 0.1533,
+      "step": 20643
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3067015935385288e-05,
+      "loss": 0.3611,
+      "step": 20644
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3066680100078923e-05,
+      "loss": 0.1809,
+      "step": 20645
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3066344264772556e-05,
+      "loss": 0.2785,
+      "step": 20646
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.306600842946619e-05,
+      "loss": 0.3825,
+      "step": 20647
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3065672594159825e-05,
+      "loss": 0.3662,
+      "step": 20648
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.306533675885346e-05,
+      "loss": 0.3038,
+      "step": 20649
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3065000923547093e-05,
+      "loss": 0.2504,
+      "step": 20650
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.306466508824073e-05,
+      "loss": 0.2886,
+      "step": 20651
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3064329252934362e-05,
+      "loss": 0.2388,
+      "step": 20652
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3063993417627995e-05,
+      "loss": 0.3362,
+      "step": 20653
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3063657582321632e-05,
+      "loss": 0.4927,
+      "step": 20654
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3063321747015265e-05,
+      "loss": 0.2901,
+      "step": 20655
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.30629859117089e-05,
+      "loss": 0.1987,
+      "step": 20656
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3062650076402534e-05,
+      "loss": 0.4615,
+      "step": 20657
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3062314241096168e-05,
+      "loss": 0.2133,
+      "step": 20658
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3061978405789801e-05,
+      "loss": 0.2201,
+      "step": 20659
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3061642570483438e-05,
+      "loss": 0.3204,
+      "step": 20660
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.306130673517707e-05,
+      "loss": 0.3977,
+      "step": 20661
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3060970899870704e-05,
+      "loss": 0.2626,
+      "step": 20662
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.306063506456434e-05,
+      "loss": 0.5326,
+      "step": 20663
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3060299229257973e-05,
+      "loss": 0.2547,
+      "step": 20664
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3059963393951608e-05,
+      "loss": 0.2211,
+      "step": 20665
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.305962755864524e-05,
+      "loss": 0.2272,
+      "step": 20666
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3059291723338877e-05,
+      "loss": 0.2835,
+      "step": 20667
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.305895588803251e-05,
+      "loss": 0.2132,
+      "step": 20668
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3058620052726143e-05,
+      "loss": 0.3324,
+      "step": 20669
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.305828421741978e-05,
+      "loss": 0.3924,
+      "step": 20670
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3057948382113412e-05,
+      "loss": 0.3197,
+      "step": 20671
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3057612546807045e-05,
+      "loss": 0.26,
+      "step": 20672
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3057276711500681e-05,
+      "loss": 0.3882,
+      "step": 20673
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3056940876194314e-05,
+      "loss": 0.3214,
+      "step": 20674
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3056605040887949e-05,
+      "loss": 0.2694,
+      "step": 20675
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3056269205581584e-05,
+      "loss": 0.3018,
+      "step": 20676
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3055933370275218e-05,
+      "loss": 0.2511,
+      "step": 20677
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3055597534968851e-05,
+      "loss": 0.1582,
+      "step": 20678
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3055261699662488e-05,
+      "loss": 0.3149,
+      "step": 20679
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.305492586435612e-05,
+      "loss": 0.3328,
+      "step": 20680
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3054590029049754e-05,
+      "loss": 0.2842,
+      "step": 20681
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.305425419374339e-05,
+      "loss": 0.313,
+      "step": 20682
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3053918358437023e-05,
+      "loss": 0.4032,
+      "step": 20683
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3053582523130657e-05,
+      "loss": 0.3746,
+      "step": 20684
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3053246687824292e-05,
+      "loss": 0.3597,
+      "step": 20685
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3052910852517927e-05,
+      "loss": 0.3723,
+      "step": 20686
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.305257501721156e-05,
+      "loss": 0.2051,
+      "step": 20687
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3052239181905196e-05,
+      "loss": 0.2392,
+      "step": 20688
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3051903346598829e-05,
+      "loss": 0.3909,
+      "step": 20689
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3051567511292462e-05,
+      "loss": 0.2396,
+      "step": 20690
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3051231675986098e-05,
+      "loss": 0.2535,
+      "step": 20691
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3050895840679731e-05,
+      "loss": 0.3701,
+      "step": 20692
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3050560005373364e-05,
+      "loss": 0.2559,
+      "step": 20693
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3050224170067e-05,
+      "loss": 0.2305,
+      "step": 20694
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3049888334760633e-05,
+      "loss": 0.3674,
+      "step": 20695
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3049552499454268e-05,
+      "loss": 0.4168,
+      "step": 20696
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3049216664147903e-05,
+      "loss": 0.2485,
+      "step": 20697
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3048880828841537e-05,
+      "loss": 0.3501,
+      "step": 20698
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.304854499353517e-05,
+      "loss": 0.2618,
+      "step": 20699
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3048209158228807e-05,
+      "loss": 0.3472,
+      "step": 20700
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.304787332292244e-05,
+      "loss": 0.3157,
+      "step": 20701
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3047537487616073e-05,
+      "loss": 0.2561,
+      "step": 20702
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3047201652309709e-05,
+      "loss": 0.347,
+      "step": 20703
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3046865817003342e-05,
+      "loss": 0.4047,
+      "step": 20704
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3046529981696977e-05,
+      "loss": 0.1947,
+      "step": 20705
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3046194146390611e-05,
+      "loss": 0.351,
+      "step": 20706
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3045858311084246e-05,
+      "loss": 0.2851,
+      "step": 20707
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3045522475777879e-05,
+      "loss": 0.3851,
+      "step": 20708
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3045186640471515e-05,
+      "loss": 0.2921,
+      "step": 20709
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3044850805165148e-05,
+      "loss": 0.1318,
+      "step": 20710
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3044514969858781e-05,
+      "loss": 0.3052,
+      "step": 20711
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3044179134552417e-05,
+      "loss": 0.271,
+      "step": 20712
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.304384329924605e-05,
+      "loss": 0.2506,
+      "step": 20713
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3043507463939683e-05,
+      "loss": 0.247,
+      "step": 20714
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.304317162863332e-05,
+      "loss": 0.3778,
+      "step": 20715
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3042835793326953e-05,
+      "loss": 0.2309,
+      "step": 20716
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3042499958020587e-05,
+      "loss": 0.3568,
+      "step": 20717
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3042164122714222e-05,
+      "loss": 0.3048,
+      "step": 20718
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3041828287407857e-05,
+      "loss": 0.3342,
+      "step": 20719
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.304149245210149e-05,
+      "loss": 0.291,
+      "step": 20720
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3041156616795126e-05,
+      "loss": 0.3194,
+      "step": 20721
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3040820781488759e-05,
+      "loss": 0.2619,
+      "step": 20722
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3040484946182392e-05,
+      "loss": 0.3151,
+      "step": 20723
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3040149110876028e-05,
+      "loss": 0.2169,
+      "step": 20724
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3039813275569661e-05,
+      "loss": 0.363,
+      "step": 20725
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3039477440263296e-05,
+      "loss": 0.3964,
+      "step": 20726
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.303914160495693e-05,
+      "loss": 0.3092,
+      "step": 20727
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3038805769650565e-05,
+      "loss": 0.2083,
+      "step": 20728
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3038469934344198e-05,
+      "loss": 0.2452,
+      "step": 20729
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3038134099037834e-05,
+      "loss": 0.2551,
+      "step": 20730
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3037798263731467e-05,
+      "loss": 0.3322,
+      "step": 20731
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.30374624284251e-05,
+      "loss": 0.2193,
+      "step": 20732
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3037126593118737e-05,
+      "loss": 0.2122,
+      "step": 20733
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.303679075781237e-05,
+      "loss": 0.2357,
+      "step": 20734
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3036454922506004e-05,
+      "loss": 0.2039,
+      "step": 20735
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3036119087199639e-05,
+      "loss": 0.238,
+      "step": 20736
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3035783251893272e-05,
+      "loss": 0.3711,
+      "step": 20737
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3035447416586906e-05,
+      "loss": 0.4257,
+      "step": 20738
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3035111581280541e-05,
+      "loss": 0.3775,
+      "step": 20739
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3034775745974176e-05,
+      "loss": 0.3345,
+      "step": 20740
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3034439910667809e-05,
+      "loss": 0.4318,
+      "step": 20741
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3034104075361445e-05,
+      "loss": 0.3006,
+      "step": 20742
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3033768240055078e-05,
+      "loss": 0.3502,
+      "step": 20743
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 1.3033432404748711e-05,
+      "loss": 0.2497,
+      "step": 20744
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3033096569442347e-05,
+      "loss": 0.3119,
+      "step": 20745
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.303276073413598e-05,
+      "loss": 0.4359,
+      "step": 20746
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3032424898829615e-05,
+      "loss": 0.1914,
+      "step": 20747
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.303208906352325e-05,
+      "loss": 0.3811,
+      "step": 20748
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3031753228216884e-05,
+      "loss": 0.4143,
+      "step": 20749
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3031417392910517e-05,
+      "loss": 0.1739,
+      "step": 20750
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3031081557604153e-05,
+      "loss": 0.2782,
+      "step": 20751
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3030745722297786e-05,
+      "loss": 0.336,
+      "step": 20752
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.303040988699142e-05,
+      "loss": 0.1926,
+      "step": 20753
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3030074051685056e-05,
+      "loss": 0.1175,
+      "step": 20754
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3029738216378689e-05,
+      "loss": 0.244,
+      "step": 20755
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3029402381072323e-05,
+      "loss": 0.3491,
+      "step": 20756
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3029066545765958e-05,
+      "loss": 0.1953,
+      "step": 20757
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3028730710459591e-05,
+      "loss": 0.3198,
+      "step": 20758
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3028394875153226e-05,
+      "loss": 0.2706,
+      "step": 20759
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.302805903984686e-05,
+      "loss": 0.2222,
+      "step": 20760
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3027723204540495e-05,
+      "loss": 0.3642,
+      "step": 20761
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3027387369234128e-05,
+      "loss": 0.2286,
+      "step": 20762
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3027051533927764e-05,
+      "loss": 0.4144,
+      "step": 20763
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3026715698621397e-05,
+      "loss": 0.2119,
+      "step": 20764
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.302637986331503e-05,
+      "loss": 0.3689,
+      "step": 20765
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3026044028008666e-05,
+      "loss": 0.3284,
+      "step": 20766
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.30257081927023e-05,
+      "loss": 0.2201,
+      "step": 20767
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3025372357395934e-05,
+      "loss": 0.281,
+      "step": 20768
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3025036522089569e-05,
+      "loss": 0.3476,
+      "step": 20769
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3024700686783203e-05,
+      "loss": 0.2954,
+      "step": 20770
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3024364851476836e-05,
+      "loss": 0.3083,
+      "step": 20771
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3024029016170473e-05,
+      "loss": 0.2563,
+      "step": 20772
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3023693180864105e-05,
+      "loss": 0.215,
+      "step": 20773
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3023357345557738e-05,
+      "loss": 0.3159,
+      "step": 20774
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3023021510251375e-05,
+      "loss": 0.2652,
+      "step": 20775
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3022685674945008e-05,
+      "loss": 0.333,
+      "step": 20776
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3022349839638642e-05,
+      "loss": 0.2297,
+      "step": 20777
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3022014004332277e-05,
+      "loss": 0.243,
+      "step": 20778
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3021678169025912e-05,
+      "loss": 0.2185,
+      "step": 20779
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3021342333719545e-05,
+      "loss": 0.2608,
+      "step": 20780
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.302100649841318e-05,
+      "loss": 0.3885,
+      "step": 20781
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3020670663106814e-05,
+      "loss": 0.3166,
+      "step": 20782
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3020334827800447e-05,
+      "loss": 0.366,
+      "step": 20783
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3019998992494083e-05,
+      "loss": 0.1214,
+      "step": 20784
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3019663157187716e-05,
+      "loss": 0.2555,
+      "step": 20785
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3019327321881349e-05,
+      "loss": 0.2809,
+      "step": 20786
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3018991486574985e-05,
+      "loss": 0.3328,
+      "step": 20787
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3018655651268618e-05,
+      "loss": 0.4005,
+      "step": 20788
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3018319815962253e-05,
+      "loss": 0.3711,
+      "step": 20789
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3017983980655888e-05,
+      "loss": 0.2794,
+      "step": 20790
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3017648145349522e-05,
+      "loss": 0.2932,
+      "step": 20791
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3017312310043155e-05,
+      "loss": 0.275,
+      "step": 20792
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3016976474736792e-05,
+      "loss": 0.3422,
+      "step": 20793
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3016640639430425e-05,
+      "loss": 0.2445,
+      "step": 20794
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3016304804124058e-05,
+      "loss": 0.3446,
+      "step": 20795
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3015968968817694e-05,
+      "loss": 0.2817,
+      "step": 20796
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3015633133511327e-05,
+      "loss": 0.2034,
+      "step": 20797
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3015297298204962e-05,
+      "loss": 0.3418,
+      "step": 20798
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3014961462898596e-05,
+      "loss": 0.3337,
+      "step": 20799
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.301462562759223e-05,
+      "loss": 0.1952,
+      "step": 20800
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3014289792285864e-05,
+      "loss": 0.264,
+      "step": 20801
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.30139539569795e-05,
+      "loss": 0.2656,
+      "step": 20802
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3013618121673133e-05,
+      "loss": 0.1321,
+      "step": 20803
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3013282286366766e-05,
+      "loss": 0.3965,
+      "step": 20804
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3012946451060402e-05,
+      "loss": 0.3072,
+      "step": 20805
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3012610615754035e-05,
+      "loss": 0.4109,
+      "step": 20806
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3012274780447668e-05,
+      "loss": 0.2395,
+      "step": 20807
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3011938945141305e-05,
+      "loss": 0.3153,
+      "step": 20808
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3011603109834938e-05,
+      "loss": 0.2661,
+      "step": 20809
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3011267274528572e-05,
+      "loss": 0.2801,
+      "step": 20810
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3010931439222207e-05,
+      "loss": 0.1967,
+      "step": 20811
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3010595603915841e-05,
+      "loss": 0.1964,
+      "step": 20812
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3010259768609474e-05,
+      "loss": 0.4229,
+      "step": 20813
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.300992393330311e-05,
+      "loss": 0.2169,
+      "step": 20814
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3009588097996744e-05,
+      "loss": 0.497,
+      "step": 20815
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3009252262690377e-05,
+      "loss": 0.3034,
+      "step": 20816
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3008916427384013e-05,
+      "loss": 0.2025,
+      "step": 20817
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3008580592077646e-05,
+      "loss": 0.225,
+      "step": 20818
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.300824475677128e-05,
+      "loss": 0.3954,
+      "step": 20819
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3007908921464915e-05,
+      "loss": 0.6011,
+      "step": 20820
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.300757308615855e-05,
+      "loss": 0.2733,
+      "step": 20821
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3007237250852183e-05,
+      "loss": 0.2916,
+      "step": 20822
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.300690141554582e-05,
+      "loss": 0.2587,
+      "step": 20823
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3006565580239452e-05,
+      "loss": 0.2508,
+      "step": 20824
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3006229744933085e-05,
+      "loss": 0.3835,
+      "step": 20825
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3005893909626721e-05,
+      "loss": 0.3152,
+      "step": 20826
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3005558074320354e-05,
+      "loss": 0.2267,
+      "step": 20827
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3005222239013987e-05,
+      "loss": 0.3098,
+      "step": 20828
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3004886403707624e-05,
+      "loss": 0.2876,
+      "step": 20829
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3004550568401257e-05,
+      "loss": 0.3104,
+      "step": 20830
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3004214733094891e-05,
+      "loss": 0.1745,
+      "step": 20831
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3003878897788526e-05,
+      "loss": 0.4241,
+      "step": 20832
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.300354306248216e-05,
+      "loss": 0.2987,
+      "step": 20833
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3003207227175794e-05,
+      "loss": 0.24,
+      "step": 20834
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.300287139186943e-05,
+      "loss": 0.2144,
+      "step": 20835
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3002535556563063e-05,
+      "loss": 0.1581,
+      "step": 20836
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3002199721256696e-05,
+      "loss": 0.3684,
+      "step": 20837
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3001863885950332e-05,
+      "loss": 0.3665,
+      "step": 20838
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3001528050643965e-05,
+      "loss": 0.3947,
+      "step": 20839
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.30011922153376e-05,
+      "loss": 0.3038,
+      "step": 20840
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3000856380031234e-05,
+      "loss": 0.5178,
+      "step": 20841
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3000520544724869e-05,
+      "loss": 0.3434,
+      "step": 20842
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.3000184709418502e-05,
+      "loss": 0.2236,
+      "step": 20843
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2999848874112135e-05,
+      "loss": 0.2763,
+      "step": 20844
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2999513038805771e-05,
+      "loss": 0.2841,
+      "step": 20845
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2999177203499404e-05,
+      "loss": 0.2448,
+      "step": 20846
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2998841368193039e-05,
+      "loss": 0.1653,
+      "step": 20847
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2998505532886674e-05,
+      "loss": 0.3377,
+      "step": 20848
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2998169697580307e-05,
+      "loss": 0.273,
+      "step": 20849
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2997833862273941e-05,
+      "loss": 0.2777,
+      "step": 20850
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2997498026967576e-05,
+      "loss": 0.3178,
+      "step": 20851
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.299716219166121e-05,
+      "loss": 0.4065,
+      "step": 20852
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2996826356354843e-05,
+      "loss": 0.2002,
+      "step": 20853
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.299649052104848e-05,
+      "loss": 0.374,
+      "step": 20854
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2996154685742113e-05,
+      "loss": 0.1902,
+      "step": 20855
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2995818850435746e-05,
+      "loss": 0.2248,
+      "step": 20856
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2995483015129382e-05,
+      "loss": 0.3844,
+      "step": 20857
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2995147179823015e-05,
+      "loss": 0.1204,
+      "step": 20858
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.299481134451665e-05,
+      "loss": 0.4635,
+      "step": 20859
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2994475509210284e-05,
+      "loss": 0.3546,
+      "step": 20860
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2994139673903919e-05,
+      "loss": 0.1832,
+      "step": 20861
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2993803838597552e-05,
+      "loss": 0.332,
+      "step": 20862
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2993468003291188e-05,
+      "loss": 0.3291,
+      "step": 20863
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2993132167984821e-05,
+      "loss": 0.3042,
+      "step": 20864
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2992796332678454e-05,
+      "loss": 0.2456,
+      "step": 20865
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.299246049737209e-05,
+      "loss": 0.2915,
+      "step": 20866
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2992124662065723e-05,
+      "loss": 0.2904,
+      "step": 20867
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2991788826759358e-05,
+      "loss": 0.3039,
+      "step": 20868
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2991452991452993e-05,
+      "loss": 0.2084,
+      "step": 20869
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2991117156146627e-05,
+      "loss": 0.3148,
+      "step": 20870
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.299078132084026e-05,
+      "loss": 0.2439,
+      "step": 20871
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2990445485533895e-05,
+      "loss": 0.3121,
+      "step": 20872
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.299010965022753e-05,
+      "loss": 0.3248,
+      "step": 20873
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2989773814921163e-05,
+      "loss": 0.2094,
+      "step": 20874
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2989437979614799e-05,
+      "loss": 0.1765,
+      "step": 20875
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2989102144308432e-05,
+      "loss": 0.2956,
+      "step": 20876
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2988766309002065e-05,
+      "loss": 0.2664,
+      "step": 20877
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2988430473695701e-05,
+      "loss": 0.283,
+      "step": 20878
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2988094638389334e-05,
+      "loss": 0.2472,
+      "step": 20879
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2987758803082969e-05,
+      "loss": 0.2746,
+      "step": 20880
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2987422967776603e-05,
+      "loss": 0.3557,
+      "step": 20881
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2987087132470238e-05,
+      "loss": 0.2263,
+      "step": 20882
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2986751297163871e-05,
+      "loss": 0.3335,
+      "step": 20883
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2986415461857507e-05,
+      "loss": 0.2759,
+      "step": 20884
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.298607962655114e-05,
+      "loss": 0.1963,
+      "step": 20885
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2985743791244773e-05,
+      "loss": 0.4264,
+      "step": 20886
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.298540795593841e-05,
+      "loss": 0.4322,
+      "step": 20887
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2985072120632043e-05,
+      "loss": 0.3615,
+      "step": 20888
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2984736285325677e-05,
+      "loss": 0.1835,
+      "step": 20889
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2984400450019312e-05,
+      "loss": 0.2092,
+      "step": 20890
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2984064614712946e-05,
+      "loss": 0.3857,
+      "step": 20891
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.298372877940658e-05,
+      "loss": 0.2771,
+      "step": 20892
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2983392944100214e-05,
+      "loss": 0.268,
+      "step": 20893
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2983057108793849e-05,
+      "loss": 0.3801,
+      "step": 20894
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2982721273487482e-05,
+      "loss": 0.4339,
+      "step": 20895
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2982385438181118e-05,
+      "loss": 0.2547,
+      "step": 20896
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2982049602874751e-05,
+      "loss": 0.2733,
+      "step": 20897
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2981713767568384e-05,
+      "loss": 0.214,
+      "step": 20898
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.298137793226202e-05,
+      "loss": 0.2178,
+      "step": 20899
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2981042096955653e-05,
+      "loss": 0.2008,
+      "step": 20900
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2980706261649288e-05,
+      "loss": 0.2328,
+      "step": 20901
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2980370426342922e-05,
+      "loss": 0.2351,
+      "step": 20902
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2980034591036557e-05,
+      "loss": 0.3831,
+      "step": 20903
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.297969875573019e-05,
+      "loss": 0.2771,
+      "step": 20904
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2979362920423826e-05,
+      "loss": 0.2303,
+      "step": 20905
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.297902708511746e-05,
+      "loss": 0.1452,
+      "step": 20906
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2978691249811092e-05,
+      "loss": 0.3114,
+      "step": 20907
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2978355414504729e-05,
+      "loss": 0.345,
+      "step": 20908
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2978019579198362e-05,
+      "loss": 0.3505,
+      "step": 20909
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2977683743891996e-05,
+      "loss": 0.3009,
+      "step": 20910
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2977347908585631e-05,
+      "loss": 0.2686,
+      "step": 20911
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2977012073279266e-05,
+      "loss": 0.4429,
+      "step": 20912
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2976676237972899e-05,
+      "loss": 0.205,
+      "step": 20913
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2976340402666535e-05,
+      "loss": 0.3163,
+      "step": 20914
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2976004567360168e-05,
+      "loss": 0.5598,
+      "step": 20915
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.29756687320538e-05,
+      "loss": 0.1546,
+      "step": 20916
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2975332896747437e-05,
+      "loss": 0.3408,
+      "step": 20917
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.297499706144107e-05,
+      "loss": 0.2358,
+      "step": 20918
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2974661226134703e-05,
+      "loss": 0.1948,
+      "step": 20919
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.297432539082834e-05,
+      "loss": 0.2756,
+      "step": 20920
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2973989555521972e-05,
+      "loss": 0.2001,
+      "step": 20921
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2973653720215607e-05,
+      "loss": 0.3481,
+      "step": 20922
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2973317884909242e-05,
+      "loss": 0.2317,
+      "step": 20923
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2972982049602876e-05,
+      "loss": 0.2763,
+      "step": 20924
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.297264621429651e-05,
+      "loss": 0.1765,
+      "step": 20925
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2972310378990146e-05,
+      "loss": 0.2446,
+      "step": 20926
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2971974543683779e-05,
+      "loss": 0.3061,
+      "step": 20927
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2971638708377411e-05,
+      "loss": 0.3129,
+      "step": 20928
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2971302873071048e-05,
+      "loss": 0.3087,
+      "step": 20929
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.297096703776468e-05,
+      "loss": 0.3972,
+      "step": 20930
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2970631202458315e-05,
+      "loss": 0.287,
+      "step": 20931
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.297029536715195e-05,
+      "loss": 0.3074,
+      "step": 20932
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2969959531845585e-05,
+      "loss": 0.3879,
+      "step": 20933
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2969623696539218e-05,
+      "loss": 0.3999,
+      "step": 20934
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2969287861232854e-05,
+      "loss": 0.2742,
+      "step": 20935
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2968952025926487e-05,
+      "loss": 0.289,
+      "step": 20936
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.296861619062012e-05,
+      "loss": 0.3419,
+      "step": 20937
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2968280355313756e-05,
+      "loss": 0.1713,
+      "step": 20938
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.296794452000739e-05,
+      "loss": 0.343,
+      "step": 20939
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2967608684701022e-05,
+      "loss": 0.3226,
+      "step": 20940
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2967272849394658e-05,
+      "loss": 0.3459,
+      "step": 20941
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 1.2966937014088291e-05,
+      "loss": 0.5117,
+      "step": 20942
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2966601178781926e-05,
+      "loss": 0.3326,
+      "step": 20943
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.296626534347556e-05,
+      "loss": 0.2741,
+      "step": 20944
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2965929508169195e-05,
+      "loss": 0.3651,
+      "step": 20945
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2965593672862828e-05,
+      "loss": 0.3083,
+      "step": 20946
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2965257837556465e-05,
+      "loss": 0.4027,
+      "step": 20947
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2964922002250098e-05,
+      "loss": 0.3357,
+      "step": 20948
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.296458616694373e-05,
+      "loss": 0.4579,
+      "step": 20949
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2964250331637367e-05,
+      "loss": 0.3928,
+      "step": 20950
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2963914496331e-05,
+      "loss": 0.3309,
+      "step": 20951
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2963578661024635e-05,
+      "loss": 0.374,
+      "step": 20952
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.296324282571827e-05,
+      "loss": 0.2756,
+      "step": 20953
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2962906990411904e-05,
+      "loss": 0.4323,
+      "step": 20954
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2962571155105537e-05,
+      "loss": 0.2203,
+      "step": 20955
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2962235319799173e-05,
+      "loss": 0.2703,
+      "step": 20956
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2961899484492806e-05,
+      "loss": 0.3882,
+      "step": 20957
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2961563649186439e-05,
+      "loss": 0.248,
+      "step": 20958
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2961227813880075e-05,
+      "loss": 0.2355,
+      "step": 20959
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2960891978573708e-05,
+      "loss": 0.4637,
+      "step": 20960
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2960556143267343e-05,
+      "loss": 0.2446,
+      "step": 20961
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2960220307960978e-05,
+      "loss": 0.4172,
+      "step": 20962
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.295988447265461e-05,
+      "loss": 0.2699,
+      "step": 20963
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2959548637348245e-05,
+      "loss": 0.1951,
+      "step": 20964
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.295921280204188e-05,
+      "loss": 0.2101,
+      "step": 20965
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2958876966735515e-05,
+      "loss": 0.2396,
+      "step": 20966
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2958541131429147e-05,
+      "loss": 0.3426,
+      "step": 20967
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2958205296122784e-05,
+      "loss": 0.4065,
+      "step": 20968
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2957869460816417e-05,
+      "loss": 0.1293,
+      "step": 20969
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.295753362551005e-05,
+      "loss": 0.3746,
+      "step": 20970
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2957197790203686e-05,
+      "loss": 0.1771,
+      "step": 20971
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2956861954897319e-05,
+      "loss": 0.3336,
+      "step": 20972
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2956526119590954e-05,
+      "loss": 0.2133,
+      "step": 20973
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2956190284284588e-05,
+      "loss": 0.3911,
+      "step": 20974
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2955854448978223e-05,
+      "loss": 0.2386,
+      "step": 20975
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2955518613671856e-05,
+      "loss": 0.165,
+      "step": 20976
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2955182778365492e-05,
+      "loss": 0.3124,
+      "step": 20977
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2954846943059125e-05,
+      "loss": 0.1991,
+      "step": 20978
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2954511107752758e-05,
+      "loss": 0.2412,
+      "step": 20979
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2954175272446394e-05,
+      "loss": 0.3087,
+      "step": 20980
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2953839437140027e-05,
+      "loss": 0.3368,
+      "step": 20981
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2953503601833662e-05,
+      "loss": 0.1721,
+      "step": 20982
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2953167766527297e-05,
+      "loss": 0.1594,
+      "step": 20983
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.295283193122093e-05,
+      "loss": 0.2666,
+      "step": 20984
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2952496095914564e-05,
+      "loss": 0.2828,
+      "step": 20985
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2952160260608199e-05,
+      "loss": 0.1643,
+      "step": 20986
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2951824425301834e-05,
+      "loss": 0.4699,
+      "step": 20987
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2951488589995467e-05,
+      "loss": 0.3807,
+      "step": 20988
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2951152754689103e-05,
+      "loss": 0.4591,
+      "step": 20989
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2950816919382736e-05,
+      "loss": 0.1463,
+      "step": 20990
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2950481084076369e-05,
+      "loss": 0.2352,
+      "step": 20991
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2950145248770005e-05,
+      "loss": 0.2697,
+      "step": 20992
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2949809413463638e-05,
+      "loss": 0.2279,
+      "step": 20993
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2949473578157273e-05,
+      "loss": 0.107,
+      "step": 20994
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2949137742850907e-05,
+      "loss": 0.2474,
+      "step": 20995
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2948801907544542e-05,
+      "loss": 0.3389,
+      "step": 20996
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2948466072238175e-05,
+      "loss": 0.4821,
+      "step": 20997
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2948130236931811e-05,
+      "loss": 0.3211,
+      "step": 20998
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2947794401625444e-05,
+      "loss": 0.1353,
+      "step": 20999
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2947458566319077e-05,
+      "loss": 0.3757,
+      "step": 21000
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2947122731012714e-05,
+      "loss": 0.2325,
+      "step": 21001
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2946786895706347e-05,
+      "loss": 0.3551,
+      "step": 21002
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2946451060399981e-05,
+      "loss": 0.3331,
+      "step": 21003
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2946115225093616e-05,
+      "loss": 0.2601,
+      "step": 21004
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.294577938978725e-05,
+      "loss": 0.2484,
+      "step": 21005
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2945443554480883e-05,
+      "loss": 0.2824,
+      "step": 21006
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2945107719174518e-05,
+      "loss": 0.4109,
+      "step": 21007
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2944771883868153e-05,
+      "loss": 0.1093,
+      "step": 21008
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2944436048561786e-05,
+      "loss": 0.2145,
+      "step": 21009
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2944100213255422e-05,
+      "loss": 0.2824,
+      "step": 21010
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2943764377949055e-05,
+      "loss": 0.3613,
+      "step": 21011
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2943428542642688e-05,
+      "loss": 0.3044,
+      "step": 21012
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2943092707336324e-05,
+      "loss": 0.186,
+      "step": 21013
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2942756872029957e-05,
+      "loss": 0.3604,
+      "step": 21014
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2942421036723592e-05,
+      "loss": 0.4572,
+      "step": 21015
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2942085201417227e-05,
+      "loss": 0.2354,
+      "step": 21016
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2941749366110861e-05,
+      "loss": 0.3923,
+      "step": 21017
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2941413530804494e-05,
+      "loss": 0.3986,
+      "step": 21018
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.294107769549813e-05,
+      "loss": 0.2857,
+      "step": 21019
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2940741860191763e-05,
+      "loss": 0.4177,
+      "step": 21020
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2940406024885396e-05,
+      "loss": 0.1838,
+      "step": 21021
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2940070189579033e-05,
+      "loss": 0.3398,
+      "step": 21022
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2939734354272666e-05,
+      "loss": 0.4466,
+      "step": 21023
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.29393985189663e-05,
+      "loss": 0.3267,
+      "step": 21024
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2939062683659933e-05,
+      "loss": 0.2577,
+      "step": 21025
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.293872684835357e-05,
+      "loss": 0.2998,
+      "step": 21026
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2938391013047203e-05,
+      "loss": 0.3248,
+      "step": 21027
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2938055177740836e-05,
+      "loss": 0.1911,
+      "step": 21028
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2937719342434472e-05,
+      "loss": 0.2043,
+      "step": 21029
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2937383507128105e-05,
+      "loss": 0.3311,
+      "step": 21030
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2937047671821738e-05,
+      "loss": 0.321,
+      "step": 21031
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2936711836515374e-05,
+      "loss": 0.261,
+      "step": 21032
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2936376001209007e-05,
+      "loss": 0.2081,
+      "step": 21033
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2936040165902642e-05,
+      "loss": 0.2061,
+      "step": 21034
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2935704330596276e-05,
+      "loss": 0.1967,
+      "step": 21035
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2935368495289911e-05,
+      "loss": 0.3296,
+      "step": 21036
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2935032659983544e-05,
+      "loss": 0.1809,
+      "step": 21037
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.293469682467718e-05,
+      "loss": 0.2408,
+      "step": 21038
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2934360989370813e-05,
+      "loss": 0.249,
+      "step": 21039
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2934025154064446e-05,
+      "loss": 0.4913,
+      "step": 21040
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2933689318758083e-05,
+      "loss": 0.4251,
+      "step": 21041
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2933353483451716e-05,
+      "loss": 0.2861,
+      "step": 21042
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.293301764814535e-05,
+      "loss": 0.1891,
+      "step": 21043
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2932681812838985e-05,
+      "loss": 0.1555,
+      "step": 21044
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.293234597753262e-05,
+      "loss": 0.2014,
+      "step": 21045
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2932010142226252e-05,
+      "loss": 0.1799,
+      "step": 21046
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2931674306919889e-05,
+      "loss": 0.274,
+      "step": 21047
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2931338471613522e-05,
+      "loss": 0.3681,
+      "step": 21048
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2931002636307155e-05,
+      "loss": 0.3746,
+      "step": 21049
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2930666801000791e-05,
+      "loss": 0.2735,
+      "step": 21050
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2930330965694424e-05,
+      "loss": 0.1663,
+      "step": 21051
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2929995130388057e-05,
+      "loss": 0.294,
+      "step": 21052
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2929659295081693e-05,
+      "loss": 0.2447,
+      "step": 21053
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2929323459775326e-05,
+      "loss": 0.1897,
+      "step": 21054
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2928987624468961e-05,
+      "loss": 0.3035,
+      "step": 21055
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2928651789162596e-05,
+      "loss": 0.396,
+      "step": 21056
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.292831595385623e-05,
+      "loss": 0.3387,
+      "step": 21057
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2927980118549863e-05,
+      "loss": 0.454,
+      "step": 21058
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.29276442832435e-05,
+      "loss": 0.2648,
+      "step": 21059
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2927308447937132e-05,
+      "loss": 0.3283,
+      "step": 21060
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2926972612630765e-05,
+      "loss": 0.2593,
+      "step": 21061
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2926636777324402e-05,
+      "loss": 0.2949,
+      "step": 21062
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2926300942018035e-05,
+      "loss": 0.2832,
+      "step": 21063
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.292596510671167e-05,
+      "loss": 0.2364,
+      "step": 21064
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2925629271405304e-05,
+      "loss": 0.2073,
+      "step": 21065
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2925293436098939e-05,
+      "loss": 0.3243,
+      "step": 21066
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2924957600792572e-05,
+      "loss": 0.227,
+      "step": 21067
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2924621765486208e-05,
+      "loss": 0.2395,
+      "step": 21068
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.292428593017984e-05,
+      "loss": 0.3953,
+      "step": 21069
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2923950094873474e-05,
+      "loss": 0.2794,
+      "step": 21070
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.292361425956711e-05,
+      "loss": 0.3116,
+      "step": 21071
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2923278424260743e-05,
+      "loss": 0.4083,
+      "step": 21072
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2922942588954378e-05,
+      "loss": 0.2128,
+      "step": 21073
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2922606753648012e-05,
+      "loss": 0.308,
+      "step": 21074
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2922270918341645e-05,
+      "loss": 0.2567,
+      "step": 21075
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.292193508303528e-05,
+      "loss": 0.2361,
+      "step": 21076
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2921599247728915e-05,
+      "loss": 0.4003,
+      "step": 21077
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.292126341242255e-05,
+      "loss": 0.303,
+      "step": 21078
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2920927577116182e-05,
+      "loss": 0.3477,
+      "step": 21079
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2920591741809819e-05,
+      "loss": 0.2246,
+      "step": 21080
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2920255906503452e-05,
+      "loss": 0.5365,
+      "step": 21081
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2919920071197084e-05,
+      "loss": 0.2025,
+      "step": 21082
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.291958423589072e-05,
+      "loss": 0.2135,
+      "step": 21083
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2919248400584354e-05,
+      "loss": 0.2656,
+      "step": 21084
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2918912565277988e-05,
+      "loss": 0.2314,
+      "step": 21085
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2918576729971623e-05,
+      "loss": 0.2575,
+      "step": 21086
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2918240894665258e-05,
+      "loss": 0.3993,
+      "step": 21087
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.291790505935889e-05,
+      "loss": 0.2901,
+      "step": 21088
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2917569224052527e-05,
+      "loss": 0.2274,
+      "step": 21089
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.291723338874616e-05,
+      "loss": 0.4476,
+      "step": 21090
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2916897553439793e-05,
+      "loss": 0.1586,
+      "step": 21091
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.291656171813343e-05,
+      "loss": 0.3163,
+      "step": 21092
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2916225882827062e-05,
+      "loss": 0.2819,
+      "step": 21093
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2915890047520697e-05,
+      "loss": 0.3749,
+      "step": 21094
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2915554212214332e-05,
+      "loss": 0.4895,
+      "step": 21095
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2915218376907966e-05,
+      "loss": 0.3521,
+      "step": 21096
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2914882541601599e-05,
+      "loss": 0.308,
+      "step": 21097
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2914546706295234e-05,
+      "loss": 0.2968,
+      "step": 21098
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2914210870988868e-05,
+      "loss": 0.238,
+      "step": 21099
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2913875035682501e-05,
+      "loss": 0.24,
+      "step": 21100
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2913539200376138e-05,
+      "loss": 0.33,
+      "step": 21101
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.291320336506977e-05,
+      "loss": 0.306,
+      "step": 21102
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2912867529763404e-05,
+      "loss": 0.295,
+      "step": 21103
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.291253169445704e-05,
+      "loss": 0.2249,
+      "step": 21104
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2912195859150673e-05,
+      "loss": 0.1616,
+      "step": 21105
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2911860023844308e-05,
+      "loss": 0.332,
+      "step": 21106
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2911524188537942e-05,
+      "loss": 0.3499,
+      "step": 21107
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2911188353231577e-05,
+      "loss": 0.3056,
+      "step": 21108
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.291085251792521e-05,
+      "loss": 0.1881,
+      "step": 21109
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2910516682618846e-05,
+      "loss": 0.1852,
+      "step": 21110
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2910180847312479e-05,
+      "loss": 0.297,
+      "step": 21111
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2909845012006112e-05,
+      "loss": 0.1969,
+      "step": 21112
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2909509176699748e-05,
+      "loss": 0.3104,
+      "step": 21113
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2909173341393381e-05,
+      "loss": 0.3019,
+      "step": 21114
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2908837506087016e-05,
+      "loss": 0.3659,
+      "step": 21115
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.290850167078065e-05,
+      "loss": 0.2977,
+      "step": 21116
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2908165835474285e-05,
+      "loss": 0.2511,
+      "step": 21117
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2907830000167918e-05,
+      "loss": 0.2893,
+      "step": 21118
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2907494164861553e-05,
+      "loss": 0.3405,
+      "step": 21119
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2907158329555188e-05,
+      "loss": 0.3172,
+      "step": 21120
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.290682249424882e-05,
+      "loss": 0.3132,
+      "step": 21121
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2906486658942457e-05,
+      "loss": 0.2348,
+      "step": 21122
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.290615082363609e-05,
+      "loss": 0.2734,
+      "step": 21123
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2905814988329723e-05,
+      "loss": 0.3926,
+      "step": 21124
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2905479153023359e-05,
+      "loss": 0.2536,
+      "step": 21125
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2905143317716992e-05,
+      "loss": 0.223,
+      "step": 21126
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2904807482410627e-05,
+      "loss": 0.3529,
+      "step": 21127
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2904471647104261e-05,
+      "loss": 0.4334,
+      "step": 21128
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2904135811797896e-05,
+      "loss": 0.3524,
+      "step": 21129
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2903799976491529e-05,
+      "loss": 0.2356,
+      "step": 21130
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2903464141185165e-05,
+      "loss": 0.277,
+      "step": 21131
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2903128305878798e-05,
+      "loss": 0.2542,
+      "step": 21132
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2902792470572431e-05,
+      "loss": 0.3225,
+      "step": 21133
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2902456635266068e-05,
+      "loss": 0.2175,
+      "step": 21134
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.29021207999597e-05,
+      "loss": 0.3116,
+      "step": 21135
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2901784964653335e-05,
+      "loss": 0.5347,
+      "step": 21136
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.290144912934697e-05,
+      "loss": 0.2256,
+      "step": 21137
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2901113294040604e-05,
+      "loss": 0.423,
+      "step": 21138
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2900777458734237e-05,
+      "loss": 0.3283,
+      "step": 21139
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2900441623427874e-05,
+      "loss": 0.2146,
+      "step": 21140
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.2900105788121507e-05,
+      "loss": 0.3912,
+      "step": 21141
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.289976995281514e-05,
+      "loss": 0.2218,
+      "step": 21142
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2899434117508776e-05,
+      "loss": 0.3611,
+      "step": 21143
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2899098282202409e-05,
+      "loss": 0.354,
+      "step": 21144
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2898762446896042e-05,
+      "loss": 0.4711,
+      "step": 21145
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2898426611589678e-05,
+      "loss": 0.4452,
+      "step": 21146
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2898090776283311e-05,
+      "loss": 0.2551,
+      "step": 21147
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2897754940976946e-05,
+      "loss": 0.3245,
+      "step": 21148
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.289741910567058e-05,
+      "loss": 0.229,
+      "step": 21149
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2897083270364215e-05,
+      "loss": 0.2332,
+      "step": 21150
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2896747435057848e-05,
+      "loss": 0.2624,
+      "step": 21151
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2896411599751484e-05,
+      "loss": 0.4155,
+      "step": 21152
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2896075764445117e-05,
+      "loss": 0.2384,
+      "step": 21153
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.289573992913875e-05,
+      "loss": 0.3973,
+      "step": 21154
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2895404093832387e-05,
+      "loss": 0.3581,
+      "step": 21155
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.289506825852602e-05,
+      "loss": 0.2353,
+      "step": 21156
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2894732423219654e-05,
+      "loss": 0.3466,
+      "step": 21157
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2894396587913289e-05,
+      "loss": 0.3944,
+      "step": 21158
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2894060752606924e-05,
+      "loss": 0.3353,
+      "step": 21159
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2893724917300556e-05,
+      "loss": 0.3435,
+      "step": 21160
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2893389081994193e-05,
+      "loss": 0.1687,
+      "step": 21161
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2893053246687826e-05,
+      "loss": 0.3767,
+      "step": 21162
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2892717411381459e-05,
+      "loss": 0.3289,
+      "step": 21163
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2892381576075095e-05,
+      "loss": 0.2686,
+      "step": 21164
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2892045740768728e-05,
+      "loss": 0.2125,
+      "step": 21165
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2891709905462361e-05,
+      "loss": 0.3745,
+      "step": 21166
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2891374070155997e-05,
+      "loss": 0.454,
+      "step": 21167
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.289103823484963e-05,
+      "loss": 0.3668,
+      "step": 21168
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2890702399543265e-05,
+      "loss": 0.4071,
+      "step": 21169
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.28903665642369e-05,
+      "loss": 0.212,
+      "step": 21170
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2890030728930534e-05,
+      "loss": 0.1914,
+      "step": 21171
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2889694893624167e-05,
+      "loss": 0.2667,
+      "step": 21172
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2889359058317804e-05,
+      "loss": 0.4608,
+      "step": 21173
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2889023223011436e-05,
+      "loss": 0.3665,
+      "step": 21174
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.288868738770507e-05,
+      "loss": 0.3098,
+      "step": 21175
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2888351552398706e-05,
+      "loss": 0.3463,
+      "step": 21176
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2888015717092339e-05,
+      "loss": 0.3785,
+      "step": 21177
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2887679881785973e-05,
+      "loss": 0.1874,
+      "step": 21178
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2887344046479608e-05,
+      "loss": 0.433,
+      "step": 21179
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2887008211173243e-05,
+      "loss": 0.3185,
+      "step": 21180
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2886672375866876e-05,
+      "loss": 0.265,
+      "step": 21181
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2886336540560512e-05,
+      "loss": 0.3035,
+      "step": 21182
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2886000705254145e-05,
+      "loss": 0.4924,
+      "step": 21183
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2885664869947778e-05,
+      "loss": 0.3098,
+      "step": 21184
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2885329034641414e-05,
+      "loss": 0.2065,
+      "step": 21185
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2884993199335047e-05,
+      "loss": 0.2773,
+      "step": 21186
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.288465736402868e-05,
+      "loss": 0.3321,
+      "step": 21187
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2884321528722316e-05,
+      "loss": 0.3853,
+      "step": 21188
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.288398569341595e-05,
+      "loss": 0.4033,
+      "step": 21189
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2883649858109584e-05,
+      "loss": 0.19,
+      "step": 21190
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2883314022803219e-05,
+      "loss": 0.3927,
+      "step": 21191
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2882978187496853e-05,
+      "loss": 0.4287,
+      "step": 21192
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2882642352190486e-05,
+      "loss": 0.2312,
+      "step": 21193
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2882306516884123e-05,
+      "loss": 0.4363,
+      "step": 21194
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2881970681577756e-05,
+      "loss": 0.2295,
+      "step": 21195
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2881634846271389e-05,
+      "loss": 0.2795,
+      "step": 21196
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2881299010965025e-05,
+      "loss": 0.3073,
+      "step": 21197
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2880963175658658e-05,
+      "loss": 0.2801,
+      "step": 21198
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2880627340352292e-05,
+      "loss": 0.3885,
+      "step": 21199
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2880291505045927e-05,
+      "loss": 0.3758,
+      "step": 21200
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2879955669739562e-05,
+      "loss": 0.2723,
+      "step": 21201
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2879619834433195e-05,
+      "loss": 0.2057,
+      "step": 21202
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2879283999126831e-05,
+      "loss": 0.2368,
+      "step": 21203
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2878948163820464e-05,
+      "loss": 0.2605,
+      "step": 21204
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2878612328514097e-05,
+      "loss": 0.2298,
+      "step": 21205
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2878276493207732e-05,
+      "loss": 0.382,
+      "step": 21206
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2877940657901366e-05,
+      "loss": 0.3469,
+      "step": 21207
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2877604822595001e-05,
+      "loss": 0.2239,
+      "step": 21208
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2877268987288634e-05,
+      "loss": 0.314,
+      "step": 21209
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2876933151982269e-05,
+      "loss": 0.3745,
+      "step": 21210
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2876597316675903e-05,
+      "loss": 0.2845,
+      "step": 21211
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2876261481369536e-05,
+      "loss": 0.2562,
+      "step": 21212
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2875925646063172e-05,
+      "loss": 0.4084,
+      "step": 21213
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2875589810756805e-05,
+      "loss": 0.2547,
+      "step": 21214
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2875253975450438e-05,
+      "loss": 0.2242,
+      "step": 21215
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2874918140144075e-05,
+      "loss": 0.2834,
+      "step": 21216
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2874582304837708e-05,
+      "loss": 0.2037,
+      "step": 21217
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2874246469531342e-05,
+      "loss": 0.313,
+      "step": 21218
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2873910634224977e-05,
+      "loss": 0.2988,
+      "step": 21219
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2873574798918612e-05,
+      "loss": 0.1957,
+      "step": 21220
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2873238963612245e-05,
+      "loss": 0.3312,
+      "step": 21221
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2872903128305881e-05,
+      "loss": 0.2334,
+      "step": 21222
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2872567292999514e-05,
+      "loss": 0.3801,
+      "step": 21223
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2872231457693147e-05,
+      "loss": 0.2939,
+      "step": 21224
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2871895622386783e-05,
+      "loss": 0.2903,
+      "step": 21225
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2871559787080416e-05,
+      "loss": 0.309,
+      "step": 21226
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.287122395177405e-05,
+      "loss": 0.2453,
+      "step": 21227
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2870888116467685e-05,
+      "loss": 0.4293,
+      "step": 21228
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.287055228116132e-05,
+      "loss": 0.361,
+      "step": 21229
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2870216445854953e-05,
+      "loss": 0.2432,
+      "step": 21230
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.286988061054859e-05,
+      "loss": 0.2519,
+      "step": 21231
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2869544775242222e-05,
+      "loss": 0.3185,
+      "step": 21232
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2869208939935855e-05,
+      "loss": 0.3118,
+      "step": 21233
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2868873104629492e-05,
+      "loss": 0.3077,
+      "step": 21234
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2868537269323125e-05,
+      "loss": 0.2411,
+      "step": 21235
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2868201434016757e-05,
+      "loss": 0.246,
+      "step": 21236
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2867865598710394e-05,
+      "loss": 0.2659,
+      "step": 21237
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2867529763404027e-05,
+      "loss": 0.3743,
+      "step": 21238
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2867193928097661e-05,
+      "loss": 0.2169,
+      "step": 21239
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2866858092791296e-05,
+      "loss": 0.1996,
+      "step": 21240
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.286652225748493e-05,
+      "loss": 0.3901,
+      "step": 21241
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2866186422178564e-05,
+      "loss": 0.2775,
+      "step": 21242
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.28658505868722e-05,
+      "loss": 0.3273,
+      "step": 21243
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2865514751565833e-05,
+      "loss": 0.4558,
+      "step": 21244
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2865178916259466e-05,
+      "loss": 0.2689,
+      "step": 21245
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2864843080953102e-05,
+      "loss": 0.1945,
+      "step": 21246
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2864507245646735e-05,
+      "loss": 0.1987,
+      "step": 21247
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.286417141034037e-05,
+      "loss": 0.2467,
+      "step": 21248
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2863835575034005e-05,
+      "loss": 0.1823,
+      "step": 21249
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.286349973972764e-05,
+      "loss": 0.3532,
+      "step": 21250
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2863163904421272e-05,
+      "loss": 0.2629,
+      "step": 21251
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2862828069114908e-05,
+      "loss": 0.2678,
+      "step": 21252
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2862492233808541e-05,
+      "loss": 0.2884,
+      "step": 21253
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2862156398502174e-05,
+      "loss": 0.3049,
+      "step": 21254
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.286182056319581e-05,
+      "loss": 0.3921,
+      "step": 21255
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2861484727889444e-05,
+      "loss": 0.5169,
+      "step": 21256
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2861148892583077e-05,
+      "loss": 0.3993,
+      "step": 21257
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2860813057276713e-05,
+      "loss": 0.3009,
+      "step": 21258
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2860477221970346e-05,
+      "loss": 0.2429,
+      "step": 21259
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.286014138666398e-05,
+      "loss": 0.4714,
+      "step": 21260
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2859805551357615e-05,
+      "loss": 0.2188,
+      "step": 21261
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.285946971605125e-05,
+      "loss": 0.2535,
+      "step": 21262
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2859133880744883e-05,
+      "loss": 0.3218,
+      "step": 21263
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2858798045438519e-05,
+      "loss": 0.2572,
+      "step": 21264
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2858462210132152e-05,
+      "loss": 0.2089,
+      "step": 21265
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2858126374825785e-05,
+      "loss": 0.1852,
+      "step": 21266
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2857790539519421e-05,
+      "loss": 0.2805,
+      "step": 21267
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2857454704213054e-05,
+      "loss": 0.2188,
+      "step": 21268
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2857118868906689e-05,
+      "loss": 0.27,
+      "step": 21269
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2856783033600324e-05,
+      "loss": 0.3224,
+      "step": 21270
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2856447198293958e-05,
+      "loss": 0.408,
+      "step": 21271
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2856111362987591e-05,
+      "loss": 0.2459,
+      "step": 21272
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2855775527681228e-05,
+      "loss": 0.3891,
+      "step": 21273
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.285543969237486e-05,
+      "loss": 0.1881,
+      "step": 21274
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2855103857068493e-05,
+      "loss": 0.2447,
+      "step": 21275
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.285476802176213e-05,
+      "loss": 0.3581,
+      "step": 21276
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2854432186455763e-05,
+      "loss": 0.1735,
+      "step": 21277
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2854096351149396e-05,
+      "loss": 0.3948,
+      "step": 21278
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2853760515843032e-05,
+      "loss": 0.1638,
+      "step": 21279
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2853424680536665e-05,
+      "loss": 0.1972,
+      "step": 21280
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.28530888452303e-05,
+      "loss": 0.2944,
+      "step": 21281
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2852753009923934e-05,
+      "loss": 0.303,
+      "step": 21282
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2852417174617569e-05,
+      "loss": 0.3295,
+      "step": 21283
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2852081339311202e-05,
+      "loss": 0.4915,
+      "step": 21284
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2851745504004838e-05,
+      "loss": 0.3101,
+      "step": 21285
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2851409668698471e-05,
+      "loss": 0.2414,
+      "step": 21286
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2851073833392104e-05,
+      "loss": 0.2975,
+      "step": 21287
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.285073799808574e-05,
+      "loss": 0.2241,
+      "step": 21288
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2850402162779373e-05,
+      "loss": 0.2663,
+      "step": 21289
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2850066327473008e-05,
+      "loss": 0.2064,
+      "step": 21290
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2849730492166643e-05,
+      "loss": 0.4308,
+      "step": 21291
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2849394656860277e-05,
+      "loss": 0.2209,
+      "step": 21292
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.284905882155391e-05,
+      "loss": 0.1771,
+      "step": 21293
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2848722986247547e-05,
+      "loss": 0.3572,
+      "step": 21294
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.284838715094118e-05,
+      "loss": 0.1474,
+      "step": 21295
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2848051315634813e-05,
+      "loss": 0.3755,
+      "step": 21296
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2847715480328449e-05,
+      "loss": 0.0811,
+      "step": 21297
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2847379645022082e-05,
+      "loss": 0.3131,
+      "step": 21298
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2847043809715717e-05,
+      "loss": 0.1965,
+      "step": 21299
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2846707974409351e-05,
+      "loss": 0.3552,
+      "step": 21300
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2846372139102984e-05,
+      "loss": 0.1921,
+      "step": 21301
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2846036303796619e-05,
+      "loss": 0.1885,
+      "step": 21302
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2845700468490253e-05,
+      "loss": 0.3736,
+      "step": 21303
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2845364633183888e-05,
+      "loss": 0.2009,
+      "step": 21304
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2845028797877521e-05,
+      "loss": 0.1487,
+      "step": 21305
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2844692962571157e-05,
+      "loss": 0.4394,
+      "step": 21306
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.284435712726479e-05,
+      "loss": 0.3001,
+      "step": 21307
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2844021291958423e-05,
+      "loss": 0.3949,
+      "step": 21308
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.284368545665206e-05,
+      "loss": 0.1752,
+      "step": 21309
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2843349621345693e-05,
+      "loss": 0.2913,
+      "step": 21310
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2843013786039327e-05,
+      "loss": 0.2354,
+      "step": 21311
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2842677950732962e-05,
+      "loss": 0.2569,
+      "step": 21312
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2842342115426597e-05,
+      "loss": 0.4059,
+      "step": 21313
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.284200628012023e-05,
+      "loss": 0.303,
+      "step": 21314
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2841670444813866e-05,
+      "loss": 0.2033,
+      "step": 21315
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2841334609507499e-05,
+      "loss": 0.2885,
+      "step": 21316
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2840998774201132e-05,
+      "loss": 0.3534,
+      "step": 21317
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2840662938894768e-05,
+      "loss": 0.3938,
+      "step": 21318
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2840327103588401e-05,
+      "loss": 0.3741,
+      "step": 21319
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2839991268282036e-05,
+      "loss": 0.2686,
+      "step": 21320
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.283965543297567e-05,
+      "loss": 0.3858,
+      "step": 21321
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2839319597669305e-05,
+      "loss": 0.2682,
+      "step": 21322
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2838983762362938e-05,
+      "loss": 0.1776,
+      "step": 21323
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2838647927056573e-05,
+      "loss": 0.3555,
+      "step": 21324
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2838312091750207e-05,
+      "loss": 0.2774,
+      "step": 21325
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.283797625644384e-05,
+      "loss": 0.3186,
+      "step": 21326
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2837640421137477e-05,
+      "loss": 0.1999,
+      "step": 21327
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.283730458583111e-05,
+      "loss": 0.2739,
+      "step": 21328
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2836968750524742e-05,
+      "loss": 0.4015,
+      "step": 21329
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2836632915218379e-05,
+      "loss": 0.2701,
+      "step": 21330
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2836297079912012e-05,
+      "loss": 0.4992,
+      "step": 21331
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2835961244605646e-05,
+      "loss": 0.3495,
+      "step": 21332
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2835625409299281e-05,
+      "loss": 0.5534,
+      "step": 21333
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2835289573992916e-05,
+      "loss": 0.1382,
+      "step": 21334
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2834953738686549e-05,
+      "loss": 0.2324,
+      "step": 21335
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2834617903380185e-05,
+      "loss": 0.454,
+      "step": 21336
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2834282068073818e-05,
+      "loss": 0.1832,
+      "step": 21337
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2833946232767451e-05,
+      "loss": 0.1834,
+      "step": 21338
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 1.2833610397461087e-05,
+      "loss": 0.2622,
+      "step": 21339
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.283327456215472e-05,
+      "loss": 0.263,
+      "step": 21340
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2832938726848355e-05,
+      "loss": 0.3497,
+      "step": 21341
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.283260289154199e-05,
+      "loss": 0.2018,
+      "step": 21342
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2832267056235624e-05,
+      "loss": 0.4574,
+      "step": 21343
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2831931220929257e-05,
+      "loss": 0.1443,
+      "step": 21344
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2831595385622892e-05,
+      "loss": 0.4177,
+      "step": 21345
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2831259550316526e-05,
+      "loss": 0.3478,
+      "step": 21346
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.283092371501016e-05,
+      "loss": 0.2613,
+      "step": 21347
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2830587879703796e-05,
+      "loss": 0.2632,
+      "step": 21348
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2830252044397429e-05,
+      "loss": 0.3129,
+      "step": 21349
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2829916209091062e-05,
+      "loss": 0.379,
+      "step": 21350
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2829580373784698e-05,
+      "loss": 0.6508,
+      "step": 21351
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2829244538478331e-05,
+      "loss": 0.3251,
+      "step": 21352
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2828908703171965e-05,
+      "loss": 0.2482,
+      "step": 21353
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.28285728678656e-05,
+      "loss": 0.2725,
+      "step": 21354
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2828237032559235e-05,
+      "loss": 0.288,
+      "step": 21355
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2827901197252868e-05,
+      "loss": 0.3975,
+      "step": 21356
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2827565361946504e-05,
+      "loss": 0.2185,
+      "step": 21357
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2827229526640137e-05,
+      "loss": 0.3656,
+      "step": 21358
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.282689369133377e-05,
+      "loss": 0.349,
+      "step": 21359
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2826557856027406e-05,
+      "loss": 0.4723,
+      "step": 21360
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.282622202072104e-05,
+      "loss": 0.2905,
+      "step": 21361
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2825886185414674e-05,
+      "loss": 0.1763,
+      "step": 21362
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2825550350108309e-05,
+      "loss": 0.3307,
+      "step": 21363
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2825214514801943e-05,
+      "loss": 0.1702,
+      "step": 21364
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2824878679495576e-05,
+      "loss": 0.1975,
+      "step": 21365
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2824542844189213e-05,
+      "loss": 0.2893,
+      "step": 21366
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2824207008882845e-05,
+      "loss": 0.2247,
+      "step": 21367
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2823871173576478e-05,
+      "loss": 0.2751,
+      "step": 21368
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2823535338270115e-05,
+      "loss": 0.1565,
+      "step": 21369
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2823199502963748e-05,
+      "loss": 0.2777,
+      "step": 21370
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.282286366765738e-05,
+      "loss": 0.319,
+      "step": 21371
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2822527832351017e-05,
+      "loss": 0.2867,
+      "step": 21372
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.282219199704465e-05,
+      "loss": 0.3199,
+      "step": 21373
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2821856161738285e-05,
+      "loss": 0.2005,
+      "step": 21374
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.282152032643192e-05,
+      "loss": 0.4873,
+      "step": 21375
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2821184491125554e-05,
+      "loss": 0.2218,
+      "step": 21376
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2820848655819187e-05,
+      "loss": 0.2316,
+      "step": 21377
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2820512820512823e-05,
+      "loss": 0.1579,
+      "step": 21378
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2820176985206456e-05,
+      "loss": 0.4739,
+      "step": 21379
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2819841149900089e-05,
+      "loss": 0.2603,
+      "step": 21380
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2819505314593725e-05,
+      "loss": 0.2852,
+      "step": 21381
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2819169479287358e-05,
+      "loss": 0.2769,
+      "step": 21382
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2818833643980993e-05,
+      "loss": 0.174,
+      "step": 21383
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2818497808674628e-05,
+      "loss": 0.5875,
+      "step": 21384
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2818161973368262e-05,
+      "loss": 0.5691,
+      "step": 21385
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2817826138061895e-05,
+      "loss": 0.2779,
+      "step": 21386
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2817490302755528e-05,
+      "loss": 0.1814,
+      "step": 21387
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2817154467449165e-05,
+      "loss": 0.4885,
+      "step": 21388
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2816818632142798e-05,
+      "loss": 0.2359,
+      "step": 21389
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2816482796836432e-05,
+      "loss": 0.284,
+      "step": 21390
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2816146961530067e-05,
+      "loss": 0.1676,
+      "step": 21391
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.28158111262237e-05,
+      "loss": 0.3075,
+      "step": 21392
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2815475290917334e-05,
+      "loss": 0.2792,
+      "step": 21393
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2815139455610969e-05,
+      "loss": 0.3736,
+      "step": 21394
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2814803620304604e-05,
+      "loss": 0.2896,
+      "step": 21395
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2814467784998237e-05,
+      "loss": 0.3617,
+      "step": 21396
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2814131949691873e-05,
+      "loss": 0.2728,
+      "step": 21397
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2813796114385506e-05,
+      "loss": 0.2319,
+      "step": 21398
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2813460279079139e-05,
+      "loss": 0.2922,
+      "step": 21399
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2813124443772775e-05,
+      "loss": 0.2043,
+      "step": 21400
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2812788608466408e-05,
+      "loss": 0.2519,
+      "step": 21401
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2812452773160043e-05,
+      "loss": 0.4818,
+      "step": 21402
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2812116937853678e-05,
+      "loss": 0.4276,
+      "step": 21403
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2811781102547312e-05,
+      "loss": 0.3411,
+      "step": 21404
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2811445267240945e-05,
+      "loss": 0.1664,
+      "step": 21405
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2811109431934581e-05,
+      "loss": 0.2185,
+      "step": 21406
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2810773596628214e-05,
+      "loss": 0.4655,
+      "step": 21407
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2810437761321847e-05,
+      "loss": 0.2401,
+      "step": 21408
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2810101926015484e-05,
+      "loss": 0.2909,
+      "step": 21409
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2809766090709117e-05,
+      "loss": 0.2838,
+      "step": 21410
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2809430255402751e-05,
+      "loss": 0.2662,
+      "step": 21411
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2809094420096386e-05,
+      "loss": 0.3184,
+      "step": 21412
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2808758584790019e-05,
+      "loss": 0.4156,
+      "step": 21413
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2808422749483654e-05,
+      "loss": 0.2583,
+      "step": 21414
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2808086914177288e-05,
+      "loss": 0.2677,
+      "step": 21415
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2807751078870923e-05,
+      "loss": 0.2584,
+      "step": 21416
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2807415243564556e-05,
+      "loss": 0.3838,
+      "step": 21417
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2807079408258192e-05,
+      "loss": 0.3101,
+      "step": 21418
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2806743572951825e-05,
+      "loss": 0.2703,
+      "step": 21419
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2806407737645458e-05,
+      "loss": 0.4641,
+      "step": 21420
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2806071902339094e-05,
+      "loss": 0.3054,
+      "step": 21421
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2805736067032727e-05,
+      "loss": 0.3919,
+      "step": 21422
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2805400231726362e-05,
+      "loss": 0.1828,
+      "step": 21423
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2805064396419997e-05,
+      "loss": 0.1994,
+      "step": 21424
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2804728561113631e-05,
+      "loss": 0.2374,
+      "step": 21425
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2804392725807264e-05,
+      "loss": 0.3332,
+      "step": 21426
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.28040568905009e-05,
+      "loss": 0.2063,
+      "step": 21427
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2803721055194534e-05,
+      "loss": 0.3556,
+      "step": 21428
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2803385219888167e-05,
+      "loss": 0.3099,
+      "step": 21429
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2803049384581803e-05,
+      "loss": 0.2609,
+      "step": 21430
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2802713549275436e-05,
+      "loss": 0.4236,
+      "step": 21431
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.280237771396907e-05,
+      "loss": 0.4896,
+      "step": 21432
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2802041878662705e-05,
+      "loss": 0.3092,
+      "step": 21433
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.280170604335634e-05,
+      "loss": 0.2834,
+      "step": 21434
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2801370208049973e-05,
+      "loss": 0.3913,
+      "step": 21435
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2801034372743607e-05,
+      "loss": 0.2621,
+      "step": 21436
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2800698537437242e-05,
+      "loss": 0.1728,
+      "step": 21437
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2800362702130875e-05,
+      "loss": 0.3168,
+      "step": 21438
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2800026866824511e-05,
+      "loss": 0.2198,
+      "step": 21439
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2799691031518144e-05,
+      "loss": 0.3855,
+      "step": 21440
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2799355196211777e-05,
+      "loss": 0.2023,
+      "step": 21441
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2799019360905414e-05,
+      "loss": 0.1693,
+      "step": 21442
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2798683525599046e-05,
+      "loss": 0.1899,
+      "step": 21443
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2798347690292681e-05,
+      "loss": 0.2837,
+      "step": 21444
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2798011854986316e-05,
+      "loss": 0.2902,
+      "step": 21445
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.279767601967995e-05,
+      "loss": 0.391,
+      "step": 21446
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2797340184373583e-05,
+      "loss": 0.261,
+      "step": 21447
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.279700434906722e-05,
+      "loss": 0.2317,
+      "step": 21448
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2796668513760853e-05,
+      "loss": 0.1848,
+      "step": 21449
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2796332678454486e-05,
+      "loss": 0.2514,
+      "step": 21450
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2795996843148122e-05,
+      "loss": 0.2433,
+      "step": 21451
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2795661007841755e-05,
+      "loss": 0.2976,
+      "step": 21452
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.279532517253539e-05,
+      "loss": 0.3551,
+      "step": 21453
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2794989337229024e-05,
+      "loss": 0.198,
+      "step": 21454
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2794653501922659e-05,
+      "loss": 0.5012,
+      "step": 21455
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2794317666616292e-05,
+      "loss": 0.2233,
+      "step": 21456
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2793981831309928e-05,
+      "loss": 0.2847,
+      "step": 21457
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2793645996003561e-05,
+      "loss": 0.2892,
+      "step": 21458
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2793310160697194e-05,
+      "loss": 0.3436,
+      "step": 21459
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.279297432539083e-05,
+      "loss": 0.2627,
+      "step": 21460
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2792638490084463e-05,
+      "loss": 0.3037,
+      "step": 21461
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2792302654778096e-05,
+      "loss": 0.4287,
+      "step": 21462
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2791966819471733e-05,
+      "loss": 0.2851,
+      "step": 21463
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2791630984165366e-05,
+      "loss": 0.2352,
+      "step": 21464
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2791295148859e-05,
+      "loss": 0.3481,
+      "step": 21465
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2790959313552635e-05,
+      "loss": 0.4784,
+      "step": 21466
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.279062347824627e-05,
+      "loss": 0.2804,
+      "step": 21467
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2790287642939903e-05,
+      "loss": 0.2082,
+      "step": 21468
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2789951807633539e-05,
+      "loss": 0.1998,
+      "step": 21469
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2789615972327172e-05,
+      "loss": 0.3147,
+      "step": 21470
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2789280137020805e-05,
+      "loss": 0.2401,
+      "step": 21471
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2788944301714441e-05,
+      "loss": 0.2284,
+      "step": 21472
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2788608466408074e-05,
+      "loss": 0.3147,
+      "step": 21473
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2788272631101709e-05,
+      "loss": 0.265,
+      "step": 21474
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2787936795795343e-05,
+      "loss": 0.2982,
+      "step": 21475
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2787600960488978e-05,
+      "loss": 0.5019,
+      "step": 21476
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2787265125182611e-05,
+      "loss": 0.2798,
+      "step": 21477
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2786929289876247e-05,
+      "loss": 0.3263,
+      "step": 21478
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.278659345456988e-05,
+      "loss": 0.217,
+      "step": 21479
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2786257619263513e-05,
+      "loss": 0.2028,
+      "step": 21480
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.278592178395715e-05,
+      "loss": 0.2485,
+      "step": 21481
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2785585948650782e-05,
+      "loss": 0.1705,
+      "step": 21482
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2785250113344415e-05,
+      "loss": 0.321,
+      "step": 21483
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2784914278038052e-05,
+      "loss": 0.3869,
+      "step": 21484
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2784578442731685e-05,
+      "loss": 0.2123,
+      "step": 21485
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.278424260742532e-05,
+      "loss": 0.2968,
+      "step": 21486
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2783906772118954e-05,
+      "loss": 0.2812,
+      "step": 21487
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2783570936812589e-05,
+      "loss": 0.1695,
+      "step": 21488
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2783235101506222e-05,
+      "loss": 0.3763,
+      "step": 21489
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2782899266199858e-05,
+      "loss": 0.1974,
+      "step": 21490
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2782563430893491e-05,
+      "loss": 0.3592,
+      "step": 21491
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2782227595587124e-05,
+      "loss": 0.3209,
+      "step": 21492
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.278189176028076e-05,
+      "loss": 0.2502,
+      "step": 21493
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2781555924974393e-05,
+      "loss": 0.2862,
+      "step": 21494
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2781220089668028e-05,
+      "loss": 0.2888,
+      "step": 21495
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2780884254361662e-05,
+      "loss": 0.4423,
+      "step": 21496
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2780548419055297e-05,
+      "loss": 0.289,
+      "step": 21497
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.278021258374893e-05,
+      "loss": 0.6247,
+      "step": 21498
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2779876748442566e-05,
+      "loss": 0.523,
+      "step": 21499
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.27795409131362e-05,
+      "loss": 0.2261,
+      "step": 21500
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2779205077829832e-05,
+      "loss": 0.2561,
+      "step": 21501
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2778869242523469e-05,
+      "loss": 0.3898,
+      "step": 21502
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2778533407217102e-05,
+      "loss": 0.3703,
+      "step": 21503
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2778197571910735e-05,
+      "loss": 0.2986,
+      "step": 21504
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2777861736604371e-05,
+      "loss": 0.3775,
+      "step": 21505
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2777525901298004e-05,
+      "loss": 0.473,
+      "step": 21506
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2777190065991639e-05,
+      "loss": 0.3049,
+      "step": 21507
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2776854230685273e-05,
+      "loss": 0.3638,
+      "step": 21508
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2776518395378908e-05,
+      "loss": 0.5642,
+      "step": 21509
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.277618256007254e-05,
+      "loss": 0.3014,
+      "step": 21510
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2775846724766177e-05,
+      "loss": 0.2383,
+      "step": 21511
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.277551088945981e-05,
+      "loss": 0.4545,
+      "step": 21512
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2775175054153443e-05,
+      "loss": 0.2247,
+      "step": 21513
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.277483921884708e-05,
+      "loss": 0.3168,
+      "step": 21514
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2774503383540712e-05,
+      "loss": 0.4788,
+      "step": 21515
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2774167548234347e-05,
+      "loss": 0.2662,
+      "step": 21516
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2773831712927982e-05,
+      "loss": 0.1479,
+      "step": 21517
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2773495877621616e-05,
+      "loss": 0.2691,
+      "step": 21518
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.277316004231525e-05,
+      "loss": 0.3505,
+      "step": 21519
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2772824207008886e-05,
+      "loss": 0.3292,
+      "step": 21520
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2772488371702518e-05,
+      "loss": 0.2722,
+      "step": 21521
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2772152536396151e-05,
+      "loss": 0.2707,
+      "step": 21522
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2771816701089788e-05,
+      "loss": 0.3835,
+      "step": 21523
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.277148086578342e-05,
+      "loss": 0.1913,
+      "step": 21524
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2771145030477055e-05,
+      "loss": 0.3608,
+      "step": 21525
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.277080919517069e-05,
+      "loss": 0.4131,
+      "step": 21526
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2770473359864323e-05,
+      "loss": 0.3171,
+      "step": 21527
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2770137524557958e-05,
+      "loss": 0.2421,
+      "step": 21528
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2769801689251592e-05,
+      "loss": 0.1749,
+      "step": 21529
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2769465853945227e-05,
+      "loss": 0.3077,
+      "step": 21530
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.276913001863886e-05,
+      "loss": 0.3836,
+      "step": 21531
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2768794183332496e-05,
+      "loss": 0.2755,
+      "step": 21532
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.276845834802613e-05,
+      "loss": 0.2573,
+      "step": 21533
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2768122512719762e-05,
+      "loss": 0.3114,
+      "step": 21534
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2767786677413398e-05,
+      "loss": 0.3799,
+      "step": 21535
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2767450842107031e-05,
+      "loss": 0.2289,
+      "step": 21536
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.2767115006800666e-05,
+      "loss": 0.2896,
+      "step": 21537
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 1.27667791714943e-05,
+      "loss": 0.1755,
+      "step": 21538
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2766443336187935e-05,
+      "loss": 0.2771,
+      "step": 21539
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2766107500881568e-05,
+      "loss": 0.4797,
+      "step": 21540
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2765771665575205e-05,
+      "loss": 0.2322,
+      "step": 21541
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2765435830268838e-05,
+      "loss": 0.3955,
+      "step": 21542
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.276509999496247e-05,
+      "loss": 0.0953,
+      "step": 21543
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2764764159656107e-05,
+      "loss": 0.1884,
+      "step": 21544
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.276442832434974e-05,
+      "loss": 0.2536,
+      "step": 21545
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2764092489043375e-05,
+      "loss": 0.2558,
+      "step": 21546
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.276375665373701e-05,
+      "loss": 0.1856,
+      "step": 21547
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2763420818430642e-05,
+      "loss": 0.4387,
+      "step": 21548
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2763084983124277e-05,
+      "loss": 0.3541,
+      "step": 21549
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2762749147817911e-05,
+      "loss": 0.1847,
+      "step": 21550
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2762413312511546e-05,
+      "loss": 0.3179,
+      "step": 21551
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2762077477205179e-05,
+      "loss": 0.287,
+      "step": 21552
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2761741641898815e-05,
+      "loss": 0.2636,
+      "step": 21553
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2761405806592448e-05,
+      "loss": 0.228,
+      "step": 21554
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2761069971286081e-05,
+      "loss": 0.2559,
+      "step": 21555
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2760734135979718e-05,
+      "loss": 0.2486,
+      "step": 21556
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.276039830067335e-05,
+      "loss": 0.2267,
+      "step": 21557
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2760062465366985e-05,
+      "loss": 0.2627,
+      "step": 21558
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.275972663006062e-05,
+      "loss": 0.3686,
+      "step": 21559
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2759390794754254e-05,
+      "loss": 0.1998,
+      "step": 21560
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2759054959447887e-05,
+      "loss": 0.3674,
+      "step": 21561
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2758719124141524e-05,
+      "loss": 0.2515,
+      "step": 21562
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2758383288835157e-05,
+      "loss": 0.2751,
+      "step": 21563
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.275804745352879e-05,
+      "loss": 0.3937,
+      "step": 21564
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2757711618222426e-05,
+      "loss": 0.4699,
+      "step": 21565
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2757375782916059e-05,
+      "loss": 0.2462,
+      "step": 21566
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2757039947609694e-05,
+      "loss": 0.3968,
+      "step": 21567
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2756704112303327e-05,
+      "loss": 0.2201,
+      "step": 21568
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2756368276996963e-05,
+      "loss": 0.2371,
+      "step": 21569
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2756032441690596e-05,
+      "loss": 0.389,
+      "step": 21570
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2755696606384229e-05,
+      "loss": 0.2569,
+      "step": 21571
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2755360771077865e-05,
+      "loss": 0.4882,
+      "step": 21572
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2755024935771498e-05,
+      "loss": 0.2888,
+      "step": 21573
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2754689100465131e-05,
+      "loss": 0.356,
+      "step": 21574
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2754353265158767e-05,
+      "loss": 0.2248,
+      "step": 21575
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.27540174298524e-05,
+      "loss": 0.3567,
+      "step": 21576
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2753681594546035e-05,
+      "loss": 0.2926,
+      "step": 21577
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.275334575923967e-05,
+      "loss": 0.3824,
+      "step": 21578
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2753009923933304e-05,
+      "loss": 0.457,
+      "step": 21579
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2752674088626937e-05,
+      "loss": 0.3041,
+      "step": 21580
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2752338253320574e-05,
+      "loss": 0.1706,
+      "step": 21581
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2752002418014207e-05,
+      "loss": 0.3877,
+      "step": 21582
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.275166658270784e-05,
+      "loss": 0.2364,
+      "step": 21583
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2751330747401476e-05,
+      "loss": 0.2894,
+      "step": 21584
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2750994912095109e-05,
+      "loss": 0.1941,
+      "step": 21585
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2750659076788743e-05,
+      "loss": 0.2767,
+      "step": 21586
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2750323241482378e-05,
+      "loss": 0.3209,
+      "step": 21587
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2749987406176013e-05,
+      "loss": 0.2059,
+      "step": 21588
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2749651570869646e-05,
+      "loss": 0.3177,
+      "step": 21589
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2749315735563282e-05,
+      "loss": 0.1884,
+      "step": 21590
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2748979900256915e-05,
+      "loss": 0.4232,
+      "step": 21591
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2748644064950548e-05,
+      "loss": 0.2245,
+      "step": 21592
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2748308229644184e-05,
+      "loss": 0.2386,
+      "step": 21593
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2747972394337817e-05,
+      "loss": 0.3375,
+      "step": 21594
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.274763655903145e-05,
+      "loss": 0.4266,
+      "step": 21595
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2747300723725087e-05,
+      "loss": 0.3465,
+      "step": 21596
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.274696488841872e-05,
+      "loss": 0.218,
+      "step": 21597
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2746629053112354e-05,
+      "loss": 0.2489,
+      "step": 21598
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2746293217805989e-05,
+      "loss": 0.4125,
+      "step": 21599
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2745957382499623e-05,
+      "loss": 0.3293,
+      "step": 21600
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2745621547193256e-05,
+      "loss": 0.2708,
+      "step": 21601
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2745285711886893e-05,
+      "loss": 0.3352,
+      "step": 21602
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2744949876580526e-05,
+      "loss": 0.199,
+      "step": 21603
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2744614041274159e-05,
+      "loss": 0.2351,
+      "step": 21604
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2744278205967795e-05,
+      "loss": 0.4208,
+      "step": 21605
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2743942370661428e-05,
+      "loss": 0.2415,
+      "step": 21606
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2743606535355063e-05,
+      "loss": 0.2661,
+      "step": 21607
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2743270700048697e-05,
+      "loss": 0.1973,
+      "step": 21608
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2742934864742332e-05,
+      "loss": 0.3819,
+      "step": 21609
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2742599029435965e-05,
+      "loss": 0.3143,
+      "step": 21610
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2742263194129601e-05,
+      "loss": 0.2823,
+      "step": 21611
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2741927358823234e-05,
+      "loss": 0.3947,
+      "step": 21612
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2741591523516867e-05,
+      "loss": 0.2938,
+      "step": 21613
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2741255688210503e-05,
+      "loss": 0.3246,
+      "step": 21614
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2740919852904136e-05,
+      "loss": 0.3303,
+      "step": 21615
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2740584017597771e-05,
+      "loss": 0.21,
+      "step": 21616
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2740248182291406e-05,
+      "loss": 0.3464,
+      "step": 21617
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2739912346985039e-05,
+      "loss": 0.3466,
+      "step": 21618
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2739576511678673e-05,
+      "loss": 0.394,
+      "step": 21619
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2739240676372308e-05,
+      "loss": 0.2177,
+      "step": 21620
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2738904841065943e-05,
+      "loss": 0.166,
+      "step": 21621
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2738569005759576e-05,
+      "loss": 0.3174,
+      "step": 21622
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2738233170453212e-05,
+      "loss": 0.2688,
+      "step": 21623
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2737897335146845e-05,
+      "loss": 0.3269,
+      "step": 21624
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2737561499840478e-05,
+      "loss": 0.3248,
+      "step": 21625
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2737225664534114e-05,
+      "loss": 0.2882,
+      "step": 21626
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2736889829227747e-05,
+      "loss": 0.2884,
+      "step": 21627
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2736553993921382e-05,
+      "loss": 0.1583,
+      "step": 21628
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2736218158615016e-05,
+      "loss": 0.559,
+      "step": 21629
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2735882323308651e-05,
+      "loss": 0.2067,
+      "step": 21630
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2735546488002284e-05,
+      "loss": 0.3414,
+      "step": 21631
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.273521065269592e-05,
+      "loss": 0.1803,
+      "step": 21632
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2734874817389553e-05,
+      "loss": 0.3901,
+      "step": 21633
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2734538982083186e-05,
+      "loss": 0.346,
+      "step": 21634
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2734203146776823e-05,
+      "loss": 0.4129,
+      "step": 21635
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2733867311470456e-05,
+      "loss": 0.3657,
+      "step": 21636
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.273353147616409e-05,
+      "loss": 0.3576,
+      "step": 21637
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2733195640857725e-05,
+      "loss": 0.2859,
+      "step": 21638
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2732859805551358e-05,
+      "loss": 0.3218,
+      "step": 21639
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2732523970244992e-05,
+      "loss": 0.1769,
+      "step": 21640
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2732188134938627e-05,
+      "loss": 0.1811,
+      "step": 21641
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2731852299632262e-05,
+      "loss": 0.2632,
+      "step": 21642
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2731516464325895e-05,
+      "loss": 0.3364,
+      "step": 21643
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2731180629019531e-05,
+      "loss": 0.4484,
+      "step": 21644
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2730844793713164e-05,
+      "loss": 0.1794,
+      "step": 21645
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2730508958406797e-05,
+      "loss": 0.3955,
+      "step": 21646
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2730173123100433e-05,
+      "loss": 0.1559,
+      "step": 21647
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2729837287794066e-05,
+      "loss": 0.224,
+      "step": 21648
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.27295014524877e-05,
+      "loss": 0.2206,
+      "step": 21649
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2729165617181335e-05,
+      "loss": 0.4536,
+      "step": 21650
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.272882978187497e-05,
+      "loss": 0.2542,
+      "step": 21651
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2728493946568603e-05,
+      "loss": 0.2455,
+      "step": 21652
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.272815811126224e-05,
+      "loss": 0.344,
+      "step": 21653
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2727822275955872e-05,
+      "loss": 0.2575,
+      "step": 21654
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2727486440649505e-05,
+      "loss": 0.3035,
+      "step": 21655
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2727150605343142e-05,
+      "loss": 0.3038,
+      "step": 21656
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2726814770036775e-05,
+      "loss": 0.3025,
+      "step": 21657
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.272647893473041e-05,
+      "loss": 0.3365,
+      "step": 21658
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2726143099424044e-05,
+      "loss": 0.3071,
+      "step": 21659
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2725807264117679e-05,
+      "loss": 0.3293,
+      "step": 21660
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2725471428811312e-05,
+      "loss": 0.1673,
+      "step": 21661
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2725135593504946e-05,
+      "loss": 0.4025,
+      "step": 21662
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.272479975819858e-05,
+      "loss": 0.3091,
+      "step": 21663
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2724463922892214e-05,
+      "loss": 0.225,
+      "step": 21664
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.272412808758585e-05,
+      "loss": 0.3238,
+      "step": 21665
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2723792252279483e-05,
+      "loss": 0.1932,
+      "step": 21666
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2723456416973116e-05,
+      "loss": 0.3825,
+      "step": 21667
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2723120581666752e-05,
+      "loss": 0.3694,
+      "step": 21668
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2722784746360385e-05,
+      "loss": 0.3451,
+      "step": 21669
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.272244891105402e-05,
+      "loss": 0.2067,
+      "step": 21670
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2722113075747655e-05,
+      "loss": 0.3534,
+      "step": 21671
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.272177724044129e-05,
+      "loss": 0.2582,
+      "step": 21672
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2721441405134922e-05,
+      "loss": 0.328,
+      "step": 21673
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2721105569828559e-05,
+      "loss": 0.3444,
+      "step": 21674
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2720769734522192e-05,
+      "loss": 0.4594,
+      "step": 21675
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2720433899215824e-05,
+      "loss": 0.3989,
+      "step": 21676
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.272009806390946e-05,
+      "loss": 0.1081,
+      "step": 21677
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2719762228603094e-05,
+      "loss": 0.292,
+      "step": 21678
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2719426393296728e-05,
+      "loss": 0.287,
+      "step": 21679
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2719090557990363e-05,
+      "loss": 0.2844,
+      "step": 21680
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2718754722683998e-05,
+      "loss": 0.4305,
+      "step": 21681
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.271841888737763e-05,
+      "loss": 0.2787,
+      "step": 21682
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2718083052071265e-05,
+      "loss": 0.134,
+      "step": 21683
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.27177472167649e-05,
+      "loss": 0.272,
+      "step": 21684
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2717411381458533e-05,
+      "loss": 0.2067,
+      "step": 21685
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.271707554615217e-05,
+      "loss": 0.1899,
+      "step": 21686
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2716739710845802e-05,
+      "loss": 0.4441,
+      "step": 21687
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2716403875539435e-05,
+      "loss": 0.2206,
+      "step": 21688
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2716068040233071e-05,
+      "loss": 0.2947,
+      "step": 21689
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2715732204926704e-05,
+      "loss": 0.2592,
+      "step": 21690
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2715396369620339e-05,
+      "loss": 0.3666,
+      "step": 21691
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2715060534313974e-05,
+      "loss": 0.2976,
+      "step": 21692
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2714724699007608e-05,
+      "loss": 0.3541,
+      "step": 21693
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2714388863701241e-05,
+      "loss": 0.3675,
+      "step": 21694
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2714053028394878e-05,
+      "loss": 0.2915,
+      "step": 21695
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.271371719308851e-05,
+      "loss": 0.4556,
+      "step": 21696
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2713381357782144e-05,
+      "loss": 0.2832,
+      "step": 21697
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.271304552247578e-05,
+      "loss": 0.1321,
+      "step": 21698
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2712709687169413e-05,
+      "loss": 0.2214,
+      "step": 21699
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2712373851863048e-05,
+      "loss": 0.3801,
+      "step": 21700
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2712038016556682e-05,
+      "loss": 0.2278,
+      "step": 21701
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2711702181250317e-05,
+      "loss": 0.2328,
+      "step": 21702
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.271136634594395e-05,
+      "loss": 0.3564,
+      "step": 21703
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2711030510637586e-05,
+      "loss": 0.4128,
+      "step": 21704
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2710694675331219e-05,
+      "loss": 0.4221,
+      "step": 21705
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2710358840024852e-05,
+      "loss": 0.1881,
+      "step": 21706
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2710023004718488e-05,
+      "loss": 0.1544,
+      "step": 21707
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2709687169412121e-05,
+      "loss": 0.1919,
+      "step": 21708
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2709351334105754e-05,
+      "loss": 0.3023,
+      "step": 21709
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.270901549879939e-05,
+      "loss": 0.2474,
+      "step": 21710
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2708679663493024e-05,
+      "loss": 0.292,
+      "step": 21711
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2708343828186658e-05,
+      "loss": 0.2931,
+      "step": 21712
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2708007992880293e-05,
+      "loss": 0.3139,
+      "step": 21713
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2707672157573928e-05,
+      "loss": 0.2423,
+      "step": 21714
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.270733632226756e-05,
+      "loss": 0.4836,
+      "step": 21715
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2707000486961197e-05,
+      "loss": 0.238,
+      "step": 21716
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.270666465165483e-05,
+      "loss": 0.2504,
+      "step": 21717
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2706328816348463e-05,
+      "loss": 0.2429,
+      "step": 21718
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2705992981042099e-05,
+      "loss": 0.2658,
+      "step": 21719
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2705657145735732e-05,
+      "loss": 0.4015,
+      "step": 21720
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2705321310429367e-05,
+      "loss": 0.4569,
+      "step": 21721
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2704985475123001e-05,
+      "loss": 0.5088,
+      "step": 21722
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2704649639816636e-05,
+      "loss": 0.2194,
+      "step": 21723
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2704313804510269e-05,
+      "loss": 0.2369,
+      "step": 21724
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2703977969203905e-05,
+      "loss": 0.3156,
+      "step": 21725
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2703642133897538e-05,
+      "loss": 0.2519,
+      "step": 21726
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2703306298591171e-05,
+      "loss": 0.2217,
+      "step": 21727
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2702970463284807e-05,
+      "loss": 0.1908,
+      "step": 21728
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.270263462797844e-05,
+      "loss": 0.3279,
+      "step": 21729
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2702298792672073e-05,
+      "loss": 0.2363,
+      "step": 21730
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.270196295736571e-05,
+      "loss": 0.179,
+      "step": 21731
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2701627122059343e-05,
+      "loss": 0.2055,
+      "step": 21732
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2701291286752977e-05,
+      "loss": 0.2423,
+      "step": 21733
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2700955451446612e-05,
+      "loss": 0.1787,
+      "step": 21734
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.2700619616140247e-05,
+      "loss": 0.2745,
+      "step": 21735
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.270028378083388e-05,
+      "loss": 0.3013,
+      "step": 21736
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2699947945527516e-05,
+      "loss": 0.5683,
+      "step": 21737
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2699612110221149e-05,
+      "loss": 0.3265,
+      "step": 21738
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2699276274914782e-05,
+      "loss": 0.1995,
+      "step": 21739
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2698940439608418e-05,
+      "loss": 0.3489,
+      "step": 21740
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2698604604302051e-05,
+      "loss": 0.3469,
+      "step": 21741
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2698268768995686e-05,
+      "loss": 0.2742,
+      "step": 21742
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.269793293368932e-05,
+      "loss": 0.3762,
+      "step": 21743
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2697597098382955e-05,
+      "loss": 0.1398,
+      "step": 21744
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2697261263076588e-05,
+      "loss": 0.3292,
+      "step": 21745
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2696925427770221e-05,
+      "loss": 0.4903,
+      "step": 21746
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2696589592463857e-05,
+      "loss": 0.4373,
+      "step": 21747
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.269625375715749e-05,
+      "loss": 0.4276,
+      "step": 21748
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2695917921851125e-05,
+      "loss": 0.3419,
+      "step": 21749
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.269558208654476e-05,
+      "loss": 0.2516,
+      "step": 21750
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2695246251238394e-05,
+      "loss": 0.2841,
+      "step": 21751
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2694910415932027e-05,
+      "loss": 0.3052,
+      "step": 21752
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2694574580625662e-05,
+      "loss": 0.4044,
+      "step": 21753
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2694238745319296e-05,
+      "loss": 0.1409,
+      "step": 21754
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.269390291001293e-05,
+      "loss": 0.4153,
+      "step": 21755
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2693567074706566e-05,
+      "loss": 0.4123,
+      "step": 21756
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2693231239400199e-05,
+      "loss": 0.1535,
+      "step": 21757
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2692895404093832e-05,
+      "loss": 0.1238,
+      "step": 21758
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2692559568787468e-05,
+      "loss": 0.3342,
+      "step": 21759
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2692223733481101e-05,
+      "loss": 0.4446,
+      "step": 21760
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2691887898174736e-05,
+      "loss": 0.3833,
+      "step": 21761
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.269155206286837e-05,
+      "loss": 0.209,
+      "step": 21762
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2691216227562005e-05,
+      "loss": 0.1686,
+      "step": 21763
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2690880392255638e-05,
+      "loss": 0.1936,
+      "step": 21764
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2690544556949274e-05,
+      "loss": 0.365,
+      "step": 21765
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2690208721642907e-05,
+      "loss": 0.2606,
+      "step": 21766
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.268987288633654e-05,
+      "loss": 0.4394,
+      "step": 21767
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2689537051030176e-05,
+      "loss": 0.2762,
+      "step": 21768
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.268920121572381e-05,
+      "loss": 0.161,
+      "step": 21769
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2688865380417444e-05,
+      "loss": 0.2179,
+      "step": 21770
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2688529545111079e-05,
+      "loss": 0.4131,
+      "step": 21771
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2688193709804713e-05,
+      "loss": 0.2636,
+      "step": 21772
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2687857874498346e-05,
+      "loss": 0.2312,
+      "step": 21773
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2687522039191981e-05,
+      "loss": 0.3073,
+      "step": 21774
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2687186203885616e-05,
+      "loss": 0.2056,
+      "step": 21775
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2686850368579249e-05,
+      "loss": 0.2916,
+      "step": 21776
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2686514533272885e-05,
+      "loss": 0.3928,
+      "step": 21777
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2686178697966518e-05,
+      "loss": 0.1602,
+      "step": 21778
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.268584286266015e-05,
+      "loss": 0.3228,
+      "step": 21779
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2685507027353787e-05,
+      "loss": 0.2033,
+      "step": 21780
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.268517119204742e-05,
+      "loss": 0.3662,
+      "step": 21781
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2684835356741055e-05,
+      "loss": 0.4586,
+      "step": 21782
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.268449952143469e-05,
+      "loss": 0.4979,
+      "step": 21783
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2684163686128324e-05,
+      "loss": 0.2414,
+      "step": 21784
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2683827850821957e-05,
+      "loss": 0.3676,
+      "step": 21785
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2683492015515593e-05,
+      "loss": 0.2878,
+      "step": 21786
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2683156180209226e-05,
+      "loss": 0.3212,
+      "step": 21787
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.268282034490286e-05,
+      "loss": 0.2859,
+      "step": 21788
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2682484509596496e-05,
+      "loss": 0.3662,
+      "step": 21789
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2682148674290129e-05,
+      "loss": 0.3109,
+      "step": 21790
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2681812838983763e-05,
+      "loss": 0.2864,
+      "step": 21791
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2681477003677398e-05,
+      "loss": 0.2788,
+      "step": 21792
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2681141168371032e-05,
+      "loss": 0.2277,
+      "step": 21793
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2680805333064665e-05,
+      "loss": 0.2767,
+      "step": 21794
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2680469497758302e-05,
+      "loss": 0.3517,
+      "step": 21795
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2680133662451935e-05,
+      "loss": 0.3152,
+      "step": 21796
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2679797827145568e-05,
+      "loss": 0.6013,
+      "step": 21797
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2679461991839204e-05,
+      "loss": 0.2451,
+      "step": 21798
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2679126156532837e-05,
+      "loss": 0.3178,
+      "step": 21799
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.267879032122647e-05,
+      "loss": 0.3359,
+      "step": 21800
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2678454485920106e-05,
+      "loss": 0.4614,
+      "step": 21801
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.267811865061374e-05,
+      "loss": 0.1859,
+      "step": 21802
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2677782815307374e-05,
+      "loss": 0.2558,
+      "step": 21803
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2677446980001009e-05,
+      "loss": 0.3156,
+      "step": 21804
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2677111144694643e-05,
+      "loss": 0.2122,
+      "step": 21805
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2676775309388276e-05,
+      "loss": 0.2485,
+      "step": 21806
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2676439474081912e-05,
+      "loss": 0.2605,
+      "step": 21807
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2676103638775545e-05,
+      "loss": 0.4854,
+      "step": 21808
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2675767803469178e-05,
+      "loss": 0.1642,
+      "step": 21809
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2675431968162815e-05,
+      "loss": 0.2279,
+      "step": 21810
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2675096132856448e-05,
+      "loss": 0.5708,
+      "step": 21811
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2674760297550082e-05,
+      "loss": 0.2978,
+      "step": 21812
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2674424462243717e-05,
+      "loss": 0.2734,
+      "step": 21813
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2674088626937352e-05,
+      "loss": 0.3858,
+      "step": 21814
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2673752791630985e-05,
+      "loss": 0.3279,
+      "step": 21815
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2673416956324621e-05,
+      "loss": 0.3841,
+      "step": 21816
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2673081121018254e-05,
+      "loss": 0.4325,
+      "step": 21817
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2672745285711887e-05,
+      "loss": 0.2504,
+      "step": 21818
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2672409450405523e-05,
+      "loss": 0.3489,
+      "step": 21819
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2672073615099156e-05,
+      "loss": 0.2134,
+      "step": 21820
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2671737779792789e-05,
+      "loss": 0.3225,
+      "step": 21821
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2671401944486425e-05,
+      "loss": 0.4671,
+      "step": 21822
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2671066109180058e-05,
+      "loss": 0.3018,
+      "step": 21823
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2670730273873693e-05,
+      "loss": 0.2378,
+      "step": 21824
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2670394438567328e-05,
+      "loss": 0.3431,
+      "step": 21825
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2670058603260962e-05,
+      "loss": 0.1877,
+      "step": 21826
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2669722767954595e-05,
+      "loss": 0.393,
+      "step": 21827
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2669386932648232e-05,
+      "loss": 0.3125,
+      "step": 21828
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2669051097341865e-05,
+      "loss": 0.309,
+      "step": 21829
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2668715262035497e-05,
+      "loss": 0.4857,
+      "step": 21830
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2668379426729134e-05,
+      "loss": 0.4073,
+      "step": 21831
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2668043591422767e-05,
+      "loss": 0.4109,
+      "step": 21832
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2667707756116401e-05,
+      "loss": 0.2599,
+      "step": 21833
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2667371920810036e-05,
+      "loss": 0.3847,
+      "step": 21834
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.266703608550367e-05,
+      "loss": 0.3994,
+      "step": 21835
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2666700250197304e-05,
+      "loss": 0.3145,
+      "step": 21836
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.266636441489094e-05,
+      "loss": 0.3084,
+      "step": 21837
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2666028579584573e-05,
+      "loss": 0.4141,
+      "step": 21838
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2665692744278206e-05,
+      "loss": 0.3173,
+      "step": 21839
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2665356908971842e-05,
+      "loss": 0.184,
+      "step": 21840
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2665021073665475e-05,
+      "loss": 0.2236,
+      "step": 21841
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2664685238359108e-05,
+      "loss": 0.2646,
+      "step": 21842
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2664349403052745e-05,
+      "loss": 0.1927,
+      "step": 21843
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2664013567746377e-05,
+      "loss": 0.2723,
+      "step": 21844
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2663677732440012e-05,
+      "loss": 0.3167,
+      "step": 21845
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2663341897133647e-05,
+      "loss": 0.2297,
+      "step": 21846
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2663006061827281e-05,
+      "loss": 0.3546,
+      "step": 21847
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2662670226520914e-05,
+      "loss": 0.3452,
+      "step": 21848
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.266233439121455e-05,
+      "loss": 0.299,
+      "step": 21849
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2661998555908184e-05,
+      "loss": 0.2916,
+      "step": 21850
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2661662720601817e-05,
+      "loss": 0.3627,
+      "step": 21851
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2661326885295453e-05,
+      "loss": 0.484,
+      "step": 21852
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2660991049989086e-05,
+      "loss": 0.3908,
+      "step": 21853
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.266065521468272e-05,
+      "loss": 0.3459,
+      "step": 21854
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2660319379376355e-05,
+      "loss": 0.2706,
+      "step": 21855
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.265998354406999e-05,
+      "loss": 0.1955,
+      "step": 21856
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2659647708763623e-05,
+      "loss": 0.2709,
+      "step": 21857
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2659311873457259e-05,
+      "loss": 0.1699,
+      "step": 21858
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2658976038150892e-05,
+      "loss": 0.2669,
+      "step": 21859
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2658640202844525e-05,
+      "loss": 0.4056,
+      "step": 21860
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2658304367538161e-05,
+      "loss": 0.4718,
+      "step": 21861
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2657968532231794e-05,
+      "loss": 0.2725,
+      "step": 21862
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2657632696925429e-05,
+      "loss": 0.154,
+      "step": 21863
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2657296861619064e-05,
+      "loss": 0.3254,
+      "step": 21864
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2656961026312697e-05,
+      "loss": 0.1628,
+      "step": 21865
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2656625191006331e-05,
+      "loss": 0.3038,
+      "step": 21866
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2656289355699966e-05,
+      "loss": 0.322,
+      "step": 21867
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.26559535203936e-05,
+      "loss": 0.6704,
+      "step": 21868
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2655617685087233e-05,
+      "loss": 0.349,
+      "step": 21869
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.265528184978087e-05,
+      "loss": 0.3795,
+      "step": 21870
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2654946014474503e-05,
+      "loss": 0.2899,
+      "step": 21871
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2654610179168136e-05,
+      "loss": 0.1896,
+      "step": 21872
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2654274343861772e-05,
+      "loss": 0.2652,
+      "step": 21873
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2653938508555405e-05,
+      "loss": 0.3627,
+      "step": 21874
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.265360267324904e-05,
+      "loss": 0.3792,
+      "step": 21875
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2653266837942674e-05,
+      "loss": 0.6008,
+      "step": 21876
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2652931002636309e-05,
+      "loss": 0.3417,
+      "step": 21877
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2652595167329942e-05,
+      "loss": 0.2886,
+      "step": 21878
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2652259332023578e-05,
+      "loss": 0.2462,
+      "step": 21879
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2651923496717211e-05,
+      "loss": 0.3584,
+      "step": 21880
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2651587661410844e-05,
+      "loss": 0.2826,
+      "step": 21881
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.265125182610448e-05,
+      "loss": 0.2942,
+      "step": 21882
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2650915990798113e-05,
+      "loss": 0.1901,
+      "step": 21883
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2650580155491748e-05,
+      "loss": 0.1844,
+      "step": 21884
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2650244320185383e-05,
+      "loss": 0.2914,
+      "step": 21885
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2649908484879017e-05,
+      "loss": 0.2465,
+      "step": 21886
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.264957264957265e-05,
+      "loss": 0.3277,
+      "step": 21887
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2649236814266285e-05,
+      "loss": 0.4057,
+      "step": 21888
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.264890097895992e-05,
+      "loss": 0.3725,
+      "step": 21889
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2648565143653553e-05,
+      "loss": 0.3887,
+      "step": 21890
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2648229308347189e-05,
+      "loss": 0.2387,
+      "step": 21891
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2647893473040822e-05,
+      "loss": 0.4255,
+      "step": 21892
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2647557637734455e-05,
+      "loss": 0.1711,
+      "step": 21893
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2647221802428091e-05,
+      "loss": 0.2041,
+      "step": 21894
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2646885967121724e-05,
+      "loss": 0.284,
+      "step": 21895
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2646550131815359e-05,
+      "loss": 0.3266,
+      "step": 21896
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2646214296508993e-05,
+      "loss": 0.283,
+      "step": 21897
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2645878461202628e-05,
+      "loss": 0.2591,
+      "step": 21898
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2645542625896261e-05,
+      "loss": 0.3196,
+      "step": 21899
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2645206790589897e-05,
+      "loss": 0.3812,
+      "step": 21900
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.264487095528353e-05,
+      "loss": 0.4672,
+      "step": 21901
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2644535119977163e-05,
+      "loss": 0.1644,
+      "step": 21902
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.26441992846708e-05,
+      "loss": 0.3411,
+      "step": 21903
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2643863449364433e-05,
+      "loss": 0.4965,
+      "step": 21904
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2643527614058067e-05,
+      "loss": 0.2871,
+      "step": 21905
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2643191778751702e-05,
+      "loss": 0.2591,
+      "step": 21906
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2642855943445337e-05,
+      "loss": 0.2955,
+      "step": 21907
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.264252010813897e-05,
+      "loss": 0.3537,
+      "step": 21908
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2642184272832604e-05,
+      "loss": 0.3559,
+      "step": 21909
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2641848437526239e-05,
+      "loss": 0.1861,
+      "step": 21910
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2641512602219872e-05,
+      "loss": 0.2927,
+      "step": 21911
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2641176766913508e-05,
+      "loss": 0.3268,
+      "step": 21912
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2640840931607141e-05,
+      "loss": 0.2206,
+      "step": 21913
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2640505096300774e-05,
+      "loss": 0.4863,
+      "step": 21914
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.264016926099441e-05,
+      "loss": 0.3354,
+      "step": 21915
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2639833425688043e-05,
+      "loss": 0.1852,
+      "step": 21916
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2639497590381678e-05,
+      "loss": 0.245,
+      "step": 21917
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2639161755075313e-05,
+      "loss": 0.1679,
+      "step": 21918
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2638825919768947e-05,
+      "loss": 0.2079,
+      "step": 21919
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.263849008446258e-05,
+      "loss": 0.4125,
+      "step": 21920
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2638154249156217e-05,
+      "loss": 0.23,
+      "step": 21921
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.263781841384985e-05,
+      "loss": 0.1667,
+      "step": 21922
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2637482578543482e-05,
+      "loss": 0.3411,
+      "step": 21923
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2637146743237119e-05,
+      "loss": 0.168,
+      "step": 21924
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2636810907930752e-05,
+      "loss": 0.1897,
+      "step": 21925
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2636475072624386e-05,
+      "loss": 0.2249,
+      "step": 21926
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.263613923731802e-05,
+      "loss": 0.2106,
+      "step": 21927
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2635803402011656e-05,
+      "loss": 0.2875,
+      "step": 21928
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2635467566705289e-05,
+      "loss": 0.3655,
+      "step": 21929
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2635131731398922e-05,
+      "loss": 0.1917,
+      "step": 21930
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2634795896092558e-05,
+      "loss": 0.4172,
+      "step": 21931
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2634460060786191e-05,
+      "loss": 0.3259,
+      "step": 21932
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2634124225479824e-05,
+      "loss": 0.536,
+      "step": 21933
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.263378839017346e-05,
+      "loss": 0.3097,
+      "step": 21934
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 1.2633452554867093e-05,
+      "loss": 0.4365,
+      "step": 21935
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2633116719560728e-05,
+      "loss": 0.2456,
+      "step": 21936
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2632780884254362e-05,
+      "loss": 0.2262,
+      "step": 21937
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2632445048947997e-05,
+      "loss": 0.3442,
+      "step": 21938
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.263210921364163e-05,
+      "loss": 0.2822,
+      "step": 21939
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2631773378335266e-05,
+      "loss": 0.1726,
+      "step": 21940
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.26314375430289e-05,
+      "loss": 0.2041,
+      "step": 21941
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2631101707722532e-05,
+      "loss": 0.121,
+      "step": 21942
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2630765872416169e-05,
+      "loss": 0.1521,
+      "step": 21943
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2630430037109802e-05,
+      "loss": 0.361,
+      "step": 21944
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2630094201803436e-05,
+      "loss": 0.2835,
+      "step": 21945
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.262975836649707e-05,
+      "loss": 0.4106,
+      "step": 21946
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2629422531190705e-05,
+      "loss": 0.248,
+      "step": 21947
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2629086695884338e-05,
+      "loss": 0.2466,
+      "step": 21948
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2628750860577975e-05,
+      "loss": 0.3812,
+      "step": 21949
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2628415025271608e-05,
+      "loss": 0.2724,
+      "step": 21950
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.262807918996524e-05,
+      "loss": 0.2888,
+      "step": 21951
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2627743354658877e-05,
+      "loss": 0.351,
+      "step": 21952
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.262740751935251e-05,
+      "loss": 0.3615,
+      "step": 21953
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2627071684046145e-05,
+      "loss": 0.3062,
+      "step": 21954
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.262673584873978e-05,
+      "loss": 0.2517,
+      "step": 21955
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2626400013433412e-05,
+      "loss": 0.2248,
+      "step": 21956
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2626064178127047e-05,
+      "loss": 0.3641,
+      "step": 21957
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2625728342820682e-05,
+      "loss": 0.2043,
+      "step": 21958
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2625392507514316e-05,
+      "loss": 0.3513,
+      "step": 21959
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2625056672207949e-05,
+      "loss": 0.3972,
+      "step": 21960
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2624720836901585e-05,
+      "loss": 0.2775,
+      "step": 21961
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2624385001595218e-05,
+      "loss": 0.3409,
+      "step": 21962
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2624049166288851e-05,
+      "loss": 0.2872,
+      "step": 21963
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2623713330982488e-05,
+      "loss": 0.1705,
+      "step": 21964
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.262337749567612e-05,
+      "loss": 0.3568,
+      "step": 21965
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2623041660369755e-05,
+      "loss": 0.339,
+      "step": 21966
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.262270582506339e-05,
+      "loss": 0.1884,
+      "step": 21967
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2622369989757025e-05,
+      "loss": 0.3499,
+      "step": 21968
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2622034154450658e-05,
+      "loss": 0.3209,
+      "step": 21969
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2621698319144294e-05,
+      "loss": 0.2178,
+      "step": 21970
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2621362483837927e-05,
+      "loss": 0.4701,
+      "step": 21971
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.262102664853156e-05,
+      "loss": 0.5385,
+      "step": 21972
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2620690813225196e-05,
+      "loss": 0.2717,
+      "step": 21973
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2620354977918829e-05,
+      "loss": 0.2886,
+      "step": 21974
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2620019142612464e-05,
+      "loss": 0.3635,
+      "step": 21975
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2619683307306098e-05,
+      "loss": 0.1141,
+      "step": 21976
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2619347471999731e-05,
+      "loss": 0.3192,
+      "step": 21977
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2619011636693366e-05,
+      "loss": 0.3884,
+      "step": 21978
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2618675801387e-05,
+      "loss": 0.3677,
+      "step": 21979
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2618339966080635e-05,
+      "loss": 0.2707,
+      "step": 21980
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2618004130774268e-05,
+      "loss": 0.3066,
+      "step": 21981
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2617668295467905e-05,
+      "loss": 0.4249,
+      "step": 21982
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2617332460161538e-05,
+      "loss": 0.2231,
+      "step": 21983
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.261699662485517e-05,
+      "loss": 0.316,
+      "step": 21984
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2616660789548807e-05,
+      "loss": 0.2936,
+      "step": 21985
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.261632495424244e-05,
+      "loss": 0.3089,
+      "step": 21986
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2615989118936074e-05,
+      "loss": 0.1635,
+      "step": 21987
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2615653283629709e-05,
+      "loss": 0.3667,
+      "step": 21988
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2615317448323344e-05,
+      "loss": 0.3284,
+      "step": 21989
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2614981613016977e-05,
+      "loss": 0.1723,
+      "step": 21990
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2614645777710613e-05,
+      "loss": 0.3002,
+      "step": 21991
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2614309942404246e-05,
+      "loss": 0.2224,
+      "step": 21992
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2613974107097879e-05,
+      "loss": 0.4043,
+      "step": 21993
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2613638271791515e-05,
+      "loss": 0.2476,
+      "step": 21994
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2613302436485148e-05,
+      "loss": 0.2033,
+      "step": 21995
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2612966601178783e-05,
+      "loss": 0.2189,
+      "step": 21996
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2612630765872418e-05,
+      "loss": 0.2287,
+      "step": 21997
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2612294930566052e-05,
+      "loss": 0.234,
+      "step": 21998
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2611959095259685e-05,
+      "loss": 0.2,
+      "step": 21999
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.261162325995332e-05,
+      "loss": 0.3686,
+      "step": 22000
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2611287424646954e-05,
+      "loss": 0.3167,
+      "step": 22001
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2610951589340587e-05,
+      "loss": 0.3767,
+      "step": 22002
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2610615754034224e-05,
+      "loss": 0.3457,
+      "step": 22003
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2610279918727857e-05,
+      "loss": 0.2335,
+      "step": 22004
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.260994408342149e-05,
+      "loss": 0.1873,
+      "step": 22005
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2609608248115126e-05,
+      "loss": 0.3385,
+      "step": 22006
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2609272412808759e-05,
+      "loss": 0.3111,
+      "step": 22007
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2608936577502394e-05,
+      "loss": 0.4261,
+      "step": 22008
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2608600742196028e-05,
+      "loss": 0.3343,
+      "step": 22009
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2608264906889663e-05,
+      "loss": 0.3551,
+      "step": 22010
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2607929071583296e-05,
+      "loss": 0.2576,
+      "step": 22011
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2607593236276932e-05,
+      "loss": 0.2554,
+      "step": 22012
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2607257400970565e-05,
+      "loss": 0.2328,
+      "step": 22013
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2606921565664198e-05,
+      "loss": 0.2257,
+      "step": 22014
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2606585730357834e-05,
+      "loss": 0.3516,
+      "step": 22015
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2606249895051467e-05,
+      "loss": 0.3135,
+      "step": 22016
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2605914059745102e-05,
+      "loss": 0.3563,
+      "step": 22017
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2605578224438737e-05,
+      "loss": 0.1696,
+      "step": 22018
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2605242389132371e-05,
+      "loss": 0.2789,
+      "step": 22019
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2604906553826004e-05,
+      "loss": 0.3087,
+      "step": 22020
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.260457071851964e-05,
+      "loss": 0.2713,
+      "step": 22021
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2604234883213274e-05,
+      "loss": 0.253,
+      "step": 22022
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2603899047906907e-05,
+      "loss": 0.4741,
+      "step": 22023
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2603563212600543e-05,
+      "loss": 0.3259,
+      "step": 22024
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2603227377294176e-05,
+      "loss": 0.3291,
+      "step": 22025
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2602891541987809e-05,
+      "loss": 0.316,
+      "step": 22026
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2602555706681445e-05,
+      "loss": 0.1957,
+      "step": 22027
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2602219871375078e-05,
+      "loss": 0.2849,
+      "step": 22028
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2601884036068713e-05,
+      "loss": 0.2338,
+      "step": 22029
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2601548200762347e-05,
+      "loss": 0.2366,
+      "step": 22030
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2601212365455982e-05,
+      "loss": 0.3155,
+      "step": 22031
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2600876530149615e-05,
+      "loss": 0.254,
+      "step": 22032
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2600540694843251e-05,
+      "loss": 0.2773,
+      "step": 22033
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2600204859536884e-05,
+      "loss": 0.2961,
+      "step": 22034
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2599869024230517e-05,
+      "loss": 0.3245,
+      "step": 22035
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2599533188924154e-05,
+      "loss": 0.3558,
+      "step": 22036
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2599197353617786e-05,
+      "loss": 0.1929,
+      "step": 22037
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2598861518311421e-05,
+      "loss": 0.216,
+      "step": 22038
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2598525683005056e-05,
+      "loss": 0.1611,
+      "step": 22039
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.259818984769869e-05,
+      "loss": 0.1712,
+      "step": 22040
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2597854012392323e-05,
+      "loss": 0.2501,
+      "step": 22041
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.259751817708596e-05,
+      "loss": 0.3429,
+      "step": 22042
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2597182341779593e-05,
+      "loss": 0.2041,
+      "step": 22043
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2596846506473226e-05,
+      "loss": 0.3592,
+      "step": 22044
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2596510671166862e-05,
+      "loss": 0.3729,
+      "step": 22045
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2596174835860495e-05,
+      "loss": 0.2869,
+      "step": 22046
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2595839000554128e-05,
+      "loss": 0.2494,
+      "step": 22047
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2595503165247764e-05,
+      "loss": 0.4401,
+      "step": 22048
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2595167329941397e-05,
+      "loss": 0.3723,
+      "step": 22049
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2594831494635032e-05,
+      "loss": 0.2624,
+      "step": 22050
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2594495659328666e-05,
+      "loss": 0.3226,
+      "step": 22051
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2594159824022301e-05,
+      "loss": 0.1623,
+      "step": 22052
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2593823988715934e-05,
+      "loss": 0.2023,
+      "step": 22053
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.259348815340957e-05,
+      "loss": 0.3564,
+      "step": 22054
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2593152318103203e-05,
+      "loss": 0.273,
+      "step": 22055
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2592816482796836e-05,
+      "loss": 0.2403,
+      "step": 22056
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2592480647490473e-05,
+      "loss": 0.205,
+      "step": 22057
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2592144812184106e-05,
+      "loss": 0.252,
+      "step": 22058
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.259180897687774e-05,
+      "loss": 0.4398,
+      "step": 22059
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2591473141571375e-05,
+      "loss": 0.3997,
+      "step": 22060
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.259113730626501e-05,
+      "loss": 0.3845,
+      "step": 22061
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2590801470958643e-05,
+      "loss": 0.4586,
+      "step": 22062
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2590465635652279e-05,
+      "loss": 0.3758,
+      "step": 22063
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2590129800345912e-05,
+      "loss": 0.288,
+      "step": 22064
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2589793965039545e-05,
+      "loss": 0.2205,
+      "step": 22065
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2589458129733181e-05,
+      "loss": 0.3624,
+      "step": 22066
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2589122294426814e-05,
+      "loss": 0.1709,
+      "step": 22067
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2588786459120447e-05,
+      "loss": 0.249,
+      "step": 22068
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2588450623814083e-05,
+      "loss": 0.3553,
+      "step": 22069
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2588114788507716e-05,
+      "loss": 0.155,
+      "step": 22070
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2587778953201351e-05,
+      "loss": 0.2064,
+      "step": 22071
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2587443117894986e-05,
+      "loss": 0.2419,
+      "step": 22072
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.258710728258862e-05,
+      "loss": 0.4226,
+      "step": 22073
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2586771447282253e-05,
+      "loss": 0.2807,
+      "step": 22074
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.258643561197589e-05,
+      "loss": 0.2192,
+      "step": 22075
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2586099776669522e-05,
+      "loss": 0.2016,
+      "step": 22076
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2585763941363155e-05,
+      "loss": 0.1471,
+      "step": 22077
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2585428106056792e-05,
+      "loss": 0.3035,
+      "step": 22078
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2585092270750425e-05,
+      "loss": 0.1809,
+      "step": 22079
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.258475643544406e-05,
+      "loss": 0.3593,
+      "step": 22080
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2584420600137694e-05,
+      "loss": 0.355,
+      "step": 22081
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2584084764831329e-05,
+      "loss": 0.2913,
+      "step": 22082
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2583748929524962e-05,
+      "loss": 0.0994,
+      "step": 22083
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2583413094218598e-05,
+      "loss": 0.1533,
+      "step": 22084
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2583077258912231e-05,
+      "loss": 0.2934,
+      "step": 22085
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2582741423605864e-05,
+      "loss": 0.3083,
+      "step": 22086
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.25824055882995e-05,
+      "loss": 0.3598,
+      "step": 22087
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2582069752993133e-05,
+      "loss": 0.2758,
+      "step": 22088
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2581733917686768e-05,
+      "loss": 0.3747,
+      "step": 22089
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2581398082380402e-05,
+      "loss": 0.2752,
+      "step": 22090
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2581062247074035e-05,
+      "loss": 0.1966,
+      "step": 22091
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.258072641176767e-05,
+      "loss": 0.4216,
+      "step": 22092
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2580390576461305e-05,
+      "loss": 0.1975,
+      "step": 22093
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.258005474115494e-05,
+      "loss": 0.2316,
+      "step": 22094
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2579718905848572e-05,
+      "loss": 0.2803,
+      "step": 22095
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2579383070542209e-05,
+      "loss": 0.3707,
+      "step": 22096
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2579047235235842e-05,
+      "loss": 0.282,
+      "step": 22097
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2578711399929475e-05,
+      "loss": 0.2714,
+      "step": 22098
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2578375564623111e-05,
+      "loss": 0.3021,
+      "step": 22099
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2578039729316744e-05,
+      "loss": 0.3618,
+      "step": 22100
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2577703894010379e-05,
+      "loss": 0.311,
+      "step": 22101
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2577368058704013e-05,
+      "loss": 0.3238,
+      "step": 22102
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2577032223397648e-05,
+      "loss": 0.3726,
+      "step": 22103
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.257669638809128e-05,
+      "loss": 0.2752,
+      "step": 22104
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2576360552784917e-05,
+      "loss": 0.4106,
+      "step": 22105
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.257602471747855e-05,
+      "loss": 0.3713,
+      "step": 22106
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2575688882172183e-05,
+      "loss": 0.3103,
+      "step": 22107
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2575353046865818e-05,
+      "loss": 0.1912,
+      "step": 22108
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2575017211559452e-05,
+      "loss": 0.1904,
+      "step": 22109
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2574681376253087e-05,
+      "loss": 0.2533,
+      "step": 22110
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.257434554094672e-05,
+      "loss": 0.421,
+      "step": 22111
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2574009705640355e-05,
+      "loss": 0.2691,
+      "step": 22112
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.257367387033399e-05,
+      "loss": 0.2163,
+      "step": 22113
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2573338035027622e-05,
+      "loss": 0.2041,
+      "step": 22114
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2573002199721258e-05,
+      "loss": 0.5479,
+      "step": 22115
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2572666364414891e-05,
+      "loss": 0.432,
+      "step": 22116
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2572330529108524e-05,
+      "loss": 0.3775,
+      "step": 22117
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.257199469380216e-05,
+      "loss": 0.4145,
+      "step": 22118
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2571658858495794e-05,
+      "loss": 0.2693,
+      "step": 22119
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2571323023189428e-05,
+      "loss": 0.3835,
+      "step": 22120
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2570987187883063e-05,
+      "loss": 0.3605,
+      "step": 22121
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2570651352576698e-05,
+      "loss": 0.2214,
+      "step": 22122
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.257031551727033e-05,
+      "loss": 0.4023,
+      "step": 22123
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2569979681963967e-05,
+      "loss": 0.3045,
+      "step": 22124
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.25696438466576e-05,
+      "loss": 0.2945,
+      "step": 22125
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2569308011351233e-05,
+      "loss": 0.3059,
+      "step": 22126
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.256897217604487e-05,
+      "loss": 0.4201,
+      "step": 22127
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2568636340738502e-05,
+      "loss": 0.3119,
+      "step": 22128
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2568300505432137e-05,
+      "loss": 0.4036,
+      "step": 22129
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2567964670125771e-05,
+      "loss": 0.3516,
+      "step": 22130
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2567628834819406e-05,
+      "loss": 0.2634,
+      "step": 22131
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2567292999513039e-05,
+      "loss": 0.4975,
+      "step": 22132
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 1.2566957164206675e-05,
+      "loss": 0.3969,
+      "step": 22133
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2566621328900308e-05,
+      "loss": 0.3959,
+      "step": 22134
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2566285493593941e-05,
+      "loss": 0.1639,
+      "step": 22135
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2565949658287578e-05,
+      "loss": 0.2202,
+      "step": 22136
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.256561382298121e-05,
+      "loss": 0.4476,
+      "step": 22137
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2565277987674844e-05,
+      "loss": 0.218,
+      "step": 22138
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.256494215236848e-05,
+      "loss": 0.2253,
+      "step": 22139
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2564606317062113e-05,
+      "loss": 0.245,
+      "step": 22140
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2564270481755747e-05,
+      "loss": 0.4096,
+      "step": 22141
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2563934646449382e-05,
+      "loss": 0.1739,
+      "step": 22142
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2563598811143017e-05,
+      "loss": 0.3443,
+      "step": 22143
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.256326297583665e-05,
+      "loss": 0.1757,
+      "step": 22144
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2562927140530286e-05,
+      "loss": 0.2618,
+      "step": 22145
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2562591305223919e-05,
+      "loss": 0.1718,
+      "step": 22146
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2562255469917552e-05,
+      "loss": 0.4416,
+      "step": 22147
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2561919634611188e-05,
+      "loss": 0.366,
+      "step": 22148
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2561583799304821e-05,
+      "loss": 0.1879,
+      "step": 22149
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2561247963998456e-05,
+      "loss": 0.203,
+      "step": 22150
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.256091212869209e-05,
+      "loss": 0.2507,
+      "step": 22151
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2560576293385725e-05,
+      "loss": 0.1719,
+      "step": 22152
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2560240458079358e-05,
+      "loss": 0.2315,
+      "step": 22153
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2559904622772994e-05,
+      "loss": 0.2721,
+      "step": 22154
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2559568787466627e-05,
+      "loss": 0.3211,
+      "step": 22155
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.255923295216026e-05,
+      "loss": 0.3369,
+      "step": 22156
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2558897116853897e-05,
+      "loss": 0.1802,
+      "step": 22157
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.255856128154753e-05,
+      "loss": 0.2931,
+      "step": 22158
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2558225446241163e-05,
+      "loss": 0.2348,
+      "step": 22159
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2557889610934799e-05,
+      "loss": 0.3158,
+      "step": 22160
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2557553775628432e-05,
+      "loss": 0.3911,
+      "step": 22161
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2557217940322067e-05,
+      "loss": 0.3589,
+      "step": 22162
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2556882105015701e-05,
+      "loss": 0.2773,
+      "step": 22163
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2556546269709336e-05,
+      "loss": 0.4134,
+      "step": 22164
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2556210434402969e-05,
+      "loss": 0.1662,
+      "step": 22165
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2555874599096605e-05,
+      "loss": 0.1871,
+      "step": 22166
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2555538763790238e-05,
+      "loss": 0.3589,
+      "step": 22167
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2555202928483871e-05,
+      "loss": 0.3343,
+      "step": 22168
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2554867093177507e-05,
+      "loss": 0.2497,
+      "step": 22169
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.255453125787114e-05,
+      "loss": 0.211,
+      "step": 22170
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2554195422564775e-05,
+      "loss": 0.2252,
+      "step": 22171
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.255385958725841e-05,
+      "loss": 0.1503,
+      "step": 22172
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2553523751952044e-05,
+      "loss": 0.3441,
+      "step": 22173
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2553187916645677e-05,
+      "loss": 0.1512,
+      "step": 22174
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2552852081339314e-05,
+      "loss": 0.3441,
+      "step": 22175
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2552516246032947e-05,
+      "loss": 0.2998,
+      "step": 22176
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.255218041072658e-05,
+      "loss": 0.5207,
+      "step": 22177
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2551844575420216e-05,
+      "loss": 0.4181,
+      "step": 22178
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2551508740113849e-05,
+      "loss": 0.2233,
+      "step": 22179
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2551172904807483e-05,
+      "loss": 0.3653,
+      "step": 22180
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2550837069501118e-05,
+      "loss": 0.227,
+      "step": 22181
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2550501234194751e-05,
+      "loss": 0.1321,
+      "step": 22182
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2550165398888386e-05,
+      "loss": 0.3093,
+      "step": 22183
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.254982956358202e-05,
+      "loss": 0.4288,
+      "step": 22184
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2549493728275655e-05,
+      "loss": 0.4153,
+      "step": 22185
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2549157892969288e-05,
+      "loss": 0.2606,
+      "step": 22186
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2548822057662924e-05,
+      "loss": 0.2631,
+      "step": 22187
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2548486222356557e-05,
+      "loss": 0.3021,
+      "step": 22188
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.254815038705019e-05,
+      "loss": 0.2925,
+      "step": 22189
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2547814551743827e-05,
+      "loss": 0.2041,
+      "step": 22190
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.254747871643746e-05,
+      "loss": 0.2296,
+      "step": 22191
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2547142881131094e-05,
+      "loss": 0.3092,
+      "step": 22192
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2546807045824729e-05,
+      "loss": 0.3218,
+      "step": 22193
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2546471210518363e-05,
+      "loss": 0.2775,
+      "step": 22194
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2546135375211996e-05,
+      "loss": 0.2908,
+      "step": 22195
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2545799539905633e-05,
+      "loss": 0.2172,
+      "step": 22196
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2545463704599266e-05,
+      "loss": 0.4815,
+      "step": 22197
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2545127869292899e-05,
+      "loss": 0.3723,
+      "step": 22198
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2544792033986535e-05,
+      "loss": 0.3481,
+      "step": 22199
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2544456198680168e-05,
+      "loss": 0.415,
+      "step": 22200
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2544120363373803e-05,
+      "loss": 0.2985,
+      "step": 22201
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2543784528067437e-05,
+      "loss": 0.3084,
+      "step": 22202
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.254344869276107e-05,
+      "loss": 0.2516,
+      "step": 22203
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2543112857454705e-05,
+      "loss": 0.2621,
+      "step": 22204
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.254277702214834e-05,
+      "loss": 0.3572,
+      "step": 22205
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2542441186841974e-05,
+      "loss": 0.3042,
+      "step": 22206
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2542105351535607e-05,
+      "loss": 0.3129,
+      "step": 22207
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2541769516229243e-05,
+      "loss": 0.32,
+      "step": 22208
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2541433680922876e-05,
+      "loss": 0.2175,
+      "step": 22209
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.254109784561651e-05,
+      "loss": 0.381,
+      "step": 22210
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2540762010310146e-05,
+      "loss": 0.2105,
+      "step": 22211
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2540426175003779e-05,
+      "loss": 0.4315,
+      "step": 22212
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2540090339697413e-05,
+      "loss": 0.2842,
+      "step": 22213
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2539754504391048e-05,
+      "loss": 0.242,
+      "step": 22214
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2539418669084683e-05,
+      "loss": 0.3335,
+      "step": 22215
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2539082833778316e-05,
+      "loss": 0.4253,
+      "step": 22216
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2538746998471952e-05,
+      "loss": 0.195,
+      "step": 22217
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2538411163165585e-05,
+      "loss": 0.2692,
+      "step": 22218
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2538075327859218e-05,
+      "loss": 0.2102,
+      "step": 22219
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2537739492552854e-05,
+      "loss": 0.2659,
+      "step": 22220
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2537403657246487e-05,
+      "loss": 0.2052,
+      "step": 22221
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2537067821940122e-05,
+      "loss": 0.4726,
+      "step": 22222
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2536731986633756e-05,
+      "loss": 0.2341,
+      "step": 22223
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2536396151327391e-05,
+      "loss": 0.3718,
+      "step": 22224
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2536060316021024e-05,
+      "loss": 0.3032,
+      "step": 22225
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2535724480714659e-05,
+      "loss": 0.236,
+      "step": 22226
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2535388645408293e-05,
+      "loss": 0.2824,
+      "step": 22227
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2535052810101926e-05,
+      "loss": 0.2699,
+      "step": 22228
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2534716974795563e-05,
+      "loss": 0.2618,
+      "step": 22229
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2534381139489196e-05,
+      "loss": 0.3104,
+      "step": 22230
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2534045304182828e-05,
+      "loss": 0.2532,
+      "step": 22231
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2533709468876465e-05,
+      "loss": 0.2295,
+      "step": 22232
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2533373633570098e-05,
+      "loss": 0.2475,
+      "step": 22233
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2533037798263732e-05,
+      "loss": 0.1972,
+      "step": 22234
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2532701962957367e-05,
+      "loss": 0.3007,
+      "step": 22235
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2532366127651002e-05,
+      "loss": 0.3883,
+      "step": 22236
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2532030292344635e-05,
+      "loss": 0.1979,
+      "step": 22237
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2531694457038271e-05,
+      "loss": 0.2056,
+      "step": 22238
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2531358621731904e-05,
+      "loss": 0.2383,
+      "step": 22239
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2531022786425537e-05,
+      "loss": 0.2493,
+      "step": 22240
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2530686951119173e-05,
+      "loss": 0.2326,
+      "step": 22241
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2530351115812806e-05,
+      "loss": 0.39,
+      "step": 22242
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.253001528050644e-05,
+      "loss": 0.2735,
+      "step": 22243
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2529679445200075e-05,
+      "loss": 0.3682,
+      "step": 22244
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.252934360989371e-05,
+      "loss": 0.3163,
+      "step": 22245
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2529007774587343e-05,
+      "loss": 0.1704,
+      "step": 22246
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.252867193928098e-05,
+      "loss": 0.2773,
+      "step": 22247
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2528336103974612e-05,
+      "loss": 0.1871,
+      "step": 22248
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2528000268668245e-05,
+      "loss": 0.4179,
+      "step": 22249
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2527664433361882e-05,
+      "loss": 0.2821,
+      "step": 22250
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2527328598055515e-05,
+      "loss": 0.1921,
+      "step": 22251
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2526992762749148e-05,
+      "loss": 0.2948,
+      "step": 22252
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2526656927442784e-05,
+      "loss": 0.1132,
+      "step": 22253
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2526321092136417e-05,
+      "loss": 0.4132,
+      "step": 22254
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2525985256830052e-05,
+      "loss": 0.3873,
+      "step": 22255
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2525649421523686e-05,
+      "loss": 0.3613,
+      "step": 22256
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.252531358621732e-05,
+      "loss": 0.2907,
+      "step": 22257
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2524977750910954e-05,
+      "loss": 0.3634,
+      "step": 22258
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.252464191560459e-05,
+      "loss": 0.1706,
+      "step": 22259
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2524306080298223e-05,
+      "loss": 0.3132,
+      "step": 22260
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2523970244991856e-05,
+      "loss": 0.3885,
+      "step": 22261
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2523634409685492e-05,
+      "loss": 0.2501,
+      "step": 22262
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2523298574379125e-05,
+      "loss": 0.3074,
+      "step": 22263
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.252296273907276e-05,
+      "loss": 0.3127,
+      "step": 22264
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2522626903766395e-05,
+      "loss": 0.406,
+      "step": 22265
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.252229106846003e-05,
+      "loss": 0.3732,
+      "step": 22266
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2521955233153662e-05,
+      "loss": 0.3197,
+      "step": 22267
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2521619397847299e-05,
+      "loss": 0.2508,
+      "step": 22268
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2521283562540932e-05,
+      "loss": 0.4161,
+      "step": 22269
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2520947727234564e-05,
+      "loss": 0.3191,
+      "step": 22270
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.25206118919282e-05,
+      "loss": 0.2285,
+      "step": 22271
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2520276056621834e-05,
+      "loss": 0.2344,
+      "step": 22272
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2519940221315467e-05,
+      "loss": 0.2146,
+      "step": 22273
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2519604386009103e-05,
+      "loss": 0.2798,
+      "step": 22274
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2519268550702736e-05,
+      "loss": 0.1858,
+      "step": 22275
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.251893271539637e-05,
+      "loss": 0.3426,
+      "step": 22276
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2518596880090005e-05,
+      "loss": 0.5017,
+      "step": 22277
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.251826104478364e-05,
+      "loss": 0.2628,
+      "step": 22278
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2517925209477273e-05,
+      "loss": 0.161,
+      "step": 22279
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.251758937417091e-05,
+      "loss": 0.1947,
+      "step": 22280
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2517253538864542e-05,
+      "loss": 0.2891,
+      "step": 22281
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2516917703558175e-05,
+      "loss": 0.3374,
+      "step": 22282
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2516581868251811e-05,
+      "loss": 0.4342,
+      "step": 22283
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2516246032945444e-05,
+      "loss": 0.2047,
+      "step": 22284
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2515910197639079e-05,
+      "loss": 0.2696,
+      "step": 22285
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2515574362332714e-05,
+      "loss": 0.4033,
+      "step": 22286
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2515238527026348e-05,
+      "loss": 0.1955,
+      "step": 22287
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2514902691719981e-05,
+      "loss": 0.4329,
+      "step": 22288
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2514566856413614e-05,
+      "loss": 0.3152,
+      "step": 22289
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.251423102110725e-05,
+      "loss": 0.3885,
+      "step": 22290
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2513895185800884e-05,
+      "loss": 0.1948,
+      "step": 22291
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2513559350494518e-05,
+      "loss": 0.3483,
+      "step": 22292
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2513223515188153e-05,
+      "loss": 0.2274,
+      "step": 22293
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2512887679881786e-05,
+      "loss": 0.3048,
+      "step": 22294
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.251255184457542e-05,
+      "loss": 0.1678,
+      "step": 22295
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2512216009269055e-05,
+      "loss": 0.2148,
+      "step": 22296
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.251188017396269e-05,
+      "loss": 0.1984,
+      "step": 22297
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2511544338656323e-05,
+      "loss": 0.4294,
+      "step": 22298
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2511208503349959e-05,
+      "loss": 0.3034,
+      "step": 22299
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2510872668043592e-05,
+      "loss": 0.5473,
+      "step": 22300
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2510536832737225e-05,
+      "loss": 0.3395,
+      "step": 22301
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2510200997430861e-05,
+      "loss": 0.2673,
+      "step": 22302
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2509865162124494e-05,
+      "loss": 0.2147,
+      "step": 22303
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2509529326818129e-05,
+      "loss": 0.3793,
+      "step": 22304
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2509193491511764e-05,
+      "loss": 0.1977,
+      "step": 22305
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2508857656205398e-05,
+      "loss": 0.2092,
+      "step": 22306
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2508521820899031e-05,
+      "loss": 0.3766,
+      "step": 22307
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2508185985592668e-05,
+      "loss": 0.3065,
+      "step": 22308
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.25078501502863e-05,
+      "loss": 0.2697,
+      "step": 22309
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2507514314979933e-05,
+      "loss": 0.3377,
+      "step": 22310
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.250717847967357e-05,
+      "loss": 0.3775,
+      "step": 22311
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2506842644367203e-05,
+      "loss": 0.2368,
+      "step": 22312
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2506506809060837e-05,
+      "loss": 0.3718,
+      "step": 22313
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2506170973754472e-05,
+      "loss": 0.2855,
+      "step": 22314
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2505835138448107e-05,
+      "loss": 0.4947,
+      "step": 22315
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.250549930314174e-05,
+      "loss": 0.4138,
+      "step": 22316
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2505163467835374e-05,
+      "loss": 0.368,
+      "step": 22317
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2504827632529009e-05,
+      "loss": 0.3447,
+      "step": 22318
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2504491797222642e-05,
+      "loss": 0.2076,
+      "step": 22319
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2504155961916278e-05,
+      "loss": 0.3114,
+      "step": 22320
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2503820126609911e-05,
+      "loss": 0.2035,
+      "step": 22321
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2503484291303544e-05,
+      "loss": 0.3954,
+      "step": 22322
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.250314845599718e-05,
+      "loss": 0.2164,
+      "step": 22323
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2502812620690813e-05,
+      "loss": 0.3844,
+      "step": 22324
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2502476785384448e-05,
+      "loss": 0.2841,
+      "step": 22325
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2502140950078083e-05,
+      "loss": 0.3405,
+      "step": 22326
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2501805114771717e-05,
+      "loss": 0.2145,
+      "step": 22327
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.250146927946535e-05,
+      "loss": 0.2351,
+      "step": 22328
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2501133444158987e-05,
+      "loss": 0.2764,
+      "step": 22329
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.250079760885262e-05,
+      "loss": 0.3725,
+      "step": 22330
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2500461773546253e-05,
+      "loss": 0.3662,
+      "step": 22331
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.2500125938239889e-05,
+      "loss": 0.2003,
+      "step": 22332
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2499790102933522e-05,
+      "loss": 0.2932,
+      "step": 22333
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2499454267627156e-05,
+      "loss": 0.3172,
+      "step": 22334
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2499118432320791e-05,
+      "loss": 0.2368,
+      "step": 22335
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2498782597014426e-05,
+      "loss": 0.3046,
+      "step": 22336
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2498446761708059e-05,
+      "loss": 0.4512,
+      "step": 22337
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2498110926401693e-05,
+      "loss": 0.2565,
+      "step": 22338
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2497775091095328e-05,
+      "loss": 0.1742,
+      "step": 22339
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2497439255788961e-05,
+      "loss": 0.2282,
+      "step": 22340
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2497103420482597e-05,
+      "loss": 0.2115,
+      "step": 22341
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.249676758517623e-05,
+      "loss": 0.3416,
+      "step": 22342
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2496431749869863e-05,
+      "loss": 0.2833,
+      "step": 22343
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.24960959145635e-05,
+      "loss": 0.3209,
+      "step": 22344
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2495760079257133e-05,
+      "loss": 0.1622,
+      "step": 22345
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2495424243950767e-05,
+      "loss": 0.2237,
+      "step": 22346
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2495088408644402e-05,
+      "loss": 0.3637,
+      "step": 22347
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2494752573338036e-05,
+      "loss": 0.2514,
+      "step": 22348
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.249441673803167e-05,
+      "loss": 0.1843,
+      "step": 22349
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2494080902725306e-05,
+      "loss": 0.277,
+      "step": 22350
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2493745067418939e-05,
+      "loss": 0.3675,
+      "step": 22351
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2493409232112572e-05,
+      "loss": 0.3465,
+      "step": 22352
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2493073396806208e-05,
+      "loss": 0.2227,
+      "step": 22353
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2492737561499841e-05,
+      "loss": 0.3212,
+      "step": 22354
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2492401726193476e-05,
+      "loss": 0.2466,
+      "step": 22355
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.249206589088711e-05,
+      "loss": 0.3417,
+      "step": 22356
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2491730055580745e-05,
+      "loss": 0.3371,
+      "step": 22357
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2491394220274378e-05,
+      "loss": 0.2772,
+      "step": 22358
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2491058384968014e-05,
+      "loss": 0.3008,
+      "step": 22359
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2490722549661647e-05,
+      "loss": 0.2007,
+      "step": 22360
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.249038671435528e-05,
+      "loss": 0.2423,
+      "step": 22361
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2490050879048916e-05,
+      "loss": 0.3056,
+      "step": 22362
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.248971504374255e-05,
+      "loss": 0.4263,
+      "step": 22363
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2489379208436182e-05,
+      "loss": 0.2644,
+      "step": 22364
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2489043373129819e-05,
+      "loss": 0.2622,
+      "step": 22365
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2488707537823452e-05,
+      "loss": 0.2549,
+      "step": 22366
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2488371702517086e-05,
+      "loss": 0.2573,
+      "step": 22367
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2488035867210721e-05,
+      "loss": 0.2294,
+      "step": 22368
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2487700031904356e-05,
+      "loss": 0.3077,
+      "step": 22369
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2487364196597989e-05,
+      "loss": 0.4246,
+      "step": 22370
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2487028361291625e-05,
+      "loss": 0.3402,
+      "step": 22371
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2486692525985258e-05,
+      "loss": 0.295,
+      "step": 22372
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.248635669067889e-05,
+      "loss": 0.2891,
+      "step": 22373
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2486020855372527e-05,
+      "loss": 0.2144,
+      "step": 22374
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.248568502006616e-05,
+      "loss": 0.4015,
+      "step": 22375
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2485349184759795e-05,
+      "loss": 0.2131,
+      "step": 22376
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.248501334945343e-05,
+      "loss": 0.4716,
+      "step": 22377
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2484677514147064e-05,
+      "loss": 0.3341,
+      "step": 22378
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2484341678840697e-05,
+      "loss": 0.252,
+      "step": 22379
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2484005843534333e-05,
+      "loss": 0.4553,
+      "step": 22380
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2483670008227966e-05,
+      "loss": 0.5145,
+      "step": 22381
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.24833341729216e-05,
+      "loss": 0.2487,
+      "step": 22382
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2482998337615236e-05,
+      "loss": 0.3018,
+      "step": 22383
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2482662502308869e-05,
+      "loss": 0.2555,
+      "step": 22384
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2482326667002501e-05,
+      "loss": 0.2245,
+      "step": 22385
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2481990831696138e-05,
+      "loss": 0.3833,
+      "step": 22386
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.248165499638977e-05,
+      "loss": 0.3539,
+      "step": 22387
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2481319161083405e-05,
+      "loss": 0.2429,
+      "step": 22388
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.248098332577704e-05,
+      "loss": 0.1602,
+      "step": 22389
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2480647490470675e-05,
+      "loss": 0.3322,
+      "step": 22390
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2480311655164308e-05,
+      "loss": 0.3799,
+      "step": 22391
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2479975819857944e-05,
+      "loss": 0.3251,
+      "step": 22392
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2479639984551577e-05,
+      "loss": 0.3263,
+      "step": 22393
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.247930414924521e-05,
+      "loss": 0.3782,
+      "step": 22394
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2478968313938846e-05,
+      "loss": 0.4131,
+      "step": 22395
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.247863247863248e-05,
+      "loss": 0.3749,
+      "step": 22396
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2478296643326114e-05,
+      "loss": 0.2745,
+      "step": 22397
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2477960808019748e-05,
+      "loss": 0.3196,
+      "step": 22398
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2477624972713383e-05,
+      "loss": 0.3024,
+      "step": 22399
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2477289137407016e-05,
+      "loss": 0.2709,
+      "step": 22400
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2476953302100652e-05,
+      "loss": 0.2722,
+      "step": 22401
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2476617466794285e-05,
+      "loss": 0.2703,
+      "step": 22402
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2476281631487918e-05,
+      "loss": 0.3785,
+      "step": 22403
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2475945796181555e-05,
+      "loss": 0.2305,
+      "step": 22404
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2475609960875188e-05,
+      "loss": 0.2638,
+      "step": 22405
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2475274125568822e-05,
+      "loss": 0.2468,
+      "step": 22406
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2474938290262457e-05,
+      "loss": 0.2978,
+      "step": 22407
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.247460245495609e-05,
+      "loss": 0.2679,
+      "step": 22408
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2474266619649725e-05,
+      "loss": 0.1939,
+      "step": 22409
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.247393078434336e-05,
+      "loss": 0.3489,
+      "step": 22410
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2473594949036994e-05,
+      "loss": 0.3907,
+      "step": 22411
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2473259113730627e-05,
+      "loss": 0.2776,
+      "step": 22412
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2472923278424263e-05,
+      "loss": 0.4321,
+      "step": 22413
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2472587443117896e-05,
+      "loss": 0.3324,
+      "step": 22414
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2472251607811529e-05,
+      "loss": 0.193,
+      "step": 22415
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2471915772505165e-05,
+      "loss": 0.3592,
+      "step": 22416
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2471579937198798e-05,
+      "loss": 0.3626,
+      "step": 22417
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2471244101892433e-05,
+      "loss": 0.1723,
+      "step": 22418
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2470908266586068e-05,
+      "loss": 0.2294,
+      "step": 22419
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2470572431279702e-05,
+      "loss": 0.2975,
+      "step": 22420
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2470236595973335e-05,
+      "loss": 0.4523,
+      "step": 22421
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2469900760666972e-05,
+      "loss": 0.1569,
+      "step": 22422
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2469564925360605e-05,
+      "loss": 0.4827,
+      "step": 22423
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2469229090054237e-05,
+      "loss": 0.2191,
+      "step": 22424
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2468893254747874e-05,
+      "loss": 0.3804,
+      "step": 22425
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2468557419441507e-05,
+      "loss": 0.3454,
+      "step": 22426
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2468221584135141e-05,
+      "loss": 0.2275,
+      "step": 22427
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2467885748828776e-05,
+      "loss": 0.2413,
+      "step": 22428
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2467549913522409e-05,
+      "loss": 0.2936,
+      "step": 22429
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2467214078216044e-05,
+      "loss": 0.3129,
+      "step": 22430
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2466878242909678e-05,
+      "loss": 0.1441,
+      "step": 22431
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2466542407603313e-05,
+      "loss": 0.2349,
+      "step": 22432
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2466206572296946e-05,
+      "loss": 0.2975,
+      "step": 22433
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2465870736990582e-05,
+      "loss": 0.2385,
+      "step": 22434
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2465534901684215e-05,
+      "loss": 0.3501,
+      "step": 22435
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2465199066377848e-05,
+      "loss": 0.416,
+      "step": 22436
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2464863231071485e-05,
+      "loss": 0.3603,
+      "step": 22437
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2464527395765117e-05,
+      "loss": 0.3051,
+      "step": 22438
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2464191560458752e-05,
+      "loss": 0.1548,
+      "step": 22439
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2463855725152387e-05,
+      "loss": 0.2534,
+      "step": 22440
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2463519889846021e-05,
+      "loss": 0.3509,
+      "step": 22441
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2463184054539654e-05,
+      "loss": 0.2866,
+      "step": 22442
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.246284821923329e-05,
+      "loss": 0.2677,
+      "step": 22443
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2462512383926924e-05,
+      "loss": 0.3572,
+      "step": 22444
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2462176548620557e-05,
+      "loss": 0.1859,
+      "step": 22445
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2461840713314193e-05,
+      "loss": 0.2723,
+      "step": 22446
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2461504878007826e-05,
+      "loss": 0.4047,
+      "step": 22447
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.246116904270146e-05,
+      "loss": 0.4022,
+      "step": 22448
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2460833207395095e-05,
+      "loss": 0.3497,
+      "step": 22449
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.246049737208873e-05,
+      "loss": 0.2888,
+      "step": 22450
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2460161536782363e-05,
+      "loss": 0.3329,
+      "step": 22451
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2459825701475997e-05,
+      "loss": 0.4675,
+      "step": 22452
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2459489866169632e-05,
+      "loss": 0.2945,
+      "step": 22453
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2459154030863265e-05,
+      "loss": 0.2161,
+      "step": 22454
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2458818195556901e-05,
+      "loss": 0.2409,
+      "step": 22455
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2458482360250534e-05,
+      "loss": 0.2801,
+      "step": 22456
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2458146524944167e-05,
+      "loss": 0.5391,
+      "step": 22457
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2457810689637804e-05,
+      "loss": 0.3418,
+      "step": 22458
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2457474854331437e-05,
+      "loss": 0.463,
+      "step": 22459
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2457139019025071e-05,
+      "loss": 0.5082,
+      "step": 22460
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2456803183718706e-05,
+      "loss": 0.2833,
+      "step": 22461
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.245646734841234e-05,
+      "loss": 0.3564,
+      "step": 22462
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2456131513105973e-05,
+      "loss": 0.3226,
+      "step": 22463
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.245579567779961e-05,
+      "loss": 0.3398,
+      "step": 22464
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2455459842493243e-05,
+      "loss": 0.1683,
+      "step": 22465
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2455124007186876e-05,
+      "loss": 0.2775,
+      "step": 22466
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2454788171880512e-05,
+      "loss": 0.2601,
+      "step": 22467
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2454452336574145e-05,
+      "loss": 0.4682,
+      "step": 22468
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.245411650126778e-05,
+      "loss": 0.2428,
+      "step": 22469
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2453780665961413e-05,
+      "loss": 0.215,
+      "step": 22470
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2453444830655049e-05,
+      "loss": 0.369,
+      "step": 22471
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2453108995348682e-05,
+      "loss": 0.2181,
+      "step": 22472
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2452773160042315e-05,
+      "loss": 0.2978,
+      "step": 22473
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2452437324735951e-05,
+      "loss": 0.2246,
+      "step": 22474
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2452101489429584e-05,
+      "loss": 0.1636,
+      "step": 22475
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2451765654123217e-05,
+      "loss": 0.2751,
+      "step": 22476
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2451429818816853e-05,
+      "loss": 0.2639,
+      "step": 22477
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2451093983510486e-05,
+      "loss": 0.2706,
+      "step": 22478
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2450758148204121e-05,
+      "loss": 0.334,
+      "step": 22479
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2450422312897756e-05,
+      "loss": 0.3772,
+      "step": 22480
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.245008647759139e-05,
+      "loss": 0.2909,
+      "step": 22481
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2449750642285023e-05,
+      "loss": 0.3707,
+      "step": 22482
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.244941480697866e-05,
+      "loss": 0.2362,
+      "step": 22483
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2449078971672293e-05,
+      "loss": 0.4929,
+      "step": 22484
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2448743136365926e-05,
+      "loss": 0.3213,
+      "step": 22485
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2448407301059562e-05,
+      "loss": 0.3917,
+      "step": 22486
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2448071465753195e-05,
+      "loss": 0.235,
+      "step": 22487
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.244773563044683e-05,
+      "loss": 0.4722,
+      "step": 22488
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2447399795140464e-05,
+      "loss": 0.4185,
+      "step": 22489
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2447063959834099e-05,
+      "loss": 0.2883,
+      "step": 22490
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2446728124527732e-05,
+      "loss": 0.458,
+      "step": 22491
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2446392289221368e-05,
+      "loss": 0.2891,
+      "step": 22492
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2446056453915001e-05,
+      "loss": 0.2332,
+      "step": 22493
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2445720618608634e-05,
+      "loss": 0.263,
+      "step": 22494
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.244538478330227e-05,
+      "loss": 0.292,
+      "step": 22495
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2445048947995903e-05,
+      "loss": 0.4217,
+      "step": 22496
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2444713112689536e-05,
+      "loss": 0.3491,
+      "step": 22497
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2444377277383173e-05,
+      "loss": 0.4007,
+      "step": 22498
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2444041442076806e-05,
+      "loss": 0.283,
+      "step": 22499
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.244370560677044e-05,
+      "loss": 0.3167,
+      "step": 22500
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2443369771464075e-05,
+      "loss": 0.3513,
+      "step": 22501
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.244303393615771e-05,
+      "loss": 0.2162,
+      "step": 22502
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2442698100851342e-05,
+      "loss": 0.1931,
+      "step": 22503
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2442362265544979e-05,
+      "loss": 0.1971,
+      "step": 22504
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2442026430238612e-05,
+      "loss": 0.2471,
+      "step": 22505
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2441690594932245e-05,
+      "loss": 0.3494,
+      "step": 22506
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2441354759625881e-05,
+      "loss": 0.2098,
+      "step": 22507
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2441018924319514e-05,
+      "loss": 0.4341,
+      "step": 22508
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2440683089013149e-05,
+      "loss": 0.3401,
+      "step": 22509
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2440347253706783e-05,
+      "loss": 0.345,
+      "step": 22510
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2440011418400418e-05,
+      "loss": 0.3977,
+      "step": 22511
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2439675583094051e-05,
+      "loss": 0.347,
+      "step": 22512
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2439339747787687e-05,
+      "loss": 0.3201,
+      "step": 22513
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.243900391248132e-05,
+      "loss": 0.3614,
+      "step": 22514
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2438668077174953e-05,
+      "loss": 0.2393,
+      "step": 22515
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.243833224186859e-05,
+      "loss": 0.2262,
+      "step": 22516
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2437996406562222e-05,
+      "loss": 0.3279,
+      "step": 22517
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2437660571255857e-05,
+      "loss": 0.3153,
+      "step": 22518
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2437324735949492e-05,
+      "loss": 0.3489,
+      "step": 22519
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2436988900643125e-05,
+      "loss": 0.4631,
+      "step": 22520
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.243665306533676e-05,
+      "loss": 0.3074,
+      "step": 22521
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2436317230030394e-05,
+      "loss": 0.3347,
+      "step": 22522
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2435981394724029e-05,
+      "loss": 0.1912,
+      "step": 22523
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2435645559417662e-05,
+      "loss": 0.4018,
+      "step": 22524
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2435309724111298e-05,
+      "loss": 0.3383,
+      "step": 22525
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2434973888804931e-05,
+      "loss": 0.3142,
+      "step": 22526
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2434638053498564e-05,
+      "loss": 0.3043,
+      "step": 22527
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.24343022181922e-05,
+      "loss": 0.3323,
+      "step": 22528
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2433966382885833e-05,
+      "loss": 0.2695,
+      "step": 22529
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 1.2433630547579468e-05,
+      "loss": 0.2112,
+      "step": 22530
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2433294712273102e-05,
+      "loss": 0.3924,
+      "step": 22531
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2432958876966737e-05,
+      "loss": 0.2961,
+      "step": 22532
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.243262304166037e-05,
+      "loss": 0.3182,
+      "step": 22533
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2432287206354006e-05,
+      "loss": 0.3834,
+      "step": 22534
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.243195137104764e-05,
+      "loss": 0.2732,
+      "step": 22535
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2431615535741272e-05,
+      "loss": 0.2547,
+      "step": 22536
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2431279700434909e-05,
+      "loss": 0.2462,
+      "step": 22537
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2430943865128542e-05,
+      "loss": 0.3212,
+      "step": 22538
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2430608029822176e-05,
+      "loss": 0.2289,
+      "step": 22539
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.243027219451581e-05,
+      "loss": 0.2108,
+      "step": 22540
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2429936359209445e-05,
+      "loss": 0.3261,
+      "step": 22541
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2429600523903078e-05,
+      "loss": 0.321,
+      "step": 22542
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2429264688596713e-05,
+      "loss": 0.279,
+      "step": 22543
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2428928853290348e-05,
+      "loss": 0.5678,
+      "step": 22544
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.242859301798398e-05,
+      "loss": 0.3457,
+      "step": 22545
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2428257182677617e-05,
+      "loss": 0.2997,
+      "step": 22546
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.242792134737125e-05,
+      "loss": 0.3057,
+      "step": 22547
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2427585512064883e-05,
+      "loss": 0.2397,
+      "step": 22548
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.242724967675852e-05,
+      "loss": 0.3817,
+      "step": 22549
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2426913841452152e-05,
+      "loss": 0.428,
+      "step": 22550
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2426578006145787e-05,
+      "loss": 0.3047,
+      "step": 22551
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2426242170839422e-05,
+      "loss": 0.45,
+      "step": 22552
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2425906335533056e-05,
+      "loss": 0.353,
+      "step": 22553
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2425570500226689e-05,
+      "loss": 0.3898,
+      "step": 22554
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2425234664920325e-05,
+      "loss": 0.4228,
+      "step": 22555
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2424898829613958e-05,
+      "loss": 0.242,
+      "step": 22556
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2424562994307591e-05,
+      "loss": 0.3065,
+      "step": 22557
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2424227159001228e-05,
+      "loss": 0.2471,
+      "step": 22558
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.242389132369486e-05,
+      "loss": 0.2878,
+      "step": 22559
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2423555488388495e-05,
+      "loss": 0.3512,
+      "step": 22560
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.242321965308213e-05,
+      "loss": 0.5257,
+      "step": 22561
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2422883817775765e-05,
+      "loss": 0.306,
+      "step": 22562
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2422547982469398e-05,
+      "loss": 0.2769,
+      "step": 22563
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2422212147163032e-05,
+      "loss": 0.2406,
+      "step": 22564
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2421876311856667e-05,
+      "loss": 0.1953,
+      "step": 22565
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.24215404765503e-05,
+      "loss": 0.5476,
+      "step": 22566
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2421204641243936e-05,
+      "loss": 0.359,
+      "step": 22567
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2420868805937569e-05,
+      "loss": 0.2441,
+      "step": 22568
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2420532970631202e-05,
+      "loss": 0.241,
+      "step": 22569
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2420197135324838e-05,
+      "loss": 0.2125,
+      "step": 22570
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2419861300018471e-05,
+      "loss": 0.2302,
+      "step": 22571
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2419525464712106e-05,
+      "loss": 0.1876,
+      "step": 22572
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.241918962940574e-05,
+      "loss": 0.3404,
+      "step": 22573
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2418853794099375e-05,
+      "loss": 0.2406,
+      "step": 22574
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2418517958793008e-05,
+      "loss": 0.2989,
+      "step": 22575
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2418182123486645e-05,
+      "loss": 0.1893,
+      "step": 22576
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2417846288180278e-05,
+      "loss": 0.2627,
+      "step": 22577
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.241751045287391e-05,
+      "loss": 0.1709,
+      "step": 22578
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2417174617567547e-05,
+      "loss": 0.2725,
+      "step": 22579
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.241683878226118e-05,
+      "loss": 0.2363,
+      "step": 22580
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2416502946954814e-05,
+      "loss": 0.1641,
+      "step": 22581
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2416167111648449e-05,
+      "loss": 0.3298,
+      "step": 22582
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2415831276342084e-05,
+      "loss": 0.3442,
+      "step": 22583
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2415495441035717e-05,
+      "loss": 0.2051,
+      "step": 22584
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2415159605729353e-05,
+      "loss": 0.5245,
+      "step": 22585
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2414823770422986e-05,
+      "loss": 0.3711,
+      "step": 22586
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2414487935116619e-05,
+      "loss": 0.2995,
+      "step": 22587
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2414152099810255e-05,
+      "loss": 0.3632,
+      "step": 22588
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2413816264503888e-05,
+      "loss": 0.1583,
+      "step": 22589
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2413480429197521e-05,
+      "loss": 0.303,
+      "step": 22590
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2413144593891158e-05,
+      "loss": 0.2236,
+      "step": 22591
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.241280875858479e-05,
+      "loss": 0.2612,
+      "step": 22592
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2412472923278425e-05,
+      "loss": 0.241,
+      "step": 22593
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.241213708797206e-05,
+      "loss": 0.3421,
+      "step": 22594
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2411801252665694e-05,
+      "loss": 0.235,
+      "step": 22595
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2411465417359327e-05,
+      "loss": 0.1936,
+      "step": 22596
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2411129582052964e-05,
+      "loss": 0.2755,
+      "step": 22597
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2410793746746597e-05,
+      "loss": 0.4457,
+      "step": 22598
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.241045791144023e-05,
+      "loss": 0.2115,
+      "step": 22599
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2410122076133866e-05,
+      "loss": 0.3176,
+      "step": 22600
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2409786240827499e-05,
+      "loss": 0.2974,
+      "step": 22601
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2409450405521134e-05,
+      "loss": 0.2719,
+      "step": 22602
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2409114570214768e-05,
+      "loss": 0.2645,
+      "step": 22603
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2408778734908403e-05,
+      "loss": 0.3716,
+      "step": 22604
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2408442899602036e-05,
+      "loss": 0.2505,
+      "step": 22605
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2408107064295672e-05,
+      "loss": 0.3704,
+      "step": 22606
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2407771228989305e-05,
+      "loss": 0.2248,
+      "step": 22607
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2407435393682938e-05,
+      "loss": 0.1504,
+      "step": 22608
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2407099558376574e-05,
+      "loss": 0.1899,
+      "step": 22609
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2406763723070207e-05,
+      "loss": 0.2008,
+      "step": 22610
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.240642788776384e-05,
+      "loss": 0.2686,
+      "step": 22611
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2406092052457477e-05,
+      "loss": 0.3209,
+      "step": 22612
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.240575621715111e-05,
+      "loss": 0.2203,
+      "step": 22613
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2405420381844744e-05,
+      "loss": 0.2924,
+      "step": 22614
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2405084546538379e-05,
+      "loss": 0.3807,
+      "step": 22615
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2404748711232014e-05,
+      "loss": 0.433,
+      "step": 22616
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2404412875925646e-05,
+      "loss": 0.2233,
+      "step": 22617
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2404077040619283e-05,
+      "loss": 0.1596,
+      "step": 22618
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2403741205312916e-05,
+      "loss": 0.504,
+      "step": 22619
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2403405370006549e-05,
+      "loss": 0.3439,
+      "step": 22620
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2403069534700185e-05,
+      "loss": 0.2451,
+      "step": 22621
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2402733699393818e-05,
+      "loss": 0.4523,
+      "step": 22622
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2402397864087453e-05,
+      "loss": 0.2597,
+      "step": 22623
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2402062028781087e-05,
+      "loss": 0.4653,
+      "step": 22624
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2401726193474722e-05,
+      "loss": 0.3247,
+      "step": 22625
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2401390358168355e-05,
+      "loss": 0.2696,
+      "step": 22626
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2401054522861991e-05,
+      "loss": 0.5486,
+      "step": 22627
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2400718687555624e-05,
+      "loss": 0.148,
+      "step": 22628
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2400382852249257e-05,
+      "loss": 0.3161,
+      "step": 22629
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2400047016942894e-05,
+      "loss": 0.4402,
+      "step": 22630
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2399711181636526e-05,
+      "loss": 0.297,
+      "step": 22631
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.239937534633016e-05,
+      "loss": 0.2332,
+      "step": 22632
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2399039511023796e-05,
+      "loss": 0.2711,
+      "step": 22633
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2398703675717429e-05,
+      "loss": 0.265,
+      "step": 22634
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2398367840411063e-05,
+      "loss": 0.0987,
+      "step": 22635
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2398032005104698e-05,
+      "loss": 0.3683,
+      "step": 22636
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2397696169798333e-05,
+      "loss": 0.3214,
+      "step": 22637
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2397360334491966e-05,
+      "loss": 0.3387,
+      "step": 22638
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2397024499185602e-05,
+      "loss": 0.281,
+      "step": 22639
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2396688663879235e-05,
+      "loss": 0.2681,
+      "step": 22640
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2396352828572868e-05,
+      "loss": 0.2754,
+      "step": 22641
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2396016993266504e-05,
+      "loss": 0.3168,
+      "step": 22642
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2395681157960137e-05,
+      "loss": 0.2331,
+      "step": 22643
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2395345322653772e-05,
+      "loss": 0.3046,
+      "step": 22644
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2395009487347406e-05,
+      "loss": 0.3053,
+      "step": 22645
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2394673652041041e-05,
+      "loss": 0.458,
+      "step": 22646
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2394337816734674e-05,
+      "loss": 0.1995,
+      "step": 22647
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.239400198142831e-05,
+      "loss": 0.2606,
+      "step": 22648
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2393666146121943e-05,
+      "loss": 0.3031,
+      "step": 22649
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2393330310815576e-05,
+      "loss": 0.2708,
+      "step": 22650
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2392994475509211e-05,
+      "loss": 0.2386,
+      "step": 22651
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2392658640202846e-05,
+      "loss": 0.3407,
+      "step": 22652
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.239232280489648e-05,
+      "loss": 0.2734,
+      "step": 22653
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2391986969590113e-05,
+      "loss": 0.2656,
+      "step": 22654
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2391651134283748e-05,
+      "loss": 0.1852,
+      "step": 22655
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2391315298977382e-05,
+      "loss": 0.4708,
+      "step": 22656
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2390979463671015e-05,
+      "loss": 0.2454,
+      "step": 22657
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2390643628364652e-05,
+      "loss": 0.2398,
+      "step": 22658
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2390307793058285e-05,
+      "loss": 0.3818,
+      "step": 22659
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2389971957751918e-05,
+      "loss": 0.3471,
+      "step": 22660
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2389636122445554e-05,
+      "loss": 0.4007,
+      "step": 22661
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2389300287139187e-05,
+      "loss": 0.3143,
+      "step": 22662
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2388964451832822e-05,
+      "loss": 0.4099,
+      "step": 22663
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2388628616526456e-05,
+      "loss": 0.2674,
+      "step": 22664
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2388292781220091e-05,
+      "loss": 0.303,
+      "step": 22665
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2387956945913724e-05,
+      "loss": 0.2603,
+      "step": 22666
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.238762111060736e-05,
+      "loss": 0.4161,
+      "step": 22667
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2387285275300993e-05,
+      "loss": 0.298,
+      "step": 22668
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2386949439994626e-05,
+      "loss": 0.3363,
+      "step": 22669
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2386613604688262e-05,
+      "loss": 0.2431,
+      "step": 22670
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2386277769381895e-05,
+      "loss": 0.2764,
+      "step": 22671
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.238594193407553e-05,
+      "loss": 0.2495,
+      "step": 22672
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2385606098769165e-05,
+      "loss": 0.2078,
+      "step": 22673
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.23852702634628e-05,
+      "loss": 0.4181,
+      "step": 22674
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2384934428156432e-05,
+      "loss": 0.3793,
+      "step": 22675
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2384598592850069e-05,
+      "loss": 0.3263,
+      "step": 22676
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2384262757543702e-05,
+      "loss": 0.2993,
+      "step": 22677
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2383926922237335e-05,
+      "loss": 0.3222,
+      "step": 22678
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2383591086930971e-05,
+      "loss": 0.1609,
+      "step": 22679
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2383255251624604e-05,
+      "loss": 0.3389,
+      "step": 22680
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2382919416318237e-05,
+      "loss": 0.3646,
+      "step": 22681
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2382583581011873e-05,
+      "loss": 0.2426,
+      "step": 22682
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2382247745705506e-05,
+      "loss": 0.2571,
+      "step": 22683
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.238191191039914e-05,
+      "loss": 0.3453,
+      "step": 22684
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2381576075092775e-05,
+      "loss": 0.3929,
+      "step": 22685
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.238124023978641e-05,
+      "loss": 0.36,
+      "step": 22686
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2380904404480043e-05,
+      "loss": 0.4493,
+      "step": 22687
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.238056856917368e-05,
+      "loss": 0.2617,
+      "step": 22688
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2380232733867312e-05,
+      "loss": 0.3353,
+      "step": 22689
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2379896898560945e-05,
+      "loss": 0.3208,
+      "step": 22690
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2379561063254582e-05,
+      "loss": 0.2054,
+      "step": 22691
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2379225227948215e-05,
+      "loss": 0.3417,
+      "step": 22692
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.237888939264185e-05,
+      "loss": 0.248,
+      "step": 22693
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2378553557335484e-05,
+      "loss": 0.2606,
+      "step": 22694
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2378217722029118e-05,
+      "loss": 0.3304,
+      "step": 22695
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2377881886722751e-05,
+      "loss": 0.2844,
+      "step": 22696
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2377546051416388e-05,
+      "loss": 0.2782,
+      "step": 22697
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.237721021611002e-05,
+      "loss": 0.472,
+      "step": 22698
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2376874380803654e-05,
+      "loss": 0.4041,
+      "step": 22699
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.237653854549729e-05,
+      "loss": 0.2683,
+      "step": 22700
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2376202710190923e-05,
+      "loss": 0.3608,
+      "step": 22701
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2375866874884556e-05,
+      "loss": 0.3023,
+      "step": 22702
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2375531039578192e-05,
+      "loss": 0.2299,
+      "step": 22703
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2375195204271825e-05,
+      "loss": 0.1514,
+      "step": 22704
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.237485936896546e-05,
+      "loss": 0.3193,
+      "step": 22705
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2374523533659095e-05,
+      "loss": 0.2795,
+      "step": 22706
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.237418769835273e-05,
+      "loss": 0.1944,
+      "step": 22707
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2373851863046362e-05,
+      "loss": 0.4117,
+      "step": 22708
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2373516027739998e-05,
+      "loss": 0.3157,
+      "step": 22709
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2373180192433631e-05,
+      "loss": 0.1571,
+      "step": 22710
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2372844357127264e-05,
+      "loss": 0.4169,
+      "step": 22711
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.23725085218209e-05,
+      "loss": 0.2318,
+      "step": 22712
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2372172686514534e-05,
+      "loss": 0.2248,
+      "step": 22713
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2371836851208168e-05,
+      "loss": 0.3798,
+      "step": 22714
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2371501015901803e-05,
+      "loss": 0.3651,
+      "step": 22715
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2371165180595438e-05,
+      "loss": 0.2508,
+      "step": 22716
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.237082934528907e-05,
+      "loss": 0.3016,
+      "step": 22717
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2370493509982707e-05,
+      "loss": 0.3112,
+      "step": 22718
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.237015767467634e-05,
+      "loss": 0.3544,
+      "step": 22719
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2369821839369973e-05,
+      "loss": 0.299,
+      "step": 22720
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.236948600406361e-05,
+      "loss": 0.2058,
+      "step": 22721
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2369150168757242e-05,
+      "loss": 0.5519,
+      "step": 22722
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2368814333450875e-05,
+      "loss": 0.2314,
+      "step": 22723
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2368478498144511e-05,
+      "loss": 0.4429,
+      "step": 22724
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2368142662838144e-05,
+      "loss": 0.3176,
+      "step": 22725
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2367806827531779e-05,
+      "loss": 0.3479,
+      "step": 22726
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2367470992225414e-05,
+      "loss": 0.2701,
+      "step": 22727
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2367135156919048e-05,
+      "loss": 0.3481,
+      "step": 22728
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.2366799321612681e-05,
+      "loss": 0.3493,
+      "step": 22729
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2366463486306318e-05,
+      "loss": 0.3497,
+      "step": 22730
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.236612765099995e-05,
+      "loss": 0.1975,
+      "step": 22731
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2365791815693584e-05,
+      "loss": 0.3904,
+      "step": 22732
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.236545598038722e-05,
+      "loss": 0.415,
+      "step": 22733
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2365120145080853e-05,
+      "loss": 0.2967,
+      "step": 22734
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2364784309774487e-05,
+      "loss": 0.2566,
+      "step": 22735
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2364448474468122e-05,
+      "loss": 0.2779,
+      "step": 22736
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2364112639161757e-05,
+      "loss": 0.2773,
+      "step": 22737
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.236377680385539e-05,
+      "loss": 0.1695,
+      "step": 22738
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2363440968549026e-05,
+      "loss": 0.1693,
+      "step": 22739
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2363105133242659e-05,
+      "loss": 0.3672,
+      "step": 22740
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2362769297936292e-05,
+      "loss": 0.1803,
+      "step": 22741
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2362433462629928e-05,
+      "loss": 0.3445,
+      "step": 22742
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2362097627323561e-05,
+      "loss": 0.3287,
+      "step": 22743
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2361761792017196e-05,
+      "loss": 0.2423,
+      "step": 22744
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.236142595671083e-05,
+      "loss": 0.4404,
+      "step": 22745
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2361090121404463e-05,
+      "loss": 0.2306,
+      "step": 22746
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2360754286098098e-05,
+      "loss": 0.3899,
+      "step": 22747
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2360418450791733e-05,
+      "loss": 0.4204,
+      "step": 22748
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2360082615485367e-05,
+      "loss": 0.2027,
+      "step": 22749
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2359746780179e-05,
+      "loss": 0.2731,
+      "step": 22750
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2359410944872637e-05,
+      "loss": 0.2804,
+      "step": 22751
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.235907510956627e-05,
+      "loss": 0.386,
+      "step": 22752
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2358739274259903e-05,
+      "loss": 0.2679,
+      "step": 22753
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2358403438953539e-05,
+      "loss": 0.3163,
+      "step": 22754
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2358067603647172e-05,
+      "loss": 0.3532,
+      "step": 22755
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2357731768340807e-05,
+      "loss": 0.2106,
+      "step": 22756
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2357395933034441e-05,
+      "loss": 0.3432,
+      "step": 22757
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2357060097728076e-05,
+      "loss": 0.3776,
+      "step": 22758
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2356724262421709e-05,
+      "loss": 0.2324,
+      "step": 22759
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2356388427115345e-05,
+      "loss": 0.2139,
+      "step": 22760
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2356052591808978e-05,
+      "loss": 0.3772,
+      "step": 22761
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2355716756502611e-05,
+      "loss": 0.3688,
+      "step": 22762
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2355380921196247e-05,
+      "loss": 0.2075,
+      "step": 22763
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.235504508588988e-05,
+      "loss": 0.3526,
+      "step": 22764
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2354709250583515e-05,
+      "loss": 0.2772,
+      "step": 22765
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.235437341527715e-05,
+      "loss": 0.3475,
+      "step": 22766
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2354037579970783e-05,
+      "loss": 0.3185,
+      "step": 22767
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2353701744664417e-05,
+      "loss": 0.2894,
+      "step": 22768
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2353365909358052e-05,
+      "loss": 0.2163,
+      "step": 22769
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2353030074051687e-05,
+      "loss": 0.4282,
+      "step": 22770
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.235269423874532e-05,
+      "loss": 0.2123,
+      "step": 22771
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2352358403438956e-05,
+      "loss": 0.2918,
+      "step": 22772
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2352022568132589e-05,
+      "loss": 0.333,
+      "step": 22773
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2351686732826222e-05,
+      "loss": 0.2102,
+      "step": 22774
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2351350897519858e-05,
+      "loss": 0.1968,
+      "step": 22775
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2351015062213491e-05,
+      "loss": 0.2316,
+      "step": 22776
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2350679226907126e-05,
+      "loss": 0.3407,
+      "step": 22777
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.235034339160076e-05,
+      "loss": 0.2693,
+      "step": 22778
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2350007556294395e-05,
+      "loss": 0.1726,
+      "step": 22779
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2349671720988028e-05,
+      "loss": 0.444,
+      "step": 22780
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2349335885681664e-05,
+      "loss": 0.2915,
+      "step": 22781
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2349000050375297e-05,
+      "loss": 0.5217,
+      "step": 22782
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.234866421506893e-05,
+      "loss": 0.4737,
+      "step": 22783
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2348328379762567e-05,
+      "loss": 0.3605,
+      "step": 22784
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.23479925444562e-05,
+      "loss": 0.4211,
+      "step": 22785
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2347656709149834e-05,
+      "loss": 0.3451,
+      "step": 22786
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2347320873843469e-05,
+      "loss": 0.361,
+      "step": 22787
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2346985038537103e-05,
+      "loss": 0.3946,
+      "step": 22788
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2346649203230736e-05,
+      "loss": 0.4011,
+      "step": 22789
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2346313367924371e-05,
+      "loss": 0.2895,
+      "step": 22790
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2345977532618006e-05,
+      "loss": 0.2509,
+      "step": 22791
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2345641697311639e-05,
+      "loss": 0.3385,
+      "step": 22792
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2345305862005275e-05,
+      "loss": 0.2407,
+      "step": 22793
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2344970026698908e-05,
+      "loss": 0.328,
+      "step": 22794
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2344634191392541e-05,
+      "loss": 0.3242,
+      "step": 22795
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2344298356086177e-05,
+      "loss": 0.3464,
+      "step": 22796
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.234396252077981e-05,
+      "loss": 0.3185,
+      "step": 22797
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2343626685473445e-05,
+      "loss": 0.2676,
+      "step": 22798
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.234329085016708e-05,
+      "loss": 0.2176,
+      "step": 22799
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2342955014860714e-05,
+      "loss": 0.3592,
+      "step": 22800
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2342619179554347e-05,
+      "loss": 0.3626,
+      "step": 22801
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2342283344247983e-05,
+      "loss": 0.2665,
+      "step": 22802
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2341947508941616e-05,
+      "loss": 0.2757,
+      "step": 22803
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.234161167363525e-05,
+      "loss": 0.3279,
+      "step": 22804
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2341275838328886e-05,
+      "loss": 0.299,
+      "step": 22805
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2340940003022519e-05,
+      "loss": 0.3018,
+      "step": 22806
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2340604167716153e-05,
+      "loss": 0.2712,
+      "step": 22807
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2340268332409788e-05,
+      "loss": 0.3209,
+      "step": 22808
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2339932497103423e-05,
+      "loss": 0.3304,
+      "step": 22809
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2339596661797056e-05,
+      "loss": 0.385,
+      "step": 22810
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2339260826490692e-05,
+      "loss": 0.2039,
+      "step": 22811
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2338924991184325e-05,
+      "loss": 0.3264,
+      "step": 22812
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2338589155877958e-05,
+      "loss": 0.49,
+      "step": 22813
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2338253320571594e-05,
+      "loss": 0.3666,
+      "step": 22814
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2337917485265227e-05,
+      "loss": 0.4132,
+      "step": 22815
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.233758164995886e-05,
+      "loss": 0.3004,
+      "step": 22816
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2337245814652496e-05,
+      "loss": 0.3654,
+      "step": 22817
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.233690997934613e-05,
+      "loss": 0.2913,
+      "step": 22818
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2336574144039764e-05,
+      "loss": 0.1867,
+      "step": 22819
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2336238308733399e-05,
+      "loss": 0.3758,
+      "step": 22820
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2335902473427033e-05,
+      "loss": 0.2914,
+      "step": 22821
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2335566638120666e-05,
+      "loss": 0.3871,
+      "step": 22822
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2335230802814303e-05,
+      "loss": 0.2486,
+      "step": 22823
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2334894967507935e-05,
+      "loss": 0.1974,
+      "step": 22824
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2334559132201568e-05,
+      "loss": 0.4129,
+      "step": 22825
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2334223296895205e-05,
+      "loss": 0.2584,
+      "step": 22826
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2333887461588838e-05,
+      "loss": 0.2921,
+      "step": 22827
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2333551626282472e-05,
+      "loss": 0.369,
+      "step": 22828
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2333215790976105e-05,
+      "loss": 0.426,
+      "step": 22829
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2332879955669742e-05,
+      "loss": 0.3131,
+      "step": 22830
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2332544120363375e-05,
+      "loss": 0.3493,
+      "step": 22831
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2332208285057008e-05,
+      "loss": 0.199,
+      "step": 22832
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2331872449750644e-05,
+      "loss": 0.2547,
+      "step": 22833
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2331536614444277e-05,
+      "loss": 0.2322,
+      "step": 22834
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2331200779137912e-05,
+      "loss": 0.3746,
+      "step": 22835
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2330864943831546e-05,
+      "loss": 0.2298,
+      "step": 22836
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2330529108525179e-05,
+      "loss": 0.3182,
+      "step": 22837
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2330193273218814e-05,
+      "loss": 0.2747,
+      "step": 22838
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2329857437912448e-05,
+      "loss": 0.1665,
+      "step": 22839
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2329521602606083e-05,
+      "loss": 0.2869,
+      "step": 22840
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2329185767299716e-05,
+      "loss": 0.1601,
+      "step": 22841
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2328849931993352e-05,
+      "loss": 0.2929,
+      "step": 22842
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2328514096686985e-05,
+      "loss": 0.3187,
+      "step": 22843
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2328178261380618e-05,
+      "loss": 0.3603,
+      "step": 22844
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2327842426074255e-05,
+      "loss": 0.4305,
+      "step": 22845
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2327506590767888e-05,
+      "loss": 0.3561,
+      "step": 22846
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2327170755461522e-05,
+      "loss": 0.3051,
+      "step": 22847
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2326834920155157e-05,
+      "loss": 0.253,
+      "step": 22848
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2326499084848792e-05,
+      "loss": 0.2437,
+      "step": 22849
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2326163249542424e-05,
+      "loss": 0.2392,
+      "step": 22850
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.232582741423606e-05,
+      "loss": 0.3107,
+      "step": 22851
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2325491578929694e-05,
+      "loss": 0.3926,
+      "step": 22852
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2325155743623327e-05,
+      "loss": 0.5353,
+      "step": 22853
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2324819908316963e-05,
+      "loss": 0.1961,
+      "step": 22854
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2324484073010596e-05,
+      "loss": 0.375,
+      "step": 22855
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.232414823770423e-05,
+      "loss": 0.398,
+      "step": 22856
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2323812402397865e-05,
+      "loss": 0.3126,
+      "step": 22857
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2323476567091498e-05,
+      "loss": 0.251,
+      "step": 22858
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2323140731785133e-05,
+      "loss": 0.3124,
+      "step": 22859
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2322804896478768e-05,
+      "loss": 0.352,
+      "step": 22860
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2322469061172402e-05,
+      "loss": 0.3095,
+      "step": 22861
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2322133225866035e-05,
+      "loss": 0.3501,
+      "step": 22862
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2321797390559671e-05,
+      "loss": 0.3555,
+      "step": 22863
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2321461555253304e-05,
+      "loss": 0.2658,
+      "step": 22864
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2321125719946937e-05,
+      "loss": 0.3845,
+      "step": 22865
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2320789884640574e-05,
+      "loss": 0.2662,
+      "step": 22866
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2320454049334207e-05,
+      "loss": 0.2909,
+      "step": 22867
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2320118214027841e-05,
+      "loss": 0.3082,
+      "step": 22868
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2319782378721476e-05,
+      "loss": 0.2934,
+      "step": 22869
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.231944654341511e-05,
+      "loss": 0.3762,
+      "step": 22870
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2319110708108744e-05,
+      "loss": 0.2327,
+      "step": 22871
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.231877487280238e-05,
+      "loss": 0.1692,
+      "step": 22872
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2318439037496013e-05,
+      "loss": 0.2989,
+      "step": 22873
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2318103202189646e-05,
+      "loss": 0.2985,
+      "step": 22874
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2317767366883282e-05,
+      "loss": 0.318,
+      "step": 22875
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2317431531576915e-05,
+      "loss": 0.2324,
+      "step": 22876
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.231709569627055e-05,
+      "loss": 0.3448,
+      "step": 22877
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2316759860964184e-05,
+      "loss": 0.1875,
+      "step": 22878
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2316424025657819e-05,
+      "loss": 0.3125,
+      "step": 22879
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2316088190351452e-05,
+      "loss": 0.4844,
+      "step": 22880
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2315752355045087e-05,
+      "loss": 0.2808,
+      "step": 22881
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2315416519738721e-05,
+      "loss": 0.2496,
+      "step": 22882
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2315080684432354e-05,
+      "loss": 0.2604,
+      "step": 22883
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.231474484912599e-05,
+      "loss": 0.3077,
+      "step": 22884
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2314409013819624e-05,
+      "loss": 0.2327,
+      "step": 22885
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2314073178513257e-05,
+      "loss": 0.2243,
+      "step": 22886
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2313737343206893e-05,
+      "loss": 0.2411,
+      "step": 22887
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2313401507900526e-05,
+      "loss": 0.3123,
+      "step": 22888
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.231306567259416e-05,
+      "loss": 0.3951,
+      "step": 22889
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2312729837287795e-05,
+      "loss": 0.1891,
+      "step": 22890
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.231239400198143e-05,
+      "loss": 0.3312,
+      "step": 22891
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2312058166675063e-05,
+      "loss": 0.3606,
+      "step": 22892
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2311722331368699e-05,
+      "loss": 0.4232,
+      "step": 22893
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2311386496062332e-05,
+      "loss": 0.2234,
+      "step": 22894
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2311050660755965e-05,
+      "loss": 0.1769,
+      "step": 22895
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2310714825449601e-05,
+      "loss": 0.1799,
+      "step": 22896
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2310378990143234e-05,
+      "loss": 0.3354,
+      "step": 22897
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2310043154836869e-05,
+      "loss": 0.4086,
+      "step": 22898
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2309707319530504e-05,
+      "loss": 0.2257,
+      "step": 22899
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2309371484224138e-05,
+      "loss": 0.1946,
+      "step": 22900
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2309035648917771e-05,
+      "loss": 0.3524,
+      "step": 22901
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2308699813611406e-05,
+      "loss": 0.1852,
+      "step": 22902
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.230836397830504e-05,
+      "loss": 0.2688,
+      "step": 22903
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2308028142998673e-05,
+      "loss": 0.244,
+      "step": 22904
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.230769230769231e-05,
+      "loss": 0.3124,
+      "step": 22905
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2307356472385943e-05,
+      "loss": 0.3019,
+      "step": 22906
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2307020637079576e-05,
+      "loss": 0.3376,
+      "step": 22907
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2306684801773212e-05,
+      "loss": 0.3678,
+      "step": 22908
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2306348966466845e-05,
+      "loss": 0.2038,
+      "step": 22909
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.230601313116048e-05,
+      "loss": 0.2522,
+      "step": 22910
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2305677295854114e-05,
+      "loss": 0.1537,
+      "step": 22911
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2305341460547749e-05,
+      "loss": 0.2108,
+      "step": 22912
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2305005625241382e-05,
+      "loss": 0.2566,
+      "step": 22913
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2304669789935018e-05,
+      "loss": 0.2142,
+      "step": 22914
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2304333954628651e-05,
+      "loss": 0.2849,
+      "step": 22915
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2303998119322284e-05,
+      "loss": 0.3339,
+      "step": 22916
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.230366228401592e-05,
+      "loss": 0.3069,
+      "step": 22917
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2303326448709553e-05,
+      "loss": 0.3056,
+      "step": 22918
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2302990613403188e-05,
+      "loss": 0.2302,
+      "step": 22919
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2302654778096823e-05,
+      "loss": 0.2576,
+      "step": 22920
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2302318942790457e-05,
+      "loss": 0.2225,
+      "step": 22921
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.230198310748409e-05,
+      "loss": 0.3076,
+      "step": 22922
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2301647272177727e-05,
+      "loss": 0.2373,
+      "step": 22923
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.230131143687136e-05,
+      "loss": 0.5644,
+      "step": 22924
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2300975601564993e-05,
+      "loss": 0.1496,
+      "step": 22925
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2300639766258629e-05,
+      "loss": 0.2731,
+      "step": 22926
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 1.2300303930952262e-05,
+      "loss": 0.1849,
+      "step": 22927
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2299968095645895e-05,
+      "loss": 0.192,
+      "step": 22928
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2299632260339531e-05,
+      "loss": 0.2334,
+      "step": 22929
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2299296425033164e-05,
+      "loss": 0.2897,
+      "step": 22930
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2298960589726799e-05,
+      "loss": 0.1936,
+      "step": 22931
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2298624754420433e-05,
+      "loss": 0.2648,
+      "step": 22932
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2298288919114068e-05,
+      "loss": 0.3186,
+      "step": 22933
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2297953083807701e-05,
+      "loss": 0.3133,
+      "step": 22934
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2297617248501337e-05,
+      "loss": 0.3054,
+      "step": 22935
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.229728141319497e-05,
+      "loss": 0.2726,
+      "step": 22936
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2296945577888603e-05,
+      "loss": 0.3044,
+      "step": 22937
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.229660974258224e-05,
+      "loss": 0.3325,
+      "step": 22938
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2296273907275873e-05,
+      "loss": 0.3989,
+      "step": 22939
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2295938071969507e-05,
+      "loss": 0.3459,
+      "step": 22940
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2295602236663142e-05,
+      "loss": 0.3724,
+      "step": 22941
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2295266401356776e-05,
+      "loss": 0.2718,
+      "step": 22942
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.229493056605041e-05,
+      "loss": 0.3708,
+      "step": 22943
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2294594730744046e-05,
+      "loss": 0.2689,
+      "step": 22944
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2294258895437679e-05,
+      "loss": 0.2703,
+      "step": 22945
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2293923060131312e-05,
+      "loss": 0.2269,
+      "step": 22946
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2293587224824948e-05,
+      "loss": 0.2225,
+      "step": 22947
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2293251389518581e-05,
+      "loss": 0.2863,
+      "step": 22948
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2292915554212214e-05,
+      "loss": 0.2136,
+      "step": 22949
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.229257971890585e-05,
+      "loss": 0.3812,
+      "step": 22950
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2292243883599483e-05,
+      "loss": 0.219,
+      "step": 22951
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2291908048293118e-05,
+      "loss": 0.3119,
+      "step": 22952
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2291572212986752e-05,
+      "loss": 0.3795,
+      "step": 22953
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2291236377680387e-05,
+      "loss": 0.234,
+      "step": 22954
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.229090054237402e-05,
+      "loss": 0.4298,
+      "step": 22955
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2290564707067656e-05,
+      "loss": 0.2192,
+      "step": 22956
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.229022887176129e-05,
+      "loss": 0.3053,
+      "step": 22957
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2289893036454922e-05,
+      "loss": 0.486,
+      "step": 22958
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2289557201148559e-05,
+      "loss": 0.1967,
+      "step": 22959
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2289221365842192e-05,
+      "loss": 0.2761,
+      "step": 22960
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2288885530535826e-05,
+      "loss": 0.3705,
+      "step": 22961
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2288549695229461e-05,
+      "loss": 0.3564,
+      "step": 22962
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2288213859923096e-05,
+      "loss": 0.3078,
+      "step": 22963
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2287878024616729e-05,
+      "loss": 0.2141,
+      "step": 22964
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2287542189310365e-05,
+      "loss": 0.2046,
+      "step": 22965
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2287206354003998e-05,
+      "loss": 0.3994,
+      "step": 22966
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.228687051869763e-05,
+      "loss": 0.2323,
+      "step": 22967
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2286534683391267e-05,
+      "loss": 0.1871,
+      "step": 22968
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.22861988480849e-05,
+      "loss": 0.496,
+      "step": 22969
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2285863012778535e-05,
+      "loss": 0.2145,
+      "step": 22970
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.228552717747217e-05,
+      "loss": 0.3286,
+      "step": 22971
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2285191342165802e-05,
+      "loss": 0.419,
+      "step": 22972
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2284855506859437e-05,
+      "loss": 0.2774,
+      "step": 22973
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2284519671553072e-05,
+      "loss": 0.1614,
+      "step": 22974
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2284183836246706e-05,
+      "loss": 0.4469,
+      "step": 22975
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.228384800094034e-05,
+      "loss": 0.3212,
+      "step": 22976
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2283512165633976e-05,
+      "loss": 0.1412,
+      "step": 22977
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2283176330327609e-05,
+      "loss": 0.2246,
+      "step": 22978
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2282840495021241e-05,
+      "loss": 0.5249,
+      "step": 22979
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2282504659714878e-05,
+      "loss": 0.3223,
+      "step": 22980
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.228216882440851e-05,
+      "loss": 0.2885,
+      "step": 22981
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2281832989102145e-05,
+      "loss": 0.2384,
+      "step": 22982
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.228149715379578e-05,
+      "loss": 0.3816,
+      "step": 22983
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2281161318489415e-05,
+      "loss": 0.3594,
+      "step": 22984
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2280825483183048e-05,
+      "loss": 0.2642,
+      "step": 22985
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2280489647876684e-05,
+      "loss": 0.2492,
+      "step": 22986
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2280153812570317e-05,
+      "loss": 0.3931,
+      "step": 22987
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.227981797726395e-05,
+      "loss": 0.1436,
+      "step": 22988
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2279482141957586e-05,
+      "loss": 0.4367,
+      "step": 22989
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.227914630665122e-05,
+      "loss": 0.4695,
+      "step": 22990
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2278810471344854e-05,
+      "loss": 0.2442,
+      "step": 22991
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2278474636038488e-05,
+      "loss": 0.2402,
+      "step": 22992
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2278138800732121e-05,
+      "loss": 0.3561,
+      "step": 22993
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2277802965425756e-05,
+      "loss": 0.1654,
+      "step": 22994
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.227746713011939e-05,
+      "loss": 0.4351,
+      "step": 22995
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2277131294813025e-05,
+      "loss": 0.6085,
+      "step": 22996
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2276795459506658e-05,
+      "loss": 0.3114,
+      "step": 22997
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2276459624200295e-05,
+      "loss": 0.2642,
+      "step": 22998
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2276123788893928e-05,
+      "loss": 0.2447,
+      "step": 22999
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.227578795358756e-05,
+      "loss": 0.3187,
+      "step": 23000
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2275452118281197e-05,
+      "loss": 0.2703,
+      "step": 23001
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.227511628297483e-05,
+      "loss": 0.2661,
+      "step": 23002
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2274780447668465e-05,
+      "loss": 0.2832,
+      "step": 23003
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.22744446123621e-05,
+      "loss": 0.3655,
+      "step": 23004
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2274108777055734e-05,
+      "loss": 0.2093,
+      "step": 23005
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2273772941749367e-05,
+      "loss": 0.5151,
+      "step": 23006
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2273437106443003e-05,
+      "loss": 0.2309,
+      "step": 23007
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2273101271136636e-05,
+      "loss": 0.1877,
+      "step": 23008
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2272765435830269e-05,
+      "loss": 0.2827,
+      "step": 23009
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2272429600523904e-05,
+      "loss": 0.3123,
+      "step": 23010
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2272093765217538e-05,
+      "loss": 0.4699,
+      "step": 23011
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2271757929911173e-05,
+      "loss": 0.2153,
+      "step": 23012
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2271422094604806e-05,
+      "loss": 0.4401,
+      "step": 23013
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2271086259298442e-05,
+      "loss": 0.2938,
+      "step": 23014
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2270750423992075e-05,
+      "loss": 0.267,
+      "step": 23015
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2270414588685708e-05,
+      "loss": 0.3296,
+      "step": 23016
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2270078753379345e-05,
+      "loss": 0.3723,
+      "step": 23017
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2269742918072977e-05,
+      "loss": 0.1447,
+      "step": 23018
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.226940708276661e-05,
+      "loss": 0.2449,
+      "step": 23019
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2269071247460247e-05,
+      "loss": 0.3086,
+      "step": 23020
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.226873541215388e-05,
+      "loss": 0.2653,
+      "step": 23021
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2268399576847514e-05,
+      "loss": 0.3371,
+      "step": 23022
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2268063741541149e-05,
+      "loss": 0.1879,
+      "step": 23023
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2267727906234784e-05,
+      "loss": 0.2127,
+      "step": 23024
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2267392070928417e-05,
+      "loss": 0.3103,
+      "step": 23025
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2267056235622053e-05,
+      "loss": 0.2026,
+      "step": 23026
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2266720400315686e-05,
+      "loss": 0.3172,
+      "step": 23027
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2266384565009319e-05,
+      "loss": 0.3096,
+      "step": 23028
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2266048729702955e-05,
+      "loss": 0.3388,
+      "step": 23029
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2265712894396588e-05,
+      "loss": 0.3086,
+      "step": 23030
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2265377059090223e-05,
+      "loss": 0.3497,
+      "step": 23031
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2265041223783857e-05,
+      "loss": 0.3538,
+      "step": 23032
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2264705388477492e-05,
+      "loss": 0.3963,
+      "step": 23033
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2264369553171125e-05,
+      "loss": 0.3971,
+      "step": 23034
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2264033717864761e-05,
+      "loss": 0.5155,
+      "step": 23035
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2263697882558394e-05,
+      "loss": 0.2665,
+      "step": 23036
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2263362047252027e-05,
+      "loss": 0.277,
+      "step": 23037
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2263026211945664e-05,
+      "loss": 0.3729,
+      "step": 23038
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2262690376639297e-05,
+      "loss": 0.319,
+      "step": 23039
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.226235454133293e-05,
+      "loss": 0.3268,
+      "step": 23040
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2262018706026566e-05,
+      "loss": 0.2562,
+      "step": 23041
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2261682870720199e-05,
+      "loss": 0.3531,
+      "step": 23042
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2261347035413833e-05,
+      "loss": 0.3108,
+      "step": 23043
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2261011200107468e-05,
+      "loss": 0.3492,
+      "step": 23044
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2260675364801103e-05,
+      "loss": 0.1812,
+      "step": 23045
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2260339529494736e-05,
+      "loss": 0.1827,
+      "step": 23046
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2260003694188372e-05,
+      "loss": 0.1904,
+      "step": 23047
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2259667858882005e-05,
+      "loss": 0.2353,
+      "step": 23048
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2259332023575638e-05,
+      "loss": 0.3096,
+      "step": 23049
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2258996188269274e-05,
+      "loss": 0.3432,
+      "step": 23050
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2258660352962907e-05,
+      "loss": 0.3496,
+      "step": 23051
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2258324517656542e-05,
+      "loss": 0.3708,
+      "step": 23052
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2257988682350177e-05,
+      "loss": 0.3716,
+      "step": 23053
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2257652847043811e-05,
+      "loss": 0.3548,
+      "step": 23054
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2257317011737444e-05,
+      "loss": 0.2781,
+      "step": 23055
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.225698117643108e-05,
+      "loss": 0.4455,
+      "step": 23056
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2256645341124713e-05,
+      "loss": 0.2524,
+      "step": 23057
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2256309505818346e-05,
+      "loss": 0.273,
+      "step": 23058
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2255973670511983e-05,
+      "loss": 0.4611,
+      "step": 23059
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2255637835205616e-05,
+      "loss": 0.2508,
+      "step": 23060
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.225530199989925e-05,
+      "loss": 0.2026,
+      "step": 23061
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2254966164592885e-05,
+      "loss": 0.3989,
+      "step": 23062
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2254630329286518e-05,
+      "loss": 0.2454,
+      "step": 23063
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2254294493980153e-05,
+      "loss": 0.2022,
+      "step": 23064
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2253958658673787e-05,
+      "loss": 0.1655,
+      "step": 23065
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2253622823367422e-05,
+      "loss": 0.3577,
+      "step": 23066
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2253286988061055e-05,
+      "loss": 0.3339,
+      "step": 23067
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2252951152754691e-05,
+      "loss": 0.1752,
+      "step": 23068
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2252615317448324e-05,
+      "loss": 0.3369,
+      "step": 23069
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2252279482141957e-05,
+      "loss": 0.3172,
+      "step": 23070
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2251943646835593e-05,
+      "loss": 0.2318,
+      "step": 23071
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2251607811529226e-05,
+      "loss": 0.1998,
+      "step": 23072
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2251271976222861e-05,
+      "loss": 0.1859,
+      "step": 23073
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2250936140916496e-05,
+      "loss": 0.273,
+      "step": 23074
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.225060030561013e-05,
+      "loss": 0.2854,
+      "step": 23075
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2250264470303763e-05,
+      "loss": 0.4772,
+      "step": 23076
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.22499286349974e-05,
+      "loss": 0.41,
+      "step": 23077
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2249592799691033e-05,
+      "loss": 0.2668,
+      "step": 23078
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2249256964384666e-05,
+      "loss": 0.3236,
+      "step": 23079
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2248921129078302e-05,
+      "loss": 0.404,
+      "step": 23080
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2248585293771935e-05,
+      "loss": 0.2371,
+      "step": 23081
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.224824945846557e-05,
+      "loss": 0.3142,
+      "step": 23082
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2247913623159204e-05,
+      "loss": 0.3137,
+      "step": 23083
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2247577787852837e-05,
+      "loss": 0.2456,
+      "step": 23084
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2247241952546472e-05,
+      "loss": 0.3945,
+      "step": 23085
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2246906117240106e-05,
+      "loss": 0.1568,
+      "step": 23086
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2246570281933741e-05,
+      "loss": 0.1953,
+      "step": 23087
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2246234446627374e-05,
+      "loss": 0.2711,
+      "step": 23088
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.224589861132101e-05,
+      "loss": 0.3369,
+      "step": 23089
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2245562776014643e-05,
+      "loss": 0.1975,
+      "step": 23090
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2245226940708276e-05,
+      "loss": 0.369,
+      "step": 23091
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2244891105401913e-05,
+      "loss": 0.1543,
+      "step": 23092
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2244555270095546e-05,
+      "loss": 0.3033,
+      "step": 23093
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.224421943478918e-05,
+      "loss": 0.2615,
+      "step": 23094
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2243883599482815e-05,
+      "loss": 0.2344,
+      "step": 23095
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.224354776417645e-05,
+      "loss": 0.4005,
+      "step": 23096
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2243211928870082e-05,
+      "loss": 0.3473,
+      "step": 23097
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2242876093563719e-05,
+      "loss": 0.4361,
+      "step": 23098
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2242540258257352e-05,
+      "loss": 0.3575,
+      "step": 23099
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2242204422950985e-05,
+      "loss": 0.1211,
+      "step": 23100
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2241868587644621e-05,
+      "loss": 0.287,
+      "step": 23101
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2241532752338254e-05,
+      "loss": 0.2091,
+      "step": 23102
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2241196917031889e-05,
+      "loss": 0.419,
+      "step": 23103
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2240861081725523e-05,
+      "loss": 0.3534,
+      "step": 23104
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2240525246419158e-05,
+      "loss": 0.2175,
+      "step": 23105
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2240189411112791e-05,
+      "loss": 0.2606,
+      "step": 23106
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2239853575806426e-05,
+      "loss": 0.3823,
+      "step": 23107
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.223951774050006e-05,
+      "loss": 0.3198,
+      "step": 23108
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2239181905193693e-05,
+      "loss": 0.1992,
+      "step": 23109
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.223884606988733e-05,
+      "loss": 0.3909,
+      "step": 23110
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2238510234580962e-05,
+      "loss": 0.2149,
+      "step": 23111
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2238174399274595e-05,
+      "loss": 0.1971,
+      "step": 23112
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2237838563968232e-05,
+      "loss": 0.3086,
+      "step": 23113
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2237502728661865e-05,
+      "loss": 0.1521,
+      "step": 23114
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.22371668933555e-05,
+      "loss": 0.2429,
+      "step": 23115
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2236831058049134e-05,
+      "loss": 0.2265,
+      "step": 23116
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2236495222742769e-05,
+      "loss": 0.3091,
+      "step": 23117
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2236159387436402e-05,
+      "loss": 0.3945,
+      "step": 23118
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2235823552130038e-05,
+      "loss": 0.3213,
+      "step": 23119
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.223548771682367e-05,
+      "loss": 0.3997,
+      "step": 23120
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2235151881517304e-05,
+      "loss": 0.192,
+      "step": 23121
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.223481604621094e-05,
+      "loss": 0.3225,
+      "step": 23122
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2234480210904573e-05,
+      "loss": 0.5758,
+      "step": 23123
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2234144375598208e-05,
+      "loss": 0.4779,
+      "step": 23124
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2233808540291842e-05,
+      "loss": 0.2579,
+      "step": 23125
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 1.2233472704985477e-05,
+      "loss": 0.2935,
+      "step": 23126
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.223313686967911e-05,
+      "loss": 0.3259,
+      "step": 23127
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2232801034372745e-05,
+      "loss": 0.3004,
+      "step": 23128
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.223246519906638e-05,
+      "loss": 0.3209,
+      "step": 23129
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2232129363760012e-05,
+      "loss": 0.2937,
+      "step": 23130
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2231793528453649e-05,
+      "loss": 0.2208,
+      "step": 23131
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2231457693147282e-05,
+      "loss": 0.3413,
+      "step": 23132
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2231121857840914e-05,
+      "loss": 0.2513,
+      "step": 23133
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.223078602253455e-05,
+      "loss": 0.1518,
+      "step": 23134
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2230450187228184e-05,
+      "loss": 0.1703,
+      "step": 23135
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2230114351921818e-05,
+      "loss": 0.6649,
+      "step": 23136
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2229778516615453e-05,
+      "loss": 0.262,
+      "step": 23137
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2229442681309088e-05,
+      "loss": 0.2788,
+      "step": 23138
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.222910684600272e-05,
+      "loss": 0.2273,
+      "step": 23139
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2228771010696357e-05,
+      "loss": 0.3829,
+      "step": 23140
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.222843517538999e-05,
+      "loss": 0.1734,
+      "step": 23141
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2228099340083623e-05,
+      "loss": 0.301,
+      "step": 23142
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.222776350477726e-05,
+      "loss": 0.252,
+      "step": 23143
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2227427669470892e-05,
+      "loss": 0.4636,
+      "step": 23144
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2227091834164527e-05,
+      "loss": 0.1924,
+      "step": 23145
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2226755998858162e-05,
+      "loss": 0.2211,
+      "step": 23146
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2226420163551796e-05,
+      "loss": 0.3708,
+      "step": 23147
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2226084328245429e-05,
+      "loss": 0.16,
+      "step": 23148
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2225748492939065e-05,
+      "loss": 0.1723,
+      "step": 23149
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2225412657632698e-05,
+      "loss": 0.3631,
+      "step": 23150
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2225076822326331e-05,
+      "loss": 0.4081,
+      "step": 23151
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2224740987019968e-05,
+      "loss": 0.2558,
+      "step": 23152
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.22244051517136e-05,
+      "loss": 0.3397,
+      "step": 23153
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2224069316407234e-05,
+      "loss": 0.2817,
+      "step": 23154
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.222373348110087e-05,
+      "loss": 0.373,
+      "step": 23155
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2223397645794503e-05,
+      "loss": 0.2774,
+      "step": 23156
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2223061810488138e-05,
+      "loss": 0.211,
+      "step": 23157
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2222725975181772e-05,
+      "loss": 0.3351,
+      "step": 23158
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2222390139875407e-05,
+      "loss": 0.3744,
+      "step": 23159
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.222205430456904e-05,
+      "loss": 0.3221,
+      "step": 23160
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2221718469262676e-05,
+      "loss": 0.1846,
+      "step": 23161
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2221382633956309e-05,
+      "loss": 0.2301,
+      "step": 23162
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2221046798649942e-05,
+      "loss": 0.4726,
+      "step": 23163
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2220710963343578e-05,
+      "loss": 0.4333,
+      "step": 23164
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2220375128037211e-05,
+      "loss": 0.3241,
+      "step": 23165
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2220039292730846e-05,
+      "loss": 0.3127,
+      "step": 23166
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.221970345742448e-05,
+      "loss": 0.3898,
+      "step": 23167
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2219367622118115e-05,
+      "loss": 0.2004,
+      "step": 23168
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2219031786811748e-05,
+      "loss": 0.2995,
+      "step": 23169
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2218695951505385e-05,
+      "loss": 0.25,
+      "step": 23170
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2218360116199018e-05,
+      "loss": 0.3623,
+      "step": 23171
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.221802428089265e-05,
+      "loss": 0.1793,
+      "step": 23172
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2217688445586287e-05,
+      "loss": 0.3215,
+      "step": 23173
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.221735261027992e-05,
+      "loss": 0.3127,
+      "step": 23174
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2217016774973553e-05,
+      "loss": 0.2005,
+      "step": 23175
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2216680939667189e-05,
+      "loss": 0.3403,
+      "step": 23176
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2216345104360822e-05,
+      "loss": 0.1985,
+      "step": 23177
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2216009269054457e-05,
+      "loss": 0.1904,
+      "step": 23178
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2215673433748091e-05,
+      "loss": 0.3266,
+      "step": 23179
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2215337598441726e-05,
+      "loss": 0.4256,
+      "step": 23180
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2215001763135359e-05,
+      "loss": 0.196,
+      "step": 23181
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2214665927828995e-05,
+      "loss": 0.2672,
+      "step": 23182
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2214330092522628e-05,
+      "loss": 0.1148,
+      "step": 23183
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2213994257216261e-05,
+      "loss": 0.3111,
+      "step": 23184
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2213658421909898e-05,
+      "loss": 0.3002,
+      "step": 23185
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.221332258660353e-05,
+      "loss": 0.2741,
+      "step": 23186
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2212986751297165e-05,
+      "loss": 0.3724,
+      "step": 23187
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.22126509159908e-05,
+      "loss": 0.2446,
+      "step": 23188
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2212315080684434e-05,
+      "loss": 0.3024,
+      "step": 23189
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2211979245378067e-05,
+      "loss": 0.3314,
+      "step": 23190
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.22116434100717e-05,
+      "loss": 0.3778,
+      "step": 23191
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2211307574765337e-05,
+      "loss": 0.3269,
+      "step": 23192
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.221097173945897e-05,
+      "loss": 0.3032,
+      "step": 23193
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2210635904152604e-05,
+      "loss": 0.4919,
+      "step": 23194
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2210300068846239e-05,
+      "loss": 0.3531,
+      "step": 23195
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2209964233539874e-05,
+      "loss": 0.2804,
+      "step": 23196
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2209628398233506e-05,
+      "loss": 0.339,
+      "step": 23197
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2209292562927141e-05,
+      "loss": 0.308,
+      "step": 23198
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2208956727620776e-05,
+      "loss": 0.2602,
+      "step": 23199
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2208620892314409e-05,
+      "loss": 0.2874,
+      "step": 23200
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2208285057008045e-05,
+      "loss": 0.2962,
+      "step": 23201
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2207949221701678e-05,
+      "loss": 0.2005,
+      "step": 23202
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2207613386395311e-05,
+      "loss": 0.1584,
+      "step": 23203
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2207277551088947e-05,
+      "loss": 0.2271,
+      "step": 23204
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.220694171578258e-05,
+      "loss": 0.3113,
+      "step": 23205
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2206605880476215e-05,
+      "loss": 0.37,
+      "step": 23206
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.220627004516985e-05,
+      "loss": 0.1755,
+      "step": 23207
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2205934209863484e-05,
+      "loss": 0.4396,
+      "step": 23208
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2205598374557117e-05,
+      "loss": 0.3786,
+      "step": 23209
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2205262539250754e-05,
+      "loss": 0.3639,
+      "step": 23210
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2204926703944386e-05,
+      "loss": 0.2803,
+      "step": 23211
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.220459086863802e-05,
+      "loss": 0.3346,
+      "step": 23212
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2204255033331656e-05,
+      "loss": 0.2572,
+      "step": 23213
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2203919198025289e-05,
+      "loss": 0.2576,
+      "step": 23214
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2203583362718923e-05,
+      "loss": 0.2237,
+      "step": 23215
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2203247527412558e-05,
+      "loss": 0.5434,
+      "step": 23216
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2202911692106193e-05,
+      "loss": 0.1298,
+      "step": 23217
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2202575856799826e-05,
+      "loss": 0.2627,
+      "step": 23218
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.220224002149346e-05,
+      "loss": 0.2982,
+      "step": 23219
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2201904186187095e-05,
+      "loss": 0.2683,
+      "step": 23220
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2201568350880728e-05,
+      "loss": 0.3251,
+      "step": 23221
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2201232515574364e-05,
+      "loss": 0.293,
+      "step": 23222
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2200896680267997e-05,
+      "loss": 0.221,
+      "step": 23223
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.220056084496163e-05,
+      "loss": 0.3713,
+      "step": 23224
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2200225009655266e-05,
+      "loss": 0.2395,
+      "step": 23225
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.21998891743489e-05,
+      "loss": 0.2527,
+      "step": 23226
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2199553339042534e-05,
+      "loss": 0.3396,
+      "step": 23227
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2199217503736169e-05,
+      "loss": 0.2841,
+      "step": 23228
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2198881668429803e-05,
+      "loss": 0.1952,
+      "step": 23229
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2198545833123436e-05,
+      "loss": 0.3063,
+      "step": 23230
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2198209997817073e-05,
+      "loss": 0.4278,
+      "step": 23231
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2197874162510706e-05,
+      "loss": 0.5453,
+      "step": 23232
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2197538327204339e-05,
+      "loss": 0.4092,
+      "step": 23233
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2197202491897975e-05,
+      "loss": 0.2474,
+      "step": 23234
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2196866656591608e-05,
+      "loss": 0.3062,
+      "step": 23235
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2196530821285243e-05,
+      "loss": 0.3662,
+      "step": 23236
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2196194985978877e-05,
+      "loss": 0.1874,
+      "step": 23237
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2195859150672512e-05,
+      "loss": 0.1997,
+      "step": 23238
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2195523315366145e-05,
+      "loss": 0.2555,
+      "step": 23239
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2195187480059781e-05,
+      "loss": 0.2031,
+      "step": 23240
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2194851644753414e-05,
+      "loss": 0.3439,
+      "step": 23241
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2194515809447047e-05,
+      "loss": 0.3471,
+      "step": 23242
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2194179974140683e-05,
+      "loss": 0.3774,
+      "step": 23243
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2193844138834316e-05,
+      "loss": 0.1967,
+      "step": 23244
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.219350830352795e-05,
+      "loss": 0.2877,
+      "step": 23245
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2193172468221586e-05,
+      "loss": 0.3082,
+      "step": 23246
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2192836632915219e-05,
+      "loss": 0.4372,
+      "step": 23247
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2192500797608853e-05,
+      "loss": 0.3256,
+      "step": 23248
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2192164962302488e-05,
+      "loss": 0.4446,
+      "step": 23249
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2191829126996122e-05,
+      "loss": 0.268,
+      "step": 23250
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2191493291689755e-05,
+      "loss": 0.2647,
+      "step": 23251
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2191157456383392e-05,
+      "loss": 0.3084,
+      "step": 23252
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2190821621077025e-05,
+      "loss": 0.2289,
+      "step": 23253
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2190485785770658e-05,
+      "loss": 0.265,
+      "step": 23254
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2190149950464294e-05,
+      "loss": 0.2914,
+      "step": 23255
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2189814115157927e-05,
+      "loss": 0.2237,
+      "step": 23256
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2189478279851562e-05,
+      "loss": 0.408,
+      "step": 23257
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2189142444545196e-05,
+      "loss": 0.2481,
+      "step": 23258
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2188806609238831e-05,
+      "loss": 0.138,
+      "step": 23259
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2188470773932464e-05,
+      "loss": 0.2979,
+      "step": 23260
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.21881349386261e-05,
+      "loss": 0.2229,
+      "step": 23261
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2187799103319733e-05,
+      "loss": 0.4023,
+      "step": 23262
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2187463268013366e-05,
+      "loss": 0.248,
+      "step": 23263
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2187127432707002e-05,
+      "loss": 0.2173,
+      "step": 23264
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2186791597400635e-05,
+      "loss": 0.2521,
+      "step": 23265
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2186455762094268e-05,
+      "loss": 0.1721,
+      "step": 23266
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2186119926787905e-05,
+      "loss": 0.2918,
+      "step": 23267
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2185784091481538e-05,
+      "loss": 0.2885,
+      "step": 23268
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2185448256175172e-05,
+      "loss": 0.3594,
+      "step": 23269
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2185112420868807e-05,
+      "loss": 0.3907,
+      "step": 23270
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2184776585562442e-05,
+      "loss": 0.1043,
+      "step": 23271
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2184440750256075e-05,
+      "loss": 0.3119,
+      "step": 23272
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2184104914949711e-05,
+      "loss": 0.304,
+      "step": 23273
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2183769079643344e-05,
+      "loss": 0.5482,
+      "step": 23274
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2183433244336977e-05,
+      "loss": 0.3859,
+      "step": 23275
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2183097409030613e-05,
+      "loss": 0.2152,
+      "step": 23276
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2182761573724246e-05,
+      "loss": 0.354,
+      "step": 23277
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.218242573841788e-05,
+      "loss": 0.4753,
+      "step": 23278
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2182089903111515e-05,
+      "loss": 0.295,
+      "step": 23279
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.218175406780515e-05,
+      "loss": 0.2109,
+      "step": 23280
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2181418232498783e-05,
+      "loss": 0.3725,
+      "step": 23281
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.218108239719242e-05,
+      "loss": 0.4763,
+      "step": 23282
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2180746561886052e-05,
+      "loss": 0.3205,
+      "step": 23283
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2180410726579685e-05,
+      "loss": 0.2537,
+      "step": 23284
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2180074891273322e-05,
+      "loss": 0.4544,
+      "step": 23285
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2179739055966955e-05,
+      "loss": 0.2435,
+      "step": 23286
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2179403220660587e-05,
+      "loss": 0.3098,
+      "step": 23287
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2179067385354224e-05,
+      "loss": 0.2495,
+      "step": 23288
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2178731550047857e-05,
+      "loss": 0.2636,
+      "step": 23289
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2178395714741491e-05,
+      "loss": 0.3918,
+      "step": 23290
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2178059879435126e-05,
+      "loss": 0.2857,
+      "step": 23291
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.217772404412876e-05,
+      "loss": 0.508,
+      "step": 23292
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2177388208822394e-05,
+      "loss": 0.3698,
+      "step": 23293
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.217705237351603e-05,
+      "loss": 0.2573,
+      "step": 23294
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2176716538209663e-05,
+      "loss": 0.243,
+      "step": 23295
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2176380702903296e-05,
+      "loss": 0.3938,
+      "step": 23296
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2176044867596932e-05,
+      "loss": 0.3867,
+      "step": 23297
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2175709032290565e-05,
+      "loss": 0.2579,
+      "step": 23298
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.21753731969842e-05,
+      "loss": 0.1989,
+      "step": 23299
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2175037361677835e-05,
+      "loss": 0.3518,
+      "step": 23300
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.217470152637147e-05,
+      "loss": 0.1897,
+      "step": 23301
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2174365691065102e-05,
+      "loss": 0.4413,
+      "step": 23302
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2174029855758738e-05,
+      "loss": 0.3095,
+      "step": 23303
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2173694020452371e-05,
+      "loss": 0.4572,
+      "step": 23304
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2173358185146004e-05,
+      "loss": 0.2666,
+      "step": 23305
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.217302234983964e-05,
+      "loss": 0.283,
+      "step": 23306
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2172686514533274e-05,
+      "loss": 0.1958,
+      "step": 23307
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2172350679226908e-05,
+      "loss": 0.304,
+      "step": 23308
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2172014843920543e-05,
+      "loss": 0.305,
+      "step": 23309
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2171679008614176e-05,
+      "loss": 0.2385,
+      "step": 23310
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.217134317330781e-05,
+      "loss": 0.2227,
+      "step": 23311
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2171007338001445e-05,
+      "loss": 0.191,
+      "step": 23312
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.217067150269508e-05,
+      "loss": 0.2628,
+      "step": 23313
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2170335667388713e-05,
+      "loss": 0.2465,
+      "step": 23314
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2169999832082349e-05,
+      "loss": 0.3627,
+      "step": 23315
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2169663996775982e-05,
+      "loss": 0.2299,
+      "step": 23316
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2169328161469615e-05,
+      "loss": 0.3117,
+      "step": 23317
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2168992326163251e-05,
+      "loss": 0.2071,
+      "step": 23318
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2168656490856884e-05,
+      "loss": 0.3566,
+      "step": 23319
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2168320655550519e-05,
+      "loss": 0.4314,
+      "step": 23320
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2167984820244154e-05,
+      "loss": 0.2485,
+      "step": 23321
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2167648984937788e-05,
+      "loss": 0.2532,
+      "step": 23322
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2167313149631421e-05,
+      "loss": 0.3424,
+      "step": 23323
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.2166977314325058e-05,
+      "loss": 0.3085,
+      "step": 23324
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.216664147901869e-05,
+      "loss": 0.5552,
+      "step": 23325
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2166305643712323e-05,
+      "loss": 0.2325,
+      "step": 23326
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.216596980840596e-05,
+      "loss": 0.1488,
+      "step": 23327
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2165633973099593e-05,
+      "loss": 0.2032,
+      "step": 23328
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2165298137793227e-05,
+      "loss": 0.2856,
+      "step": 23329
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2164962302486862e-05,
+      "loss": 0.203,
+      "step": 23330
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2164626467180497e-05,
+      "loss": 0.3625,
+      "step": 23331
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.216429063187413e-05,
+      "loss": 0.1759,
+      "step": 23332
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2163954796567764e-05,
+      "loss": 0.325,
+      "step": 23333
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2163618961261399e-05,
+      "loss": 0.233,
+      "step": 23334
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2163283125955032e-05,
+      "loss": 0.2896,
+      "step": 23335
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2162947290648668e-05,
+      "loss": 0.3934,
+      "step": 23336
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2162611455342301e-05,
+      "loss": 0.2204,
+      "step": 23337
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2162275620035934e-05,
+      "loss": 0.3753,
+      "step": 23338
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.216193978472957e-05,
+      "loss": 0.3156,
+      "step": 23339
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2161603949423203e-05,
+      "loss": 0.3789,
+      "step": 23340
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2161268114116838e-05,
+      "loss": 0.3887,
+      "step": 23341
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2160932278810473e-05,
+      "loss": 0.2962,
+      "step": 23342
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2160596443504107e-05,
+      "loss": 0.4145,
+      "step": 23343
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.216026060819774e-05,
+      "loss": 0.4083,
+      "step": 23344
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2159924772891377e-05,
+      "loss": 0.4671,
+      "step": 23345
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.215958893758501e-05,
+      "loss": 0.3931,
+      "step": 23346
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2159253102278643e-05,
+      "loss": 0.395,
+      "step": 23347
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2158917266972279e-05,
+      "loss": 0.2051,
+      "step": 23348
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2158581431665912e-05,
+      "loss": 0.2958,
+      "step": 23349
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2158245596359547e-05,
+      "loss": 0.4352,
+      "step": 23350
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2157909761053181e-05,
+      "loss": 0.3477,
+      "step": 23351
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2157573925746816e-05,
+      "loss": 0.2695,
+      "step": 23352
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2157238090440449e-05,
+      "loss": 0.2141,
+      "step": 23353
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2156902255134083e-05,
+      "loss": 0.3028,
+      "step": 23354
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2156566419827718e-05,
+      "loss": 0.2726,
+      "step": 23355
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2156230584521351e-05,
+      "loss": 0.244,
+      "step": 23356
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2155894749214987e-05,
+      "loss": 0.3957,
+      "step": 23357
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.215555891390862e-05,
+      "loss": 0.3568,
+      "step": 23358
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2155223078602253e-05,
+      "loss": 0.3411,
+      "step": 23359
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.215488724329589e-05,
+      "loss": 0.3562,
+      "step": 23360
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2154551407989523e-05,
+      "loss": 0.3523,
+      "step": 23361
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2154215572683157e-05,
+      "loss": 0.477,
+      "step": 23362
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2153879737376792e-05,
+      "loss": 0.3694,
+      "step": 23363
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2153543902070427e-05,
+      "loss": 0.248,
+      "step": 23364
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.215320806676406e-05,
+      "loss": 0.3306,
+      "step": 23365
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2152872231457696e-05,
+      "loss": 0.3677,
+      "step": 23366
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2152536396151329e-05,
+      "loss": 0.3551,
+      "step": 23367
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2152200560844962e-05,
+      "loss": 0.3079,
+      "step": 23368
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2151864725538598e-05,
+      "loss": 0.2553,
+      "step": 23369
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2151528890232231e-05,
+      "loss": 0.3081,
+      "step": 23370
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2151193054925866e-05,
+      "loss": 0.1973,
+      "step": 23371
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2150857219619499e-05,
+      "loss": 0.2421,
+      "step": 23372
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2150521384313135e-05,
+      "loss": 0.3934,
+      "step": 23373
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2150185549006768e-05,
+      "loss": 0.3189,
+      "step": 23374
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2149849713700401e-05,
+      "loss": 0.4323,
+      "step": 23375
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2149513878394037e-05,
+      "loss": 0.2236,
+      "step": 23376
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.214917804308767e-05,
+      "loss": 0.3375,
+      "step": 23377
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2148842207781303e-05,
+      "loss": 0.3568,
+      "step": 23378
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.214850637247494e-05,
+      "loss": 0.5045,
+      "step": 23379
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2148170537168572e-05,
+      "loss": 0.3542,
+      "step": 23380
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2147834701862207e-05,
+      "loss": 0.1536,
+      "step": 23381
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2147498866555842e-05,
+      "loss": 0.3043,
+      "step": 23382
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2147163031249476e-05,
+      "loss": 0.2448,
+      "step": 23383
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.214682719594311e-05,
+      "loss": 0.3267,
+      "step": 23384
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2146491360636746e-05,
+      "loss": 0.1884,
+      "step": 23385
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2146155525330379e-05,
+      "loss": 0.2435,
+      "step": 23386
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2145819690024012e-05,
+      "loss": 0.1713,
+      "step": 23387
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2145483854717648e-05,
+      "loss": 0.4358,
+      "step": 23388
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2145148019411281e-05,
+      "loss": 0.2073,
+      "step": 23389
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2144812184104916e-05,
+      "loss": 0.2874,
+      "step": 23390
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.214447634879855e-05,
+      "loss": 0.2341,
+      "step": 23391
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2144140513492185e-05,
+      "loss": 0.2904,
+      "step": 23392
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2143804678185818e-05,
+      "loss": 0.3776,
+      "step": 23393
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2143468842879454e-05,
+      "loss": 0.286,
+      "step": 23394
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2143133007573087e-05,
+      "loss": 0.3745,
+      "step": 23395
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.214279717226672e-05,
+      "loss": 0.2704,
+      "step": 23396
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2142461336960356e-05,
+      "loss": 0.1815,
+      "step": 23397
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.214212550165399e-05,
+      "loss": 0.5983,
+      "step": 23398
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2141789666347624e-05,
+      "loss": 0.2684,
+      "step": 23399
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2141453831041259e-05,
+      "loss": 0.1599,
+      "step": 23400
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2141117995734892e-05,
+      "loss": 0.2158,
+      "step": 23401
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2140782160428526e-05,
+      "loss": 0.1611,
+      "step": 23402
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2140446325122161e-05,
+      "loss": 0.263,
+      "step": 23403
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2140110489815795e-05,
+      "loss": 0.3859,
+      "step": 23404
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2139774654509428e-05,
+      "loss": 0.2093,
+      "step": 23405
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2139438819203065e-05,
+      "loss": 0.2805,
+      "step": 23406
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2139102983896698e-05,
+      "loss": 0.1808,
+      "step": 23407
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.213876714859033e-05,
+      "loss": 0.2629,
+      "step": 23408
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2138431313283967e-05,
+      "loss": 0.4382,
+      "step": 23409
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.21380954779776e-05,
+      "loss": 0.4826,
+      "step": 23410
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2137759642671235e-05,
+      "loss": 0.2608,
+      "step": 23411
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.213742380736487e-05,
+      "loss": 0.2073,
+      "step": 23412
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2137087972058504e-05,
+      "loss": 0.3035,
+      "step": 23413
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2136752136752137e-05,
+      "loss": 0.3642,
+      "step": 23414
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2136416301445773e-05,
+      "loss": 0.3664,
+      "step": 23415
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2136080466139406e-05,
+      "loss": 0.363,
+      "step": 23416
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2135744630833039e-05,
+      "loss": 0.2824,
+      "step": 23417
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2135408795526675e-05,
+      "loss": 0.4553,
+      "step": 23418
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2135072960220308e-05,
+      "loss": 0.2431,
+      "step": 23419
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2134737124913943e-05,
+      "loss": 0.4562,
+      "step": 23420
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2134401289607578e-05,
+      "loss": 0.4338,
+      "step": 23421
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.213406545430121e-05,
+      "loss": 0.3351,
+      "step": 23422
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2133729618994845e-05,
+      "loss": 0.3854,
+      "step": 23423
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.213339378368848e-05,
+      "loss": 0.3002,
+      "step": 23424
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2133057948382115e-05,
+      "loss": 0.3036,
+      "step": 23425
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2132722113075748e-05,
+      "loss": 0.3477,
+      "step": 23426
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2132386277769384e-05,
+      "loss": 0.3709,
+      "step": 23427
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2132050442463017e-05,
+      "loss": 0.2211,
+      "step": 23428
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.213171460715665e-05,
+      "loss": 0.3529,
+      "step": 23429
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2131378771850286e-05,
+      "loss": 0.3296,
+      "step": 23430
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2131042936543919e-05,
+      "loss": 0.3978,
+      "step": 23431
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2130707101237554e-05,
+      "loss": 0.308,
+      "step": 23432
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2130371265931188e-05,
+      "loss": 0.3551,
+      "step": 23433
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2130035430624823e-05,
+      "loss": 0.2108,
+      "step": 23434
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2129699595318456e-05,
+      "loss": 0.2591,
+      "step": 23435
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2129363760012092e-05,
+      "loss": 0.2453,
+      "step": 23436
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2129027924705725e-05,
+      "loss": 0.2052,
+      "step": 23437
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2128692089399358e-05,
+      "loss": 0.3371,
+      "step": 23438
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2128356254092995e-05,
+      "loss": 0.4945,
+      "step": 23439
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2128020418786628e-05,
+      "loss": 0.2963,
+      "step": 23440
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2127684583480262e-05,
+      "loss": 0.3134,
+      "step": 23441
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2127348748173897e-05,
+      "loss": 0.2163,
+      "step": 23442
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2127012912867532e-05,
+      "loss": 0.1932,
+      "step": 23443
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2126677077561164e-05,
+      "loss": 0.1512,
+      "step": 23444
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2126341242254799e-05,
+      "loss": 0.5219,
+      "step": 23445
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2126005406948434e-05,
+      "loss": 0.174,
+      "step": 23446
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2125669571642067e-05,
+      "loss": 0.2367,
+      "step": 23447
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2125333736335703e-05,
+      "loss": 0.3109,
+      "step": 23448
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2124997901029336e-05,
+      "loss": 0.3049,
+      "step": 23449
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2124662065722969e-05,
+      "loss": 0.3283,
+      "step": 23450
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2124326230416605e-05,
+      "loss": 0.2699,
+      "step": 23451
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2123990395110238e-05,
+      "loss": 0.5343,
+      "step": 23452
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2123654559803873e-05,
+      "loss": 0.5899,
+      "step": 23453
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2123318724497508e-05,
+      "loss": 0.2324,
+      "step": 23454
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2122982889191142e-05,
+      "loss": 0.2865,
+      "step": 23455
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2122647053884775e-05,
+      "loss": 0.2528,
+      "step": 23456
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2122311218578411e-05,
+      "loss": 0.4709,
+      "step": 23457
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2121975383272044e-05,
+      "loss": 0.4316,
+      "step": 23458
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2121639547965677e-05,
+      "loss": 0.4111,
+      "step": 23459
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2121303712659314e-05,
+      "loss": 0.2892,
+      "step": 23460
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2120967877352947e-05,
+      "loss": 0.2944,
+      "step": 23461
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2120632042046581e-05,
+      "loss": 0.3776,
+      "step": 23462
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2120296206740216e-05,
+      "loss": 0.2291,
+      "step": 23463
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.211996037143385e-05,
+      "loss": 0.4563,
+      "step": 23464
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2119624536127484e-05,
+      "loss": 0.3633,
+      "step": 23465
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.211928870082112e-05,
+      "loss": 0.3269,
+      "step": 23466
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2118952865514753e-05,
+      "loss": 0.2089,
+      "step": 23467
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2118617030208386e-05,
+      "loss": 0.3114,
+      "step": 23468
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2118281194902022e-05,
+      "loss": 0.4706,
+      "step": 23469
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2117945359595655e-05,
+      "loss": 0.3221,
+      "step": 23470
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2117609524289288e-05,
+      "loss": 0.2946,
+      "step": 23471
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2117273688982924e-05,
+      "loss": 0.2604,
+      "step": 23472
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2116937853676557e-05,
+      "loss": 0.3158,
+      "step": 23473
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2116602018370192e-05,
+      "loss": 0.3413,
+      "step": 23474
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2116266183063827e-05,
+      "loss": 0.4587,
+      "step": 23475
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2115930347757461e-05,
+      "loss": 0.1996,
+      "step": 23476
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2115594512451094e-05,
+      "loss": 0.2408,
+      "step": 23477
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.211525867714473e-05,
+      "loss": 0.2446,
+      "step": 23478
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2114922841838364e-05,
+      "loss": 0.2191,
+      "step": 23479
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2114587006531997e-05,
+      "loss": 0.1581,
+      "step": 23480
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2114251171225633e-05,
+      "loss": 0.3966,
+      "step": 23481
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2113915335919266e-05,
+      "loss": 0.3723,
+      "step": 23482
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.21135795006129e-05,
+      "loss": 0.2101,
+      "step": 23483
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2113243665306535e-05,
+      "loss": 0.3049,
+      "step": 23484
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.211290783000017e-05,
+      "loss": 0.2307,
+      "step": 23485
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2112571994693803e-05,
+      "loss": 0.3573,
+      "step": 23486
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2112236159387439e-05,
+      "loss": 0.4084,
+      "step": 23487
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2111900324081072e-05,
+      "loss": 0.3266,
+      "step": 23488
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2111564488774705e-05,
+      "loss": 0.3144,
+      "step": 23489
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2111228653468341e-05,
+      "loss": 0.2769,
+      "step": 23490
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2110892818161974e-05,
+      "loss": 0.2846,
+      "step": 23491
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2110556982855607e-05,
+      "loss": 0.2226,
+      "step": 23492
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2110221147549244e-05,
+      "loss": 0.2557,
+      "step": 23493
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2109885312242876e-05,
+      "loss": 0.2667,
+      "step": 23494
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2109549476936511e-05,
+      "loss": 0.2778,
+      "step": 23495
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2109213641630146e-05,
+      "loss": 0.2152,
+      "step": 23496
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.210887780632378e-05,
+      "loss": 0.2074,
+      "step": 23497
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2108541971017413e-05,
+      "loss": 0.2868,
+      "step": 23498
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.210820613571105e-05,
+      "loss": 0.2028,
+      "step": 23499
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2107870300404683e-05,
+      "loss": 0.2517,
+      "step": 23500
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2107534465098316e-05,
+      "loss": 0.2322,
+      "step": 23501
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2107198629791952e-05,
+      "loss": 0.2664,
+      "step": 23502
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2106862794485585e-05,
+      "loss": 0.2816,
+      "step": 23503
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.210652695917922e-05,
+      "loss": 0.3688,
+      "step": 23504
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2106191123872854e-05,
+      "loss": 0.2135,
+      "step": 23505
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2105855288566489e-05,
+      "loss": 0.2216,
+      "step": 23506
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2105519453260122e-05,
+      "loss": 0.2912,
+      "step": 23507
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2105183617953758e-05,
+      "loss": 0.2988,
+      "step": 23508
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2104847782647391e-05,
+      "loss": 0.261,
+      "step": 23509
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2104511947341024e-05,
+      "loss": 0.2714,
+      "step": 23510
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.210417611203466e-05,
+      "loss": 0.2505,
+      "step": 23511
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2103840276728293e-05,
+      "loss": 0.3952,
+      "step": 23512
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2103504441421926e-05,
+      "loss": 0.279,
+      "step": 23513
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2103168606115563e-05,
+      "loss": 0.1909,
+      "step": 23514
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2102832770809196e-05,
+      "loss": 0.386,
+      "step": 23515
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.210249693550283e-05,
+      "loss": 0.2983,
+      "step": 23516
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2102161100196465e-05,
+      "loss": 0.3399,
+      "step": 23517
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.21018252648901e-05,
+      "loss": 0.4068,
+      "step": 23518
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2101489429583733e-05,
+      "loss": 0.2395,
+      "step": 23519
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2101153594277369e-05,
+      "loss": 0.2594,
+      "step": 23520
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2100817758971002e-05,
+      "loss": 0.4384,
+      "step": 23521
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2100481923664635e-05,
+      "loss": 0.2814,
+      "step": 23522
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 1.2100146088358271e-05,
+      "loss": 0.2035,
+      "step": 23523
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2099810253051904e-05,
+      "loss": 0.3717,
+      "step": 23524
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2099474417745539e-05,
+      "loss": 0.3581,
+      "step": 23525
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2099138582439173e-05,
+      "loss": 0.2915,
+      "step": 23526
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2098802747132808e-05,
+      "loss": 0.3698,
+      "step": 23527
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2098466911826441e-05,
+      "loss": 0.2098,
+      "step": 23528
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2098131076520077e-05,
+      "loss": 0.2391,
+      "step": 23529
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.209779524121371e-05,
+      "loss": 0.4206,
+      "step": 23530
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2097459405907343e-05,
+      "loss": 0.2067,
+      "step": 23531
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.209712357060098e-05,
+      "loss": 0.4839,
+      "step": 23532
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2096787735294612e-05,
+      "loss": 0.4614,
+      "step": 23533
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2096451899988247e-05,
+      "loss": 0.3357,
+      "step": 23534
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2096116064681882e-05,
+      "loss": 0.1703,
+      "step": 23535
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2095780229375515e-05,
+      "loss": 0.3306,
+      "step": 23536
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.209544439406915e-05,
+      "loss": 0.3193,
+      "step": 23537
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2095108558762784e-05,
+      "loss": 0.2544,
+      "step": 23538
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2094772723456419e-05,
+      "loss": 0.3165,
+      "step": 23539
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2094436888150052e-05,
+      "loss": 0.3524,
+      "step": 23540
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2094101052843688e-05,
+      "loss": 0.2789,
+      "step": 23541
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2093765217537321e-05,
+      "loss": 0.3082,
+      "step": 23542
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2093429382230954e-05,
+      "loss": 0.2989,
+      "step": 23543
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.209309354692459e-05,
+      "loss": 0.3574,
+      "step": 23544
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2092757711618223e-05,
+      "loss": 0.2135,
+      "step": 23545
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2092421876311858e-05,
+      "loss": 0.3811,
+      "step": 23546
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2092086041005492e-05,
+      "loss": 0.2806,
+      "step": 23547
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2091750205699127e-05,
+      "loss": 0.2774,
+      "step": 23548
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.209141437039276e-05,
+      "loss": 0.1645,
+      "step": 23549
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2091078535086396e-05,
+      "loss": 0.3968,
+      "step": 23550
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.209074269978003e-05,
+      "loss": 0.3421,
+      "step": 23551
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2090406864473662e-05,
+      "loss": 0.2912,
+      "step": 23552
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2090071029167297e-05,
+      "loss": 0.1817,
+      "step": 23553
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2089735193860932e-05,
+      "loss": 0.2191,
+      "step": 23554
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2089399358554566e-05,
+      "loss": 0.4334,
+      "step": 23555
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.20890635232482e-05,
+      "loss": 0.2081,
+      "step": 23556
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2088727687941834e-05,
+      "loss": 0.4509,
+      "step": 23557
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2088391852635469e-05,
+      "loss": 0.2307,
+      "step": 23558
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2088056017329101e-05,
+      "loss": 0.1589,
+      "step": 23559
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2087720182022738e-05,
+      "loss": 0.2211,
+      "step": 23560
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.208738434671637e-05,
+      "loss": 0.2921,
+      "step": 23561
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2087048511410004e-05,
+      "loss": 0.2583,
+      "step": 23562
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.208671267610364e-05,
+      "loss": 0.2575,
+      "step": 23563
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2086376840797273e-05,
+      "loss": 0.1158,
+      "step": 23564
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2086041005490908e-05,
+      "loss": 0.315,
+      "step": 23565
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2085705170184542e-05,
+      "loss": 0.2156,
+      "step": 23566
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2085369334878177e-05,
+      "loss": 0.1871,
+      "step": 23567
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.208503349957181e-05,
+      "loss": 0.1851,
+      "step": 23568
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2084697664265446e-05,
+      "loss": 0.3002,
+      "step": 23569
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.208436182895908e-05,
+      "loss": 0.2132,
+      "step": 23570
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2084025993652712e-05,
+      "loss": 0.3108,
+      "step": 23571
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2083690158346348e-05,
+      "loss": 0.3015,
+      "step": 23572
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2083354323039981e-05,
+      "loss": 0.2987,
+      "step": 23573
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2083018487733616e-05,
+      "loss": 0.1566,
+      "step": 23574
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.208268265242725e-05,
+      "loss": 0.2825,
+      "step": 23575
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2082346817120885e-05,
+      "loss": 0.2931,
+      "step": 23576
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2082010981814518e-05,
+      "loss": 0.453,
+      "step": 23577
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2081675146508155e-05,
+      "loss": 0.3096,
+      "step": 23578
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2081339311201788e-05,
+      "loss": 0.2486,
+      "step": 23579
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.208100347589542e-05,
+      "loss": 0.2935,
+      "step": 23580
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2080667640589057e-05,
+      "loss": 0.1817,
+      "step": 23581
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.208033180528269e-05,
+      "loss": 0.2264,
+      "step": 23582
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2079995969976323e-05,
+      "loss": 0.1757,
+      "step": 23583
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.207966013466996e-05,
+      "loss": 0.5881,
+      "step": 23584
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2079324299363592e-05,
+      "loss": 0.2749,
+      "step": 23585
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2078988464057227e-05,
+      "loss": 0.2004,
+      "step": 23586
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2078652628750861e-05,
+      "loss": 0.2676,
+      "step": 23587
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2078316793444496e-05,
+      "loss": 0.2176,
+      "step": 23588
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2077980958138129e-05,
+      "loss": 0.2082,
+      "step": 23589
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2077645122831765e-05,
+      "loss": 0.2416,
+      "step": 23590
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2077309287525398e-05,
+      "loss": 0.251,
+      "step": 23591
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2076973452219031e-05,
+      "loss": 0.3645,
+      "step": 23592
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2076637616912668e-05,
+      "loss": 0.2055,
+      "step": 23593
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.20763017816063e-05,
+      "loss": 0.3491,
+      "step": 23594
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2075965946299935e-05,
+      "loss": 0.1386,
+      "step": 23595
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.207563011099357e-05,
+      "loss": 0.34,
+      "step": 23596
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2075294275687205e-05,
+      "loss": 0.2523,
+      "step": 23597
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2074958440380837e-05,
+      "loss": 0.1817,
+      "step": 23598
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2074622605074474e-05,
+      "loss": 0.3229,
+      "step": 23599
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2074286769768107e-05,
+      "loss": 0.238,
+      "step": 23600
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.207395093446174e-05,
+      "loss": 0.3875,
+      "step": 23601
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2073615099155376e-05,
+      "loss": 0.095,
+      "step": 23602
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2073279263849009e-05,
+      "loss": 0.2916,
+      "step": 23603
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2072943428542642e-05,
+      "loss": 0.3401,
+      "step": 23604
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2072607593236278e-05,
+      "loss": 0.4046,
+      "step": 23605
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2072271757929911e-05,
+      "loss": 0.2792,
+      "step": 23606
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2071935922623546e-05,
+      "loss": 0.2649,
+      "step": 23607
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.207160008731718e-05,
+      "loss": 0.3566,
+      "step": 23608
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2071264252010815e-05,
+      "loss": 0.3988,
+      "step": 23609
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2070928416704448e-05,
+      "loss": 0.1822,
+      "step": 23610
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2070592581398084e-05,
+      "loss": 0.3606,
+      "step": 23611
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2070256746091717e-05,
+      "loss": 0.3048,
+      "step": 23612
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.206992091078535e-05,
+      "loss": 0.1913,
+      "step": 23613
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2069585075478987e-05,
+      "loss": 0.3756,
+      "step": 23614
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.206924924017262e-05,
+      "loss": 0.2446,
+      "step": 23615
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2068913404866254e-05,
+      "loss": 0.2532,
+      "step": 23616
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2068577569559889e-05,
+      "loss": 0.2298,
+      "step": 23617
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2068241734253524e-05,
+      "loss": 0.2403,
+      "step": 23618
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2067905898947157e-05,
+      "loss": 0.1327,
+      "step": 23619
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2067570063640793e-05,
+      "loss": 0.3798,
+      "step": 23620
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2067234228334426e-05,
+      "loss": 0.2224,
+      "step": 23621
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2066898393028059e-05,
+      "loss": 0.2372,
+      "step": 23622
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2066562557721695e-05,
+      "loss": 0.245,
+      "step": 23623
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2066226722415328e-05,
+      "loss": 0.4264,
+      "step": 23624
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2065890887108963e-05,
+      "loss": 0.3031,
+      "step": 23625
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2065555051802597e-05,
+      "loss": 0.2247,
+      "step": 23626
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.206521921649623e-05,
+      "loss": 0.2485,
+      "step": 23627
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2064883381189865e-05,
+      "loss": 0.348,
+      "step": 23628
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.20645475458835e-05,
+      "loss": 0.2925,
+      "step": 23629
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2064211710577134e-05,
+      "loss": 0.3029,
+      "step": 23630
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2063875875270767e-05,
+      "loss": 0.228,
+      "step": 23631
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2063540039964404e-05,
+      "loss": 0.3241,
+      "step": 23632
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2063204204658037e-05,
+      "loss": 0.303,
+      "step": 23633
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.206286836935167e-05,
+      "loss": 0.4706,
+      "step": 23634
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2062532534045306e-05,
+      "loss": 0.3448,
+      "step": 23635
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2062196698738939e-05,
+      "loss": 0.2899,
+      "step": 23636
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2061860863432573e-05,
+      "loss": 0.1334,
+      "step": 23637
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2061525028126208e-05,
+      "loss": 0.2885,
+      "step": 23638
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2061189192819843e-05,
+      "loss": 0.2443,
+      "step": 23639
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2060853357513476e-05,
+      "loss": 0.3524,
+      "step": 23640
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2060517522207112e-05,
+      "loss": 0.2017,
+      "step": 23641
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2060181686900745e-05,
+      "loss": 0.3505,
+      "step": 23642
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2059845851594378e-05,
+      "loss": 0.2415,
+      "step": 23643
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2059510016288014e-05,
+      "loss": 0.333,
+      "step": 23644
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2059174180981647e-05,
+      "loss": 0.3703,
+      "step": 23645
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2058838345675282e-05,
+      "loss": 0.4458,
+      "step": 23646
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2058502510368917e-05,
+      "loss": 0.2347,
+      "step": 23647
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.205816667506255e-05,
+      "loss": 0.2678,
+      "step": 23648
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2057830839756184e-05,
+      "loss": 0.3036,
+      "step": 23649
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2057495004449819e-05,
+      "loss": 0.5345,
+      "step": 23650
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2057159169143453e-05,
+      "loss": 0.5513,
+      "step": 23651
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2056823333837086e-05,
+      "loss": 0.416,
+      "step": 23652
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2056487498530723e-05,
+      "loss": 0.3184,
+      "step": 23653
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2056151663224356e-05,
+      "loss": 0.2547,
+      "step": 23654
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2055815827917989e-05,
+      "loss": 0.3448,
+      "step": 23655
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2055479992611625e-05,
+      "loss": 0.2956,
+      "step": 23656
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2055144157305258e-05,
+      "loss": 0.31,
+      "step": 23657
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2054808321998893e-05,
+      "loss": 0.2233,
+      "step": 23658
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2054472486692527e-05,
+      "loss": 0.1951,
+      "step": 23659
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2054136651386162e-05,
+      "loss": 0.4036,
+      "step": 23660
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2053800816079795e-05,
+      "loss": 0.5052,
+      "step": 23661
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2053464980773431e-05,
+      "loss": 0.314,
+      "step": 23662
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2053129145467064e-05,
+      "loss": 0.4017,
+      "step": 23663
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2052793310160697e-05,
+      "loss": 0.3076,
+      "step": 23664
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2052457474854333e-05,
+      "loss": 0.3453,
+      "step": 23665
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2052121639547966e-05,
+      "loss": 0.1776,
+      "step": 23666
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2051785804241601e-05,
+      "loss": 0.2534,
+      "step": 23667
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2051449968935236e-05,
+      "loss": 0.4239,
+      "step": 23668
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.205111413362887e-05,
+      "loss": 0.3072,
+      "step": 23669
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2050778298322503e-05,
+      "loss": 0.2404,
+      "step": 23670
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2050442463016138e-05,
+      "loss": 0.2441,
+      "step": 23671
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2050106627709773e-05,
+      "loss": 0.2997,
+      "step": 23672
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2049770792403406e-05,
+      "loss": 0.1996,
+      "step": 23673
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2049434957097042e-05,
+      "loss": 0.1833,
+      "step": 23674
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2049099121790675e-05,
+      "loss": 0.257,
+      "step": 23675
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2048763286484308e-05,
+      "loss": 0.1174,
+      "step": 23676
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2048427451177944e-05,
+      "loss": 0.3227,
+      "step": 23677
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2048091615871577e-05,
+      "loss": 0.3181,
+      "step": 23678
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2047755780565212e-05,
+      "loss": 0.2991,
+      "step": 23679
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2047419945258846e-05,
+      "loss": 0.3185,
+      "step": 23680
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2047084109952481e-05,
+      "loss": 0.3035,
+      "step": 23681
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2046748274646114e-05,
+      "loss": 0.4775,
+      "step": 23682
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.204641243933975e-05,
+      "loss": 0.3206,
+      "step": 23683
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2046076604033383e-05,
+      "loss": 0.2241,
+      "step": 23684
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2045740768727016e-05,
+      "loss": 0.328,
+      "step": 23685
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2045404933420653e-05,
+      "loss": 0.3051,
+      "step": 23686
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2045069098114286e-05,
+      "loss": 0.2467,
+      "step": 23687
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.204473326280792e-05,
+      "loss": 0.2277,
+      "step": 23688
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2044397427501555e-05,
+      "loss": 0.3789,
+      "step": 23689
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.204406159219519e-05,
+      "loss": 0.18,
+      "step": 23690
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2043725756888822e-05,
+      "loss": 0.2322,
+      "step": 23691
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2043389921582457e-05,
+      "loss": 0.3393,
+      "step": 23692
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2043054086276092e-05,
+      "loss": 0.2829,
+      "step": 23693
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2042718250969725e-05,
+      "loss": 0.2864,
+      "step": 23694
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2042382415663361e-05,
+      "loss": 0.371,
+      "step": 23695
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2042046580356994e-05,
+      "loss": 0.1917,
+      "step": 23696
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2041710745050627e-05,
+      "loss": 0.3387,
+      "step": 23697
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2041374909744263e-05,
+      "loss": 0.2326,
+      "step": 23698
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2041039074437896e-05,
+      "loss": 0.2369,
+      "step": 23699
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.204070323913153e-05,
+      "loss": 0.492,
+      "step": 23700
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2040367403825165e-05,
+      "loss": 0.4134,
+      "step": 23701
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.20400315685188e-05,
+      "loss": 0.3297,
+      "step": 23702
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2039695733212433e-05,
+      "loss": 0.1299,
+      "step": 23703
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.203935989790607e-05,
+      "loss": 0.2399,
+      "step": 23704
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2039024062599702e-05,
+      "loss": 0.2761,
+      "step": 23705
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2038688227293335e-05,
+      "loss": 0.3104,
+      "step": 23706
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2038352391986972e-05,
+      "loss": 0.1854,
+      "step": 23707
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2038016556680605e-05,
+      "loss": 0.1043,
+      "step": 23708
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.203768072137424e-05,
+      "loss": 0.2772,
+      "step": 23709
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2037344886067874e-05,
+      "loss": 0.3931,
+      "step": 23710
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2037009050761509e-05,
+      "loss": 0.1882,
+      "step": 23711
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2036673215455142e-05,
+      "loss": 0.1837,
+      "step": 23712
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2036337380148778e-05,
+      "loss": 0.4577,
+      "step": 23713
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.203600154484241e-05,
+      "loss": 0.2645,
+      "step": 23714
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2035665709536044e-05,
+      "loss": 0.326,
+      "step": 23715
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.203532987422968e-05,
+      "loss": 0.204,
+      "step": 23716
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2034994038923313e-05,
+      "loss": 0.3611,
+      "step": 23717
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2034658203616946e-05,
+      "loss": 0.2516,
+      "step": 23718
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2034322368310582e-05,
+      "loss": 0.4091,
+      "step": 23719
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.2033986533004215e-05,
+      "loss": 0.4514,
+      "step": 23720
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 1.203365069769785e-05,
+      "loss": 0.3582,
+      "step": 23721
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2033314862391485e-05,
+      "loss": 0.2779,
+      "step": 23722
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.203297902708512e-05,
+      "loss": 0.3009,
+      "step": 23723
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2032643191778752e-05,
+      "loss": 0.2534,
+      "step": 23724
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2032307356472389e-05,
+      "loss": 0.267,
+      "step": 23725
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2031971521166022e-05,
+      "loss": 0.182,
+      "step": 23726
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2031635685859654e-05,
+      "loss": 0.1581,
+      "step": 23727
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.203129985055329e-05,
+      "loss": 0.3278,
+      "step": 23728
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2030964015246924e-05,
+      "loss": 0.3425,
+      "step": 23729
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2030628179940558e-05,
+      "loss": 0.4147,
+      "step": 23730
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2030292344634191e-05,
+      "loss": 0.2215,
+      "step": 23731
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2029956509327828e-05,
+      "loss": 0.3529,
+      "step": 23732
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.202962067402146e-05,
+      "loss": 0.1539,
+      "step": 23733
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2029284838715094e-05,
+      "loss": 0.2417,
+      "step": 23734
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.202894900340873e-05,
+      "loss": 0.3696,
+      "step": 23735
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2028613168102363e-05,
+      "loss": 0.2609,
+      "step": 23736
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2028277332795998e-05,
+      "loss": 0.2083,
+      "step": 23737
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2027941497489632e-05,
+      "loss": 0.2923,
+      "step": 23738
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2027605662183265e-05,
+      "loss": 0.4035,
+      "step": 23739
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.20272698268769e-05,
+      "loss": 0.3902,
+      "step": 23740
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2026933991570534e-05,
+      "loss": 0.3318,
+      "step": 23741
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2026598156264169e-05,
+      "loss": 0.4526,
+      "step": 23742
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2026262320957802e-05,
+      "loss": 0.2519,
+      "step": 23743
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2025926485651438e-05,
+      "loss": 0.5195,
+      "step": 23744
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2025590650345071e-05,
+      "loss": 0.2566,
+      "step": 23745
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2025254815038704e-05,
+      "loss": 0.3677,
+      "step": 23746
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.202491897973234e-05,
+      "loss": 0.4111,
+      "step": 23747
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2024583144425974e-05,
+      "loss": 0.2995,
+      "step": 23748
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2024247309119608e-05,
+      "loss": 0.2924,
+      "step": 23749
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2023911473813243e-05,
+      "loss": 0.5397,
+      "step": 23750
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2023575638506878e-05,
+      "loss": 0.2107,
+      "step": 23751
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.202323980320051e-05,
+      "loss": 0.2257,
+      "step": 23752
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2022903967894147e-05,
+      "loss": 0.2794,
+      "step": 23753
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.202256813258778e-05,
+      "loss": 0.211,
+      "step": 23754
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2022232297281413e-05,
+      "loss": 0.3711,
+      "step": 23755
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2021896461975049e-05,
+      "loss": 0.1305,
+      "step": 23756
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2021560626668682e-05,
+      "loss": 0.2783,
+      "step": 23757
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2021224791362317e-05,
+      "loss": 0.2864,
+      "step": 23758
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2020888956055951e-05,
+      "loss": 0.1872,
+      "step": 23759
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2020553120749586e-05,
+      "loss": 0.1752,
+      "step": 23760
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2020217285443219e-05,
+      "loss": 0.1945,
+      "step": 23761
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2019881450136854e-05,
+      "loss": 0.5014,
+      "step": 23762
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2019545614830488e-05,
+      "loss": 0.3019,
+      "step": 23763
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2019209779524121e-05,
+      "loss": 0.2247,
+      "step": 23764
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2018873944217758e-05,
+      "loss": 0.3581,
+      "step": 23765
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.201853810891139e-05,
+      "loss": 0.3161,
+      "step": 23766
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2018202273605023e-05,
+      "loss": 0.468,
+      "step": 23767
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.201786643829866e-05,
+      "loss": 0.2371,
+      "step": 23768
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2017530602992293e-05,
+      "loss": 0.2203,
+      "step": 23769
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2017194767685927e-05,
+      "loss": 0.3195,
+      "step": 23770
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2016858932379562e-05,
+      "loss": 0.2699,
+      "step": 23771
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2016523097073197e-05,
+      "loss": 0.1968,
+      "step": 23772
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.201618726176683e-05,
+      "loss": 0.2577,
+      "step": 23773
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2015851426460466e-05,
+      "loss": 0.435,
+      "step": 23774
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2015515591154099e-05,
+      "loss": 0.3126,
+      "step": 23775
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2015179755847732e-05,
+      "loss": 0.2693,
+      "step": 23776
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2014843920541368e-05,
+      "loss": 0.1991,
+      "step": 23777
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2014508085235001e-05,
+      "loss": 0.2785,
+      "step": 23778
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2014172249928636e-05,
+      "loss": 0.246,
+      "step": 23779
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.201383641462227e-05,
+      "loss": 0.3193,
+      "step": 23780
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2013500579315905e-05,
+      "loss": 0.4452,
+      "step": 23781
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2013164744009538e-05,
+      "loss": 0.4224,
+      "step": 23782
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2012828908703173e-05,
+      "loss": 0.2725,
+      "step": 23783
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2012493073396807e-05,
+      "loss": 0.258,
+      "step": 23784
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.201215723809044e-05,
+      "loss": 0.2537,
+      "step": 23785
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2011821402784077e-05,
+      "loss": 0.2956,
+      "step": 23786
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.201148556747771e-05,
+      "loss": 0.5347,
+      "step": 23787
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2011149732171343e-05,
+      "loss": 0.4813,
+      "step": 23788
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2010813896864979e-05,
+      "loss": 0.5528,
+      "step": 23789
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2010478061558612e-05,
+      "loss": 0.3079,
+      "step": 23790
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2010142226252246e-05,
+      "loss": 0.172,
+      "step": 23791
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2009806390945881e-05,
+      "loss": 0.3429,
+      "step": 23792
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2009470555639516e-05,
+      "loss": 0.21,
+      "step": 23793
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2009134720333149e-05,
+      "loss": 0.2171,
+      "step": 23794
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2008798885026785e-05,
+      "loss": 0.2389,
+      "step": 23795
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2008463049720418e-05,
+      "loss": 0.142,
+      "step": 23796
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2008127214414051e-05,
+      "loss": 0.5143,
+      "step": 23797
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2007791379107687e-05,
+      "loss": 0.4366,
+      "step": 23798
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.200745554380132e-05,
+      "loss": 0.2191,
+      "step": 23799
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2007119708494955e-05,
+      "loss": 0.2349,
+      "step": 23800
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.200678387318859e-05,
+      "loss": 0.5375,
+      "step": 23801
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2006448037882224e-05,
+      "loss": 0.3343,
+      "step": 23802
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2006112202575857e-05,
+      "loss": 0.2078,
+      "step": 23803
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2005776367269494e-05,
+      "loss": 0.3438,
+      "step": 23804
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2005440531963126e-05,
+      "loss": 0.1259,
+      "step": 23805
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.200510469665676e-05,
+      "loss": 0.3803,
+      "step": 23806
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2004768861350396e-05,
+      "loss": 0.2218,
+      "step": 23807
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2004433026044029e-05,
+      "loss": 0.3171,
+      "step": 23808
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2004097190737662e-05,
+      "loss": 0.467,
+      "step": 23809
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2003761355431298e-05,
+      "loss": 0.2267,
+      "step": 23810
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2003425520124931e-05,
+      "loss": 0.2637,
+      "step": 23811
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2003089684818566e-05,
+      "loss": 0.3877,
+      "step": 23812
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.20027538495122e-05,
+      "loss": 0.3668,
+      "step": 23813
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2002418014205835e-05,
+      "loss": 0.3244,
+      "step": 23814
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2002082178899468e-05,
+      "loss": 0.2364,
+      "step": 23815
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2001746343593104e-05,
+      "loss": 0.1913,
+      "step": 23816
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2001410508286737e-05,
+      "loss": 0.1527,
+      "step": 23817
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.200107467298037e-05,
+      "loss": 0.3657,
+      "step": 23818
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2000738837674006e-05,
+      "loss": 0.2695,
+      "step": 23819
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.200040300236764e-05,
+      "loss": 0.3196,
+      "step": 23820
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.2000067167061274e-05,
+      "loss": 0.3078,
+      "step": 23821
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1999731331754909e-05,
+      "loss": 0.4572,
+      "step": 23822
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1999395496448543e-05,
+      "loss": 0.3102,
+      "step": 23823
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1999059661142176e-05,
+      "loss": 0.3508,
+      "step": 23824
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1998723825835813e-05,
+      "loss": 0.2834,
+      "step": 23825
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1998387990529446e-05,
+      "loss": 0.4209,
+      "step": 23826
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1998052155223079e-05,
+      "loss": 0.2407,
+      "step": 23827
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1997716319916715e-05,
+      "loss": 0.2749,
+      "step": 23828
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1997380484610348e-05,
+      "loss": 0.2636,
+      "step": 23829
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.199704464930398e-05,
+      "loss": 0.1609,
+      "step": 23830
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1996708813997617e-05,
+      "loss": 0.2734,
+      "step": 23831
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.199637297869125e-05,
+      "loss": 0.19,
+      "step": 23832
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1996037143384885e-05,
+      "loss": 0.2435,
+      "step": 23833
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.199570130807852e-05,
+      "loss": 0.4347,
+      "step": 23834
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1995365472772154e-05,
+      "loss": 0.2613,
+      "step": 23835
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1995029637465787e-05,
+      "loss": 0.252,
+      "step": 23836
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1994693802159423e-05,
+      "loss": 0.3049,
+      "step": 23837
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1994357966853056e-05,
+      "loss": 0.3019,
+      "step": 23838
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.199402213154669e-05,
+      "loss": 0.4133,
+      "step": 23839
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1993686296240326e-05,
+      "loss": 0.226,
+      "step": 23840
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1993350460933959e-05,
+      "loss": 0.3724,
+      "step": 23841
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1993014625627593e-05,
+      "loss": 0.5011,
+      "step": 23842
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1992678790321228e-05,
+      "loss": 0.2085,
+      "step": 23843
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1992342955014862e-05,
+      "loss": 0.4622,
+      "step": 23844
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1992007119708495e-05,
+      "loss": 0.0979,
+      "step": 23845
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1991671284402132e-05,
+      "loss": 0.2161,
+      "step": 23846
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1991335449095765e-05,
+      "loss": 0.2593,
+      "step": 23847
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1990999613789398e-05,
+      "loss": 0.3247,
+      "step": 23848
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1990663778483034e-05,
+      "loss": 0.5082,
+      "step": 23849
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1990327943176667e-05,
+      "loss": 0.1871,
+      "step": 23850
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.19899921078703e-05,
+      "loss": 0.5357,
+      "step": 23851
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1989656272563936e-05,
+      "loss": 0.3544,
+      "step": 23852
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.198932043725757e-05,
+      "loss": 0.2893,
+      "step": 23853
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1988984601951204e-05,
+      "loss": 0.3187,
+      "step": 23854
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1988648766644839e-05,
+      "loss": 0.2674,
+      "step": 23855
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1988312931338473e-05,
+      "loss": 0.2804,
+      "step": 23856
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1987977096032106e-05,
+      "loss": 0.2146,
+      "step": 23857
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1987641260725742e-05,
+      "loss": 0.3896,
+      "step": 23858
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1987305425419375e-05,
+      "loss": 0.2088,
+      "step": 23859
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1986969590113008e-05,
+      "loss": 0.3559,
+      "step": 23860
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1986633754806645e-05,
+      "loss": 0.2009,
+      "step": 23861
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1986297919500278e-05,
+      "loss": 0.2262,
+      "step": 23862
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1985962084193912e-05,
+      "loss": 0.292,
+      "step": 23863
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1985626248887547e-05,
+      "loss": 0.431,
+      "step": 23864
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1985290413581182e-05,
+      "loss": 0.2061,
+      "step": 23865
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1984954578274815e-05,
+      "loss": 0.2405,
+      "step": 23866
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1984618742968451e-05,
+      "loss": 0.1909,
+      "step": 23867
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1984282907662084e-05,
+      "loss": 0.5505,
+      "step": 23868
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1983947072355717e-05,
+      "loss": 0.2919,
+      "step": 23869
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1983611237049353e-05,
+      "loss": 0.3214,
+      "step": 23870
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1983275401742986e-05,
+      "loss": 0.3125,
+      "step": 23871
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.198293956643662e-05,
+      "loss": 0.1435,
+      "step": 23872
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1982603731130255e-05,
+      "loss": 0.2939,
+      "step": 23873
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1982267895823888e-05,
+      "loss": 0.3104,
+      "step": 23874
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1981932060517523e-05,
+      "loss": 0.2721,
+      "step": 23875
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1981596225211158e-05,
+      "loss": 0.3232,
+      "step": 23876
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1981260389904792e-05,
+      "loss": 0.292,
+      "step": 23877
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1980924554598425e-05,
+      "loss": 0.3289,
+      "step": 23878
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1980588719292062e-05,
+      "loss": 0.4445,
+      "step": 23879
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1980252883985695e-05,
+      "loss": 0.4008,
+      "step": 23880
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1979917048679327e-05,
+      "loss": 0.2047,
+      "step": 23881
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1979581213372964e-05,
+      "loss": 0.2583,
+      "step": 23882
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1979245378066597e-05,
+      "loss": 0.2438,
+      "step": 23883
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1978909542760231e-05,
+      "loss": 0.2649,
+      "step": 23884
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1978573707453866e-05,
+      "loss": 0.3419,
+      "step": 23885
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.19782378721475e-05,
+      "loss": 0.1686,
+      "step": 23886
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1977902036841134e-05,
+      "loss": 0.2086,
+      "step": 23887
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.197756620153477e-05,
+      "loss": 0.3203,
+      "step": 23888
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1977230366228403e-05,
+      "loss": 0.1906,
+      "step": 23889
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1976894530922036e-05,
+      "loss": 0.1882,
+      "step": 23890
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1976558695615672e-05,
+      "loss": 0.2071,
+      "step": 23891
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1976222860309305e-05,
+      "loss": 0.4237,
+      "step": 23892
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.197588702500294e-05,
+      "loss": 0.446,
+      "step": 23893
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1975551189696575e-05,
+      "loss": 0.2962,
+      "step": 23894
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.197521535439021e-05,
+      "loss": 0.272,
+      "step": 23895
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1974879519083842e-05,
+      "loss": 0.3417,
+      "step": 23896
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1974543683777477e-05,
+      "loss": 0.2532,
+      "step": 23897
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1974207848471111e-05,
+      "loss": 0.2447,
+      "step": 23898
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1973872013164744e-05,
+      "loss": 0.3648,
+      "step": 23899
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.197353617785838e-05,
+      "loss": 0.2951,
+      "step": 23900
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1973200342552014e-05,
+      "loss": 0.3975,
+      "step": 23901
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1972864507245647e-05,
+      "loss": 0.254,
+      "step": 23902
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1972528671939283e-05,
+      "loss": 0.2337,
+      "step": 23903
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1972192836632916e-05,
+      "loss": 0.3149,
+      "step": 23904
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.197185700132655e-05,
+      "loss": 0.2724,
+      "step": 23905
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1971521166020185e-05,
+      "loss": 0.3106,
+      "step": 23906
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.197118533071382e-05,
+      "loss": 0.2099,
+      "step": 23907
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1970849495407453e-05,
+      "loss": 0.3009,
+      "step": 23908
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1970513660101089e-05,
+      "loss": 0.3554,
+      "step": 23909
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1970177824794722e-05,
+      "loss": 0.3242,
+      "step": 23910
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1969841989488355e-05,
+      "loss": 0.2796,
+      "step": 23911
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.196950615418199e-05,
+      "loss": 0.1811,
+      "step": 23912
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1969170318875624e-05,
+      "loss": 0.315,
+      "step": 23913
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1968834483569259e-05,
+      "loss": 0.2438,
+      "step": 23914
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1968498648262892e-05,
+      "loss": 0.3571,
+      "step": 23915
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1968162812956528e-05,
+      "loss": 0.2597,
+      "step": 23916
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1967826977650161e-05,
+      "loss": 0.3977,
+      "step": 23917
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1967491142343794e-05,
+      "loss": 0.3608,
+      "step": 23918
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.196715530703743e-05,
+      "loss": 0.1392,
+      "step": 23919
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.1966819471731063e-05,
+      "loss": 0.2018,
+      "step": 23920
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1966483636424696e-05,
+      "loss": 0.1855,
+      "step": 23921
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1966147801118333e-05,
+      "loss": 0.3418,
+      "step": 23922
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1965811965811966e-05,
+      "loss": 0.514,
+      "step": 23923
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.19654761305056e-05,
+      "loss": 0.2827,
+      "step": 23924
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1965140295199235e-05,
+      "loss": 0.2808,
+      "step": 23925
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.196480445989287e-05,
+      "loss": 0.3089,
+      "step": 23926
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1964468624586503e-05,
+      "loss": 0.3449,
+      "step": 23927
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1964132789280139e-05,
+      "loss": 0.3011,
+      "step": 23928
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1963796953973772e-05,
+      "loss": 0.206,
+      "step": 23929
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1963461118667405e-05,
+      "loss": 0.2002,
+      "step": 23930
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1963125283361041e-05,
+      "loss": 0.2287,
+      "step": 23931
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1962789448054674e-05,
+      "loss": 0.461,
+      "step": 23932
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1962453612748309e-05,
+      "loss": 0.3491,
+      "step": 23933
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1962117777441943e-05,
+      "loss": 0.1875,
+      "step": 23934
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1961781942135578e-05,
+      "loss": 0.272,
+      "step": 23935
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1961446106829211e-05,
+      "loss": 0.2099,
+      "step": 23936
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1961110271522847e-05,
+      "loss": 0.3598,
+      "step": 23937
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.196077443621648e-05,
+      "loss": 0.2836,
+      "step": 23938
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1960438600910113e-05,
+      "loss": 0.2694,
+      "step": 23939
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.196010276560375e-05,
+      "loss": 0.2702,
+      "step": 23940
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1959766930297383e-05,
+      "loss": 0.4049,
+      "step": 23941
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1959431094991016e-05,
+      "loss": 0.3915,
+      "step": 23942
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1959095259684652e-05,
+      "loss": 0.1924,
+      "step": 23943
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1958759424378285e-05,
+      "loss": 0.4216,
+      "step": 23944
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.195842358907192e-05,
+      "loss": 0.3825,
+      "step": 23945
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1958087753765554e-05,
+      "loss": 0.3925,
+      "step": 23946
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1957751918459189e-05,
+      "loss": 0.2861,
+      "step": 23947
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1957416083152822e-05,
+      "loss": 0.1835,
+      "step": 23948
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1957080247846458e-05,
+      "loss": 0.1601,
+      "step": 23949
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1956744412540091e-05,
+      "loss": 0.1376,
+      "step": 23950
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1956408577233724e-05,
+      "loss": 0.2089,
+      "step": 23951
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.195607274192736e-05,
+      "loss": 0.4159,
+      "step": 23952
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1955736906620993e-05,
+      "loss": 0.4199,
+      "step": 23953
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1955401071314628e-05,
+      "loss": 0.4057,
+      "step": 23954
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1955065236008263e-05,
+      "loss": 0.1742,
+      "step": 23955
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1954729400701897e-05,
+      "loss": 0.2539,
+      "step": 23956
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.195439356539553e-05,
+      "loss": 0.2531,
+      "step": 23957
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1954057730089167e-05,
+      "loss": 0.229,
+      "step": 23958
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.19537218947828e-05,
+      "loss": 0.3095,
+      "step": 23959
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1953386059476432e-05,
+      "loss": 0.2635,
+      "step": 23960
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1953050224170069e-05,
+      "loss": 0.2536,
+      "step": 23961
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1952714388863702e-05,
+      "loss": 0.3452,
+      "step": 23962
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1952378553557336e-05,
+      "loss": 0.295,
+      "step": 23963
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1952042718250971e-05,
+      "loss": 0.2591,
+      "step": 23964
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1951706882944604e-05,
+      "loss": 0.3242,
+      "step": 23965
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1951371047638239e-05,
+      "loss": 0.3474,
+      "step": 23966
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1951035212331873e-05,
+      "loss": 0.2512,
+      "step": 23967
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1950699377025508e-05,
+      "loss": 0.4889,
+      "step": 23968
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1950363541719141e-05,
+      "loss": 0.1946,
+      "step": 23969
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1950027706412777e-05,
+      "loss": 0.32,
+      "step": 23970
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.194969187110641e-05,
+      "loss": 0.2459,
+      "step": 23971
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1949356035800043e-05,
+      "loss": 0.3439,
+      "step": 23972
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.194902020049368e-05,
+      "loss": 0.2506,
+      "step": 23973
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1948684365187312e-05,
+      "loss": 0.1687,
+      "step": 23974
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1948348529880947e-05,
+      "loss": 0.228,
+      "step": 23975
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1948012694574582e-05,
+      "loss": 0.3462,
+      "step": 23976
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1947676859268216e-05,
+      "loss": 0.2619,
+      "step": 23977
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.194734102396185e-05,
+      "loss": 0.2815,
+      "step": 23978
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1947005188655486e-05,
+      "loss": 0.302,
+      "step": 23979
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1946669353349119e-05,
+      "loss": 0.2267,
+      "step": 23980
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1946333518042752e-05,
+      "loss": 0.1965,
+      "step": 23981
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1945997682736388e-05,
+      "loss": 0.2671,
+      "step": 23982
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1945661847430021e-05,
+      "loss": 0.2568,
+      "step": 23983
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1945326012123656e-05,
+      "loss": 0.2953,
+      "step": 23984
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.194499017681729e-05,
+      "loss": 0.2325,
+      "step": 23985
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1944654341510925e-05,
+      "loss": 0.253,
+      "step": 23986
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1944318506204558e-05,
+      "loss": 0.1871,
+      "step": 23987
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1943982670898192e-05,
+      "loss": 0.2982,
+      "step": 23988
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1943646835591827e-05,
+      "loss": 0.3422,
+      "step": 23989
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.194331100028546e-05,
+      "loss": 0.4461,
+      "step": 23990
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1942975164979096e-05,
+      "loss": 0.5055,
+      "step": 23991
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.194263932967273e-05,
+      "loss": 0.2764,
+      "step": 23992
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1942303494366362e-05,
+      "loss": 0.2474,
+      "step": 23993
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1941967659059999e-05,
+      "loss": 0.2559,
+      "step": 23994
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1941631823753632e-05,
+      "loss": 0.4144,
+      "step": 23995
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1941295988447266e-05,
+      "loss": 0.4406,
+      "step": 23996
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.19409601531409e-05,
+      "loss": 0.3586,
+      "step": 23997
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1940624317834535e-05,
+      "loss": 0.4511,
+      "step": 23998
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1940288482528168e-05,
+      "loss": 0.2825,
+      "step": 23999
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1939952647221805e-05,
+      "loss": 0.2969,
+      "step": 24000
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1939616811915438e-05,
+      "loss": 0.3539,
+      "step": 24001
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.193928097660907e-05,
+      "loss": 0.1888,
+      "step": 24002
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1938945141302707e-05,
+      "loss": 0.2701,
+      "step": 24003
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.193860930599634e-05,
+      "loss": 0.5343,
+      "step": 24004
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1938273470689975e-05,
+      "loss": 0.3344,
+      "step": 24005
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.193793763538361e-05,
+      "loss": 0.3689,
+      "step": 24006
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1937601800077244e-05,
+      "loss": 0.3246,
+      "step": 24007
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1937265964770877e-05,
+      "loss": 0.3156,
+      "step": 24008
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1936930129464512e-05,
+      "loss": 0.2207,
+      "step": 24009
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1936594294158146e-05,
+      "loss": 0.2554,
+      "step": 24010
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1936258458851779e-05,
+      "loss": 0.2902,
+      "step": 24011
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1935922623545415e-05,
+      "loss": 0.3716,
+      "step": 24012
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1935586788239048e-05,
+      "loss": 0.5144,
+      "step": 24013
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1935250952932681e-05,
+      "loss": 0.2969,
+      "step": 24014
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1934915117626318e-05,
+      "loss": 0.3613,
+      "step": 24015
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.193457928231995e-05,
+      "loss": 0.3385,
+      "step": 24016
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1934243447013585e-05,
+      "loss": 0.1441,
+      "step": 24017
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.193390761170722e-05,
+      "loss": 0.3653,
+      "step": 24018
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1933571776400855e-05,
+      "loss": 0.1574,
+      "step": 24019
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1933235941094488e-05,
+      "loss": 0.3085,
+      "step": 24020
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1932900105788124e-05,
+      "loss": 0.2508,
+      "step": 24021
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1932564270481757e-05,
+      "loss": 0.2653,
+      "step": 24022
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.193222843517539e-05,
+      "loss": 0.1889,
+      "step": 24023
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1931892599869026e-05,
+      "loss": 0.3385,
+      "step": 24024
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1931556764562659e-05,
+      "loss": 0.4622,
+      "step": 24025
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1931220929256294e-05,
+      "loss": 0.3953,
+      "step": 24026
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1930885093949928e-05,
+      "loss": 0.272,
+      "step": 24027
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1930549258643563e-05,
+      "loss": 0.2744,
+      "step": 24028
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1930213423337196e-05,
+      "loss": 0.198,
+      "step": 24029
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1929877588030832e-05,
+      "loss": 0.2085,
+      "step": 24030
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1929541752724465e-05,
+      "loss": 0.2837,
+      "step": 24031
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1929205917418098e-05,
+      "loss": 0.3008,
+      "step": 24032
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1928870082111735e-05,
+      "loss": 0.2203,
+      "step": 24033
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1928534246805368e-05,
+      "loss": 0.219,
+      "step": 24034
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1928198411499e-05,
+      "loss": 0.4435,
+      "step": 24035
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1927862576192637e-05,
+      "loss": 0.3687,
+      "step": 24036
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.192752674088627e-05,
+      "loss": 0.4339,
+      "step": 24037
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1927190905579904e-05,
+      "loss": 0.2815,
+      "step": 24038
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1926855070273539e-05,
+      "loss": 0.4997,
+      "step": 24039
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1926519234967174e-05,
+      "loss": 0.3184,
+      "step": 24040
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1926183399660807e-05,
+      "loss": 0.131,
+      "step": 24041
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1925847564354443e-05,
+      "loss": 0.3288,
+      "step": 24042
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1925511729048076e-05,
+      "loss": 0.4866,
+      "step": 24043
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1925175893741709e-05,
+      "loss": 0.259,
+      "step": 24044
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1924840058435345e-05,
+      "loss": 0.2863,
+      "step": 24045
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1924504223128978e-05,
+      "loss": 0.2532,
+      "step": 24046
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1924168387822613e-05,
+      "loss": 0.2322,
+      "step": 24047
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1923832552516248e-05,
+      "loss": 0.1992,
+      "step": 24048
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1923496717209882e-05,
+      "loss": 0.2696,
+      "step": 24049
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1923160881903515e-05,
+      "loss": 0.1919,
+      "step": 24050
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1922825046597151e-05,
+      "loss": 0.1325,
+      "step": 24051
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1922489211290784e-05,
+      "loss": 0.1532,
+      "step": 24052
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1922153375984417e-05,
+      "loss": 0.247,
+      "step": 24053
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1921817540678054e-05,
+      "loss": 0.359,
+      "step": 24054
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1921481705371687e-05,
+      "loss": 0.401,
+      "step": 24055
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.192114587006532e-05,
+      "loss": 0.2637,
+      "step": 24056
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1920810034758956e-05,
+      "loss": 0.3566,
+      "step": 24057
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1920474199452589e-05,
+      "loss": 0.3816,
+      "step": 24058
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1920138364146224e-05,
+      "loss": 0.2382,
+      "step": 24059
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1919802528839858e-05,
+      "loss": 0.1999,
+      "step": 24060
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1919466693533493e-05,
+      "loss": 0.3083,
+      "step": 24061
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1919130858227126e-05,
+      "loss": 0.4529,
+      "step": 24062
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1918795022920762e-05,
+      "loss": 0.1722,
+      "step": 24063
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1918459187614395e-05,
+      "loss": 0.2369,
+      "step": 24064
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1918123352308028e-05,
+      "loss": 0.2529,
+      "step": 24065
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1917787517001664e-05,
+      "loss": 0.2614,
+      "step": 24066
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1917451681695297e-05,
+      "loss": 0.3276,
+      "step": 24067
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1917115846388932e-05,
+      "loss": 0.1572,
+      "step": 24068
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1916780011082567e-05,
+      "loss": 0.235,
+      "step": 24069
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1916444175776201e-05,
+      "loss": 0.1943,
+      "step": 24070
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1916108340469834e-05,
+      "loss": 0.393,
+      "step": 24071
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.191577250516347e-05,
+      "loss": 0.2689,
+      "step": 24072
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1915436669857104e-05,
+      "loss": 0.1785,
+      "step": 24073
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1915100834550737e-05,
+      "loss": 0.2887,
+      "step": 24074
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1914764999244373e-05,
+      "loss": 0.3275,
+      "step": 24075
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1914429163938006e-05,
+      "loss": 0.402,
+      "step": 24076
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1914093328631639e-05,
+      "loss": 0.3793,
+      "step": 24077
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1913757493325275e-05,
+      "loss": 0.3248,
+      "step": 24078
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1913421658018908e-05,
+      "loss": 0.1715,
+      "step": 24079
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1913085822712543e-05,
+      "loss": 0.3077,
+      "step": 24080
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1912749987406177e-05,
+      "loss": 0.2919,
+      "step": 24081
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1912414152099812e-05,
+      "loss": 0.3937,
+      "step": 24082
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1912078316793445e-05,
+      "loss": 0.1762,
+      "step": 24083
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1911742481487081e-05,
+      "loss": 0.3204,
+      "step": 24084
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1911406646180714e-05,
+      "loss": 0.2419,
+      "step": 24085
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1911070810874347e-05,
+      "loss": 0.2843,
+      "step": 24086
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1910734975567984e-05,
+      "loss": 0.3202,
+      "step": 24087
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1910399140261616e-05,
+      "loss": 0.323,
+      "step": 24088
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1910063304955251e-05,
+      "loss": 0.4747,
+      "step": 24089
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1909727469648886e-05,
+      "loss": 0.3326,
+      "step": 24090
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.190939163434252e-05,
+      "loss": 0.325,
+      "step": 24091
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1909055799036153e-05,
+      "loss": 0.2988,
+      "step": 24092
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1908719963729786e-05,
+      "loss": 0.462,
+      "step": 24093
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1908384128423423e-05,
+      "loss": 0.1843,
+      "step": 24094
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1908048293117056e-05,
+      "loss": 0.1802,
+      "step": 24095
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.190771245781069e-05,
+      "loss": 0.398,
+      "step": 24096
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1907376622504325e-05,
+      "loss": 0.3063,
+      "step": 24097
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.190704078719796e-05,
+      "loss": 0.2742,
+      "step": 24098
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1906704951891593e-05,
+      "loss": 0.2482,
+      "step": 24099
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1906369116585227e-05,
+      "loss": 0.2715,
+      "step": 24100
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1906033281278862e-05,
+      "loss": 0.1328,
+      "step": 24101
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1905697445972495e-05,
+      "loss": 0.3133,
+      "step": 24102
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1905361610666131e-05,
+      "loss": 0.2861,
+      "step": 24103
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1905025775359764e-05,
+      "loss": 0.2669,
+      "step": 24104
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1904689940053397e-05,
+      "loss": 0.3628,
+      "step": 24105
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1904354104747033e-05,
+      "loss": 0.203,
+      "step": 24106
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1904018269440666e-05,
+      "loss": 0.3542,
+      "step": 24107
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1903682434134301e-05,
+      "loss": 0.154,
+      "step": 24108
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1903346598827936e-05,
+      "loss": 0.3048,
+      "step": 24109
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.190301076352157e-05,
+      "loss": 0.3136,
+      "step": 24110
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1902674928215203e-05,
+      "loss": 0.2328,
+      "step": 24111
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.190233909290884e-05,
+      "loss": 0.2648,
+      "step": 24112
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1902003257602473e-05,
+      "loss": 0.1729,
+      "step": 24113
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1901667422296105e-05,
+      "loss": 0.1862,
+      "step": 24114
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1901331586989742e-05,
+      "loss": 0.2426,
+      "step": 24115
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1900995751683375e-05,
+      "loss": 0.284,
+      "step": 24116
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.190065991637701e-05,
+      "loss": 0.4652,
+      "step": 24117
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 1.1900324081070644e-05,
+      "loss": 0.3231,
+      "step": 24118
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1899988245764279e-05,
+      "loss": 0.3127,
+      "step": 24119
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1899652410457912e-05,
+      "loss": 0.4316,
+      "step": 24120
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1899316575151548e-05,
+      "loss": 0.2876,
+      "step": 24121
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1898980739845181e-05,
+      "loss": 0.2594,
+      "step": 24122
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1898644904538814e-05,
+      "loss": 0.1442,
+      "step": 24123
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.189830906923245e-05,
+      "loss": 0.1791,
+      "step": 24124
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1897973233926083e-05,
+      "loss": 0.2425,
+      "step": 24125
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1897637398619716e-05,
+      "loss": 0.3331,
+      "step": 24126
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1897301563313352e-05,
+      "loss": 0.3624,
+      "step": 24127
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1896965728006985e-05,
+      "loss": 0.4162,
+      "step": 24128
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.189662989270062e-05,
+      "loss": 0.3385,
+      "step": 24129
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1896294057394255e-05,
+      "loss": 0.3376,
+      "step": 24130
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.189595822208789e-05,
+      "loss": 0.2645,
+      "step": 24131
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1895622386781522e-05,
+      "loss": 0.3268,
+      "step": 24132
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1895286551475159e-05,
+      "loss": 0.2441,
+      "step": 24133
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1894950716168792e-05,
+      "loss": 0.2709,
+      "step": 24134
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1894614880862425e-05,
+      "loss": 0.4781,
+      "step": 24135
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1894279045556061e-05,
+      "loss": 0.2765,
+      "step": 24136
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1893943210249694e-05,
+      "loss": 0.3954,
+      "step": 24137
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1893607374943329e-05,
+      "loss": 0.353,
+      "step": 24138
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1893271539636963e-05,
+      "loss": 0.274,
+      "step": 24139
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1892935704330598e-05,
+      "loss": 0.3736,
+      "step": 24140
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.189259986902423e-05,
+      "loss": 0.2021,
+      "step": 24141
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1892264033717867e-05,
+      "loss": 0.1847,
+      "step": 24142
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.18919281984115e-05,
+      "loss": 0.2851,
+      "step": 24143
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1891592363105133e-05,
+      "loss": 0.3329,
+      "step": 24144
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.189125652779877e-05,
+      "loss": 0.2312,
+      "step": 24145
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1890920692492402e-05,
+      "loss": 0.3823,
+      "step": 24146
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1890584857186035e-05,
+      "loss": 0.334,
+      "step": 24147
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1890249021879672e-05,
+      "loss": 0.3073,
+      "step": 24148
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1889913186573305e-05,
+      "loss": 0.4236,
+      "step": 24149
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.188957735126694e-05,
+      "loss": 0.3061,
+      "step": 24150
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1889241515960574e-05,
+      "loss": 0.2564,
+      "step": 24151
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1888905680654209e-05,
+      "loss": 0.2951,
+      "step": 24152
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1888569845347841e-05,
+      "loss": 0.287,
+      "step": 24153
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1888234010041478e-05,
+      "loss": 0.3314,
+      "step": 24154
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.188789817473511e-05,
+      "loss": 0.3039,
+      "step": 24155
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1887562339428744e-05,
+      "loss": 0.2926,
+      "step": 24156
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.188722650412238e-05,
+      "loss": 0.2508,
+      "step": 24157
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1886890668816013e-05,
+      "loss": 0.3211,
+      "step": 24158
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1886554833509648e-05,
+      "loss": 0.3385,
+      "step": 24159
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1886218998203282e-05,
+      "loss": 0.4217,
+      "step": 24160
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1885883162896917e-05,
+      "loss": 0.2701,
+      "step": 24161
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.188554732759055e-05,
+      "loss": 0.2717,
+      "step": 24162
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1885211492284186e-05,
+      "loss": 0.3418,
+      "step": 24163
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.188487565697782e-05,
+      "loss": 0.2777,
+      "step": 24164
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1884539821671452e-05,
+      "loss": 0.309,
+      "step": 24165
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1884203986365088e-05,
+      "loss": 0.2634,
+      "step": 24166
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1883868151058721e-05,
+      "loss": 0.2725,
+      "step": 24167
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1883532315752354e-05,
+      "loss": 0.2625,
+      "step": 24168
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.188319648044599e-05,
+      "loss": 0.2762,
+      "step": 24169
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1882860645139624e-05,
+      "loss": 0.1514,
+      "step": 24170
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1882524809833258e-05,
+      "loss": 0.3391,
+      "step": 24171
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1882188974526893e-05,
+      "loss": 0.3235,
+      "step": 24172
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1881853139220528e-05,
+      "loss": 0.2511,
+      "step": 24173
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.188151730391416e-05,
+      "loss": 0.4334,
+      "step": 24174
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1881181468607797e-05,
+      "loss": 0.34,
+      "step": 24175
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.188084563330143e-05,
+      "loss": 0.2897,
+      "step": 24176
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1880509797995063e-05,
+      "loss": 0.2706,
+      "step": 24177
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.18801739626887e-05,
+      "loss": 0.2881,
+      "step": 24178
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1879838127382332e-05,
+      "loss": 0.3212,
+      "step": 24179
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1879502292075967e-05,
+      "loss": 0.2385,
+      "step": 24180
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1879166456769601e-05,
+      "loss": 0.3655,
+      "step": 24181
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1878830621463236e-05,
+      "loss": 0.4671,
+      "step": 24182
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1878494786156869e-05,
+      "loss": 0.3204,
+      "step": 24183
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1878158950850505e-05,
+      "loss": 0.4562,
+      "step": 24184
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1877823115544138e-05,
+      "loss": 0.2587,
+      "step": 24185
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1877487280237771e-05,
+      "loss": 0.2937,
+      "step": 24186
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1877151444931408e-05,
+      "loss": 0.265,
+      "step": 24187
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.187681560962504e-05,
+      "loss": 0.2785,
+      "step": 24188
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1876479774318675e-05,
+      "loss": 0.3125,
+      "step": 24189
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.187614393901231e-05,
+      "loss": 0.1396,
+      "step": 24190
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1875808103705943e-05,
+      "loss": 0.2778,
+      "step": 24191
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1875472268399577e-05,
+      "loss": 0.3188,
+      "step": 24192
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1875136433093212e-05,
+      "loss": 0.3779,
+      "step": 24193
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1874800597786847e-05,
+      "loss": 0.186,
+      "step": 24194
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.187446476248048e-05,
+      "loss": 0.3406,
+      "step": 24195
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1874128927174116e-05,
+      "loss": 0.2519,
+      "step": 24196
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1873793091867749e-05,
+      "loss": 0.2261,
+      "step": 24197
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1873457256561382e-05,
+      "loss": 0.2841,
+      "step": 24198
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1873121421255018e-05,
+      "loss": 0.3584,
+      "step": 24199
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1872785585948651e-05,
+      "loss": 0.2428,
+      "step": 24200
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1872449750642286e-05,
+      "loss": 0.3116,
+      "step": 24201
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.187211391533592e-05,
+      "loss": 0.3481,
+      "step": 24202
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1871778080029555e-05,
+      "loss": 0.2612,
+      "step": 24203
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1871442244723188e-05,
+      "loss": 0.4088,
+      "step": 24204
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1871106409416824e-05,
+      "loss": 0.2139,
+      "step": 24205
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1870770574110457e-05,
+      "loss": 0.3653,
+      "step": 24206
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.187043473880409e-05,
+      "loss": 0.1998,
+      "step": 24207
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1870098903497727e-05,
+      "loss": 0.1889,
+      "step": 24208
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.186976306819136e-05,
+      "loss": 0.3751,
+      "step": 24209
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1869427232884994e-05,
+      "loss": 0.3456,
+      "step": 24210
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1869091397578629e-05,
+      "loss": 0.2676,
+      "step": 24211
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1868755562272262e-05,
+      "loss": 0.2118,
+      "step": 24212
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1868419726965897e-05,
+      "loss": 0.1245,
+      "step": 24213
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1868083891659531e-05,
+      "loss": 0.3159,
+      "step": 24214
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1867748056353166e-05,
+      "loss": 0.2202,
+      "step": 24215
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1867412221046799e-05,
+      "loss": 0.1353,
+      "step": 24216
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1867076385740435e-05,
+      "loss": 0.3634,
+      "step": 24217
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1866740550434068e-05,
+      "loss": 0.3677,
+      "step": 24218
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1866404715127701e-05,
+      "loss": 0.2594,
+      "step": 24219
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1866068879821337e-05,
+      "loss": 0.3906,
+      "step": 24220
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.186573304451497e-05,
+      "loss": 0.1609,
+      "step": 24221
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1865397209208605e-05,
+      "loss": 0.2089,
+      "step": 24222
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.186506137390224e-05,
+      "loss": 0.2371,
+      "step": 24223
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1864725538595874e-05,
+      "loss": 0.3369,
+      "step": 24224
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1864389703289507e-05,
+      "loss": 0.2915,
+      "step": 24225
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1864053867983144e-05,
+      "loss": 0.391,
+      "step": 24226
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1863718032676777e-05,
+      "loss": 0.2844,
+      "step": 24227
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.186338219737041e-05,
+      "loss": 0.2644,
+      "step": 24228
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1863046362064046e-05,
+      "loss": 0.225,
+      "step": 24229
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1862710526757679e-05,
+      "loss": 0.3463,
+      "step": 24230
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1862374691451313e-05,
+      "loss": 0.3053,
+      "step": 24231
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1862038856144948e-05,
+      "loss": 0.2967,
+      "step": 24232
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1861703020838583e-05,
+      "loss": 0.3418,
+      "step": 24233
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1861367185532216e-05,
+      "loss": 0.276,
+      "step": 24234
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.186103135022585e-05,
+      "loss": 0.1879,
+      "step": 24235
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1860695514919485e-05,
+      "loss": 0.1272,
+      "step": 24236
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1860359679613118e-05,
+      "loss": 0.291,
+      "step": 24237
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1860023844306754e-05,
+      "loss": 0.3444,
+      "step": 24238
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1859688009000387e-05,
+      "loss": 0.3088,
+      "step": 24239
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.185935217369402e-05,
+      "loss": 0.4051,
+      "step": 24240
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1859016338387657e-05,
+      "loss": 0.3566,
+      "step": 24241
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.185868050308129e-05,
+      "loss": 0.4319,
+      "step": 24242
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1858344667774924e-05,
+      "loss": 0.2766,
+      "step": 24243
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1858008832468559e-05,
+      "loss": 0.2943,
+      "step": 24244
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1857672997162193e-05,
+      "loss": 0.325,
+      "step": 24245
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1857337161855826e-05,
+      "loss": 0.3027,
+      "step": 24246
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1857001326549463e-05,
+      "loss": 0.1919,
+      "step": 24247
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1856665491243096e-05,
+      "loss": 0.3233,
+      "step": 24248
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1856329655936729e-05,
+      "loss": 0.244,
+      "step": 24249
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1855993820630365e-05,
+      "loss": 0.2413,
+      "step": 24250
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1855657985323998e-05,
+      "loss": 0.3175,
+      "step": 24251
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1855322150017633e-05,
+      "loss": 0.4603,
+      "step": 24252
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1854986314711267e-05,
+      "loss": 0.389,
+      "step": 24253
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1854650479404902e-05,
+      "loss": 0.2895,
+      "step": 24254
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1854314644098535e-05,
+      "loss": 0.3693,
+      "step": 24255
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1853978808792171e-05,
+      "loss": 0.1743,
+      "step": 24256
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1853642973485804e-05,
+      "loss": 0.3606,
+      "step": 24257
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1853307138179437e-05,
+      "loss": 0.2962,
+      "step": 24258
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1852971302873073e-05,
+      "loss": 0.2867,
+      "step": 24259
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1852635467566706e-05,
+      "loss": 0.2943,
+      "step": 24260
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.185229963226034e-05,
+      "loss": 0.4054,
+      "step": 24261
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1851963796953976e-05,
+      "loss": 0.2534,
+      "step": 24262
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1851627961647609e-05,
+      "loss": 0.3664,
+      "step": 24263
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1851292126341243e-05,
+      "loss": 0.2578,
+      "step": 24264
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1850956291034878e-05,
+      "loss": 0.3029,
+      "step": 24265
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1850620455728513e-05,
+      "loss": 0.2132,
+      "step": 24266
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1850284620422146e-05,
+      "loss": 0.2955,
+      "step": 24267
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1849948785115782e-05,
+      "loss": 0.3014,
+      "step": 24268
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1849612949809415e-05,
+      "loss": 0.3262,
+      "step": 24269
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1849277114503048e-05,
+      "loss": 0.4656,
+      "step": 24270
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1848941279196684e-05,
+      "loss": 0.2622,
+      "step": 24271
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1848605443890317e-05,
+      "loss": 0.2693,
+      "step": 24272
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1848269608583952e-05,
+      "loss": 0.3922,
+      "step": 24273
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1847933773277585e-05,
+      "loss": 0.1769,
+      "step": 24274
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1847597937971221e-05,
+      "loss": 0.3105,
+      "step": 24275
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1847262102664854e-05,
+      "loss": 0.481,
+      "step": 24276
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1846926267358487e-05,
+      "loss": 0.1841,
+      "step": 24277
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1846590432052123e-05,
+      "loss": 0.3495,
+      "step": 24278
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1846254596745756e-05,
+      "loss": 0.4151,
+      "step": 24279
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1845918761439391e-05,
+      "loss": 0.3337,
+      "step": 24280
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1845582926133026e-05,
+      "loss": 0.1798,
+      "step": 24281
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1845247090826658e-05,
+      "loss": 0.2933,
+      "step": 24282
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1844911255520293e-05,
+      "loss": 0.4312,
+      "step": 24283
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1844575420213928e-05,
+      "loss": 0.3232,
+      "step": 24284
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1844239584907562e-05,
+      "loss": 0.3571,
+      "step": 24285
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1843903749601195e-05,
+      "loss": 0.3657,
+      "step": 24286
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1843567914294832e-05,
+      "loss": 0.3189,
+      "step": 24287
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1843232078988465e-05,
+      "loss": 0.3783,
+      "step": 24288
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1842896243682098e-05,
+      "loss": 0.3872,
+      "step": 24289
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1842560408375734e-05,
+      "loss": 0.2646,
+      "step": 24290
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1842224573069367e-05,
+      "loss": 0.1817,
+      "step": 24291
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1841888737763002e-05,
+      "loss": 0.373,
+      "step": 24292
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1841552902456636e-05,
+      "loss": 0.2817,
+      "step": 24293
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.184121706715027e-05,
+      "loss": 0.3581,
+      "step": 24294
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1840881231843904e-05,
+      "loss": 0.251,
+      "step": 24295
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.184054539653754e-05,
+      "loss": 0.2713,
+      "step": 24296
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1840209561231173e-05,
+      "loss": 0.3605,
+      "step": 24297
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1839873725924806e-05,
+      "loss": 0.2924,
+      "step": 24298
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1839537890618442e-05,
+      "loss": 0.2844,
+      "step": 24299
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1839202055312075e-05,
+      "loss": 0.3259,
+      "step": 24300
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.183886622000571e-05,
+      "loss": 0.2199,
+      "step": 24301
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1838530384699345e-05,
+      "loss": 0.2202,
+      "step": 24302
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1838194549392978e-05,
+      "loss": 0.235,
+      "step": 24303
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1837858714086612e-05,
+      "loss": 0.3273,
+      "step": 24304
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1837522878780247e-05,
+      "loss": 0.3254,
+      "step": 24305
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1837187043473882e-05,
+      "loss": 0.2859,
+      "step": 24306
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1836851208167514e-05,
+      "loss": 0.4534,
+      "step": 24307
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.183651537286115e-05,
+      "loss": 0.3798,
+      "step": 24308
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1836179537554784e-05,
+      "loss": 0.3242,
+      "step": 24309
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1835843702248417e-05,
+      "loss": 0.3051,
+      "step": 24310
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1835507866942053e-05,
+      "loss": 0.3034,
+      "step": 24311
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1835172031635686e-05,
+      "loss": 0.2412,
+      "step": 24312
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.183483619632932e-05,
+      "loss": 0.2381,
+      "step": 24313
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1834500361022955e-05,
+      "loss": 0.3301,
+      "step": 24314
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.183416452571659e-05,
+      "loss": 0.4671,
+      "step": 24315
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.1833828690410223e-05,
+      "loss": 0.2896,
+      "step": 24316
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.183349285510386e-05,
+      "loss": 0.2674,
+      "step": 24317
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1833157019797492e-05,
+      "loss": 0.2157,
+      "step": 24318
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1832821184491125e-05,
+      "loss": 0.3313,
+      "step": 24319
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1832485349184762e-05,
+      "loss": 0.3237,
+      "step": 24320
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1832149513878394e-05,
+      "loss": 0.2076,
+      "step": 24321
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1831813678572029e-05,
+      "loss": 0.2601,
+      "step": 24322
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1831477843265664e-05,
+      "loss": 0.2292,
+      "step": 24323
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1831142007959298e-05,
+      "loss": 0.3474,
+      "step": 24324
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1830806172652931e-05,
+      "loss": 0.3825,
+      "step": 24325
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1830470337346566e-05,
+      "loss": 0.1726,
+      "step": 24326
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.18301345020402e-05,
+      "loss": 0.2259,
+      "step": 24327
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1829798666733834e-05,
+      "loss": 0.2489,
+      "step": 24328
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.182946283142747e-05,
+      "loss": 0.2644,
+      "step": 24329
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1829126996121103e-05,
+      "loss": 0.2029,
+      "step": 24330
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1828791160814736e-05,
+      "loss": 0.2354,
+      "step": 24331
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1828455325508372e-05,
+      "loss": 0.2228,
+      "step": 24332
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1828119490202005e-05,
+      "loss": 0.2482,
+      "step": 24333
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.182778365489564e-05,
+      "loss": 0.3699,
+      "step": 24334
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1827447819589274e-05,
+      "loss": 0.4034,
+      "step": 24335
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1827111984282909e-05,
+      "loss": 0.5388,
+      "step": 24336
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1826776148976542e-05,
+      "loss": 0.3935,
+      "step": 24337
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1826440313670178e-05,
+      "loss": 0.3181,
+      "step": 24338
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1826104478363811e-05,
+      "loss": 0.2927,
+      "step": 24339
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1825768643057444e-05,
+      "loss": 0.2407,
+      "step": 24340
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.182543280775108e-05,
+      "loss": 0.2308,
+      "step": 24341
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1825096972444714e-05,
+      "loss": 0.1439,
+      "step": 24342
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1824761137138348e-05,
+      "loss": 0.1438,
+      "step": 24343
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1824425301831983e-05,
+      "loss": 0.2344,
+      "step": 24344
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1824089466525618e-05,
+      "loss": 0.4616,
+      "step": 24345
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.182375363121925e-05,
+      "loss": 0.4063,
+      "step": 24346
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1823417795912885e-05,
+      "loss": 0.5007,
+      "step": 24347
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.182308196060652e-05,
+      "loss": 0.1467,
+      "step": 24348
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1822746125300153e-05,
+      "loss": 0.2557,
+      "step": 24349
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1822410289993789e-05,
+      "loss": 0.2822,
+      "step": 24350
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1822074454687422e-05,
+      "loss": 0.4689,
+      "step": 24351
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1821738619381055e-05,
+      "loss": 0.2308,
+      "step": 24352
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1821402784074691e-05,
+      "loss": 0.3592,
+      "step": 24353
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1821066948768324e-05,
+      "loss": 0.1952,
+      "step": 24354
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1820731113461959e-05,
+      "loss": 0.3811,
+      "step": 24355
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1820395278155594e-05,
+      "loss": 0.1219,
+      "step": 24356
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1820059442849228e-05,
+      "loss": 0.2693,
+      "step": 24357
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1819723607542861e-05,
+      "loss": 0.5764,
+      "step": 24358
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1819387772236498e-05,
+      "loss": 0.4316,
+      "step": 24359
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.181905193693013e-05,
+      "loss": 0.1817,
+      "step": 24360
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1818716101623763e-05,
+      "loss": 0.2936,
+      "step": 24361
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.18183802663174e-05,
+      "loss": 0.2208,
+      "step": 24362
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1818044431011033e-05,
+      "loss": 0.3477,
+      "step": 24363
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1817708595704667e-05,
+      "loss": 0.2969,
+      "step": 24364
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1817372760398302e-05,
+      "loss": 0.1944,
+      "step": 24365
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1817036925091937e-05,
+      "loss": 0.3409,
+      "step": 24366
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.181670108978557e-05,
+      "loss": 0.4179,
+      "step": 24367
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1816365254479206e-05,
+      "loss": 0.3946,
+      "step": 24368
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1816029419172839e-05,
+      "loss": 0.4443,
+      "step": 24369
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1815693583866472e-05,
+      "loss": 0.3294,
+      "step": 24370
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1815357748560108e-05,
+      "loss": 0.1974,
+      "step": 24371
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1815021913253741e-05,
+      "loss": 0.2425,
+      "step": 24372
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1814686077947374e-05,
+      "loss": 0.2891,
+      "step": 24373
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.181435024264101e-05,
+      "loss": 0.2596,
+      "step": 24374
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1814014407334643e-05,
+      "loss": 0.3336,
+      "step": 24375
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1813678572028278e-05,
+      "loss": 0.2352,
+      "step": 24376
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1813342736721913e-05,
+      "loss": 0.2232,
+      "step": 24377
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1813006901415547e-05,
+      "loss": 0.2579,
+      "step": 24378
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.181267106610918e-05,
+      "loss": 0.3202,
+      "step": 24379
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1812335230802817e-05,
+      "loss": 0.4826,
+      "step": 24380
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.181199939549645e-05,
+      "loss": 0.3107,
+      "step": 24381
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1811663560190083e-05,
+      "loss": 0.4315,
+      "step": 24382
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1811327724883719e-05,
+      "loss": 0.5125,
+      "step": 24383
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1810991889577352e-05,
+      "loss": 0.2327,
+      "step": 24384
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1810656054270986e-05,
+      "loss": 0.3365,
+      "step": 24385
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1810320218964621e-05,
+      "loss": 0.2487,
+      "step": 24386
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1809984383658256e-05,
+      "loss": 0.3335,
+      "step": 24387
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1809648548351889e-05,
+      "loss": 0.3008,
+      "step": 24388
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1809312713045525e-05,
+      "loss": 0.2314,
+      "step": 24389
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1808976877739158e-05,
+      "loss": 0.302,
+      "step": 24390
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1808641042432791e-05,
+      "loss": 0.2716,
+      "step": 24391
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1808305207126427e-05,
+      "loss": 0.2558,
+      "step": 24392
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.180796937182006e-05,
+      "loss": 0.294,
+      "step": 24393
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1807633536513693e-05,
+      "loss": 0.2798,
+      "step": 24394
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.180729770120733e-05,
+      "loss": 0.3071,
+      "step": 24395
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1806961865900963e-05,
+      "loss": 0.2422,
+      "step": 24396
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1806626030594597e-05,
+      "loss": 0.478,
+      "step": 24397
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1806290195288232e-05,
+      "loss": 0.2708,
+      "step": 24398
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1805954359981866e-05,
+      "loss": 0.205,
+      "step": 24399
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.18056185246755e-05,
+      "loss": 0.2113,
+      "step": 24400
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1805282689369136e-05,
+      "loss": 0.5113,
+      "step": 24401
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1804946854062769e-05,
+      "loss": 0.1811,
+      "step": 24402
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1804611018756402e-05,
+      "loss": 0.2079,
+      "step": 24403
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1804275183450038e-05,
+      "loss": 0.3835,
+      "step": 24404
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1803939348143671e-05,
+      "loss": 0.2776,
+      "step": 24405
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1803603512837306e-05,
+      "loss": 0.2818,
+      "step": 24406
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.180326767753094e-05,
+      "loss": 0.3866,
+      "step": 24407
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1802931842224575e-05,
+      "loss": 0.2747,
+      "step": 24408
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1802596006918208e-05,
+      "loss": 0.1471,
+      "step": 24409
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1802260171611844e-05,
+      "loss": 0.4316,
+      "step": 24410
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1801924336305477e-05,
+      "loss": 0.2575,
+      "step": 24411
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.180158850099911e-05,
+      "loss": 0.2663,
+      "step": 24412
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1801252665692746e-05,
+      "loss": 0.2359,
+      "step": 24413
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.180091683038638e-05,
+      "loss": 0.3183,
+      "step": 24414
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1800580995080014e-05,
+      "loss": 0.5682,
+      "step": 24415
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1800245159773649e-05,
+      "loss": 0.3005,
+      "step": 24416
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1799909324467282e-05,
+      "loss": 0.225,
+      "step": 24417
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1799573489160916e-05,
+      "loss": 0.3666,
+      "step": 24418
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1799237653854551e-05,
+      "loss": 0.2615,
+      "step": 24419
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1798901818548186e-05,
+      "loss": 0.3322,
+      "step": 24420
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1798565983241819e-05,
+      "loss": 0.2224,
+      "step": 24421
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1798230147935455e-05,
+      "loss": 0.2158,
+      "step": 24422
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1797894312629088e-05,
+      "loss": 0.3181,
+      "step": 24423
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.179755847732272e-05,
+      "loss": 0.2663,
+      "step": 24424
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1797222642016357e-05,
+      "loss": 0.2543,
+      "step": 24425
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.179688680670999e-05,
+      "loss": 0.3254,
+      "step": 24426
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1796550971403625e-05,
+      "loss": 0.293,
+      "step": 24427
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.179621513609726e-05,
+      "loss": 0.3248,
+      "step": 24428
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1795879300790894e-05,
+      "loss": 0.351,
+      "step": 24429
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1795543465484527e-05,
+      "loss": 0.2851,
+      "step": 24430
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1795207630178163e-05,
+      "loss": 0.4573,
+      "step": 24431
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1794871794871796e-05,
+      "loss": 0.3666,
+      "step": 24432
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.179453595956543e-05,
+      "loss": 0.2821,
+      "step": 24433
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1794200124259066e-05,
+      "loss": 0.2235,
+      "step": 24434
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1793864288952699e-05,
+      "loss": 0.3875,
+      "step": 24435
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1793528453646333e-05,
+      "loss": 0.3425,
+      "step": 24436
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1793192618339968e-05,
+      "loss": 0.398,
+      "step": 24437
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.17928567830336e-05,
+      "loss": 0.3029,
+      "step": 24438
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1792520947727235e-05,
+      "loss": 0.233,
+      "step": 24439
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.179218511242087e-05,
+      "loss": 0.4437,
+      "step": 24440
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1791849277114505e-05,
+      "loss": 0.3919,
+      "step": 24441
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1791513441808138e-05,
+      "loss": 0.2543,
+      "step": 24442
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1791177606501774e-05,
+      "loss": 0.191,
+      "step": 24443
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1790841771195407e-05,
+      "loss": 0.5588,
+      "step": 24444
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.179050593588904e-05,
+      "loss": 0.3447,
+      "step": 24445
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1790170100582676e-05,
+      "loss": 0.3757,
+      "step": 24446
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.178983426527631e-05,
+      "loss": 0.2416,
+      "step": 24447
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1789498429969944e-05,
+      "loss": 0.2089,
+      "step": 24448
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1789162594663579e-05,
+      "loss": 0.3099,
+      "step": 24449
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1788826759357213e-05,
+      "loss": 0.2822,
+      "step": 24450
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1788490924050846e-05,
+      "loss": 0.3441,
+      "step": 24451
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1788155088744482e-05,
+      "loss": 0.381,
+      "step": 24452
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1787819253438115e-05,
+      "loss": 0.2982,
+      "step": 24453
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1787483418131748e-05,
+      "loss": 0.3177,
+      "step": 24454
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1787147582825383e-05,
+      "loss": 0.2382,
+      "step": 24455
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1786811747519018e-05,
+      "loss": 0.23,
+      "step": 24456
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1786475912212652e-05,
+      "loss": 0.254,
+      "step": 24457
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1786140076906285e-05,
+      "loss": 0.1924,
+      "step": 24458
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1785804241599922e-05,
+      "loss": 0.2268,
+      "step": 24459
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1785468406293555e-05,
+      "loss": 0.4357,
+      "step": 24460
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1785132570987187e-05,
+      "loss": 0.1969,
+      "step": 24461
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1784796735680824e-05,
+      "loss": 0.2748,
+      "step": 24462
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1784460900374457e-05,
+      "loss": 0.4325,
+      "step": 24463
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.178412506506809e-05,
+      "loss": 0.3131,
+      "step": 24464
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1783789229761726e-05,
+      "loss": 0.2113,
+      "step": 24465
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1783453394455359e-05,
+      "loss": 0.2361,
+      "step": 24466
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1783117559148994e-05,
+      "loss": 0.2516,
+      "step": 24467
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1782781723842628e-05,
+      "loss": 0.4207,
+      "step": 24468
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1782445888536263e-05,
+      "loss": 0.4488,
+      "step": 24469
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1782110053229896e-05,
+      "loss": 0.202,
+      "step": 24470
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1781774217923532e-05,
+      "loss": 0.2599,
+      "step": 24471
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1781438382617165e-05,
+      "loss": 0.3221,
+      "step": 24472
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1781102547310798e-05,
+      "loss": 0.172,
+      "step": 24473
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1780766712004435e-05,
+      "loss": 0.1917,
+      "step": 24474
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1780430876698067e-05,
+      "loss": 0.1608,
+      "step": 24475
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1780095041391702e-05,
+      "loss": 0.2695,
+      "step": 24476
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1779759206085337e-05,
+      "loss": 0.347,
+      "step": 24477
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1779423370778971e-05,
+      "loss": 0.4577,
+      "step": 24478
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1779087535472604e-05,
+      "loss": 0.1955,
+      "step": 24479
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.177875170016624e-05,
+      "loss": 0.417,
+      "step": 24480
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1778415864859874e-05,
+      "loss": 0.1914,
+      "step": 24481
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1778080029553507e-05,
+      "loss": 0.334,
+      "step": 24482
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1777744194247143e-05,
+      "loss": 0.2464,
+      "step": 24483
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1777408358940776e-05,
+      "loss": 0.2567,
+      "step": 24484
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1777072523634409e-05,
+      "loss": 0.3163,
+      "step": 24485
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1776736688328045e-05,
+      "loss": 0.3198,
+      "step": 24486
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1776400853021678e-05,
+      "loss": 0.3331,
+      "step": 24487
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1776065017715313e-05,
+      "loss": 0.1576,
+      "step": 24488
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1775729182408947e-05,
+      "loss": 0.2146,
+      "step": 24489
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1775393347102582e-05,
+      "loss": 0.4372,
+      "step": 24490
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1775057511796215e-05,
+      "loss": 0.3279,
+      "step": 24491
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1774721676489851e-05,
+      "loss": 0.3825,
+      "step": 24492
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1774385841183484e-05,
+      "loss": 0.2358,
+      "step": 24493
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1774050005877117e-05,
+      "loss": 0.2128,
+      "step": 24494
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1773714170570754e-05,
+      "loss": 0.3692,
+      "step": 24495
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1773378335264387e-05,
+      "loss": 0.3132,
+      "step": 24496
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1773042499958021e-05,
+      "loss": 0.2011,
+      "step": 24497
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1772706664651656e-05,
+      "loss": 0.1582,
+      "step": 24498
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.177237082934529e-05,
+      "loss": 0.4473,
+      "step": 24499
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1772034994038923e-05,
+      "loss": 0.2568,
+      "step": 24500
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.177169915873256e-05,
+      "loss": 0.2535,
+      "step": 24501
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1771363323426193e-05,
+      "loss": 0.3543,
+      "step": 24502
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1771027488119826e-05,
+      "loss": 0.1797,
+      "step": 24503
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1770691652813462e-05,
+      "loss": 0.1417,
+      "step": 24504
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1770355817507095e-05,
+      "loss": 0.4416,
+      "step": 24505
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1770019982200728e-05,
+      "loss": 0.2353,
+      "step": 24506
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1769684146894364e-05,
+      "loss": 0.4087,
+      "step": 24507
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1769348311587997e-05,
+      "loss": 0.4259,
+      "step": 24508
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1769012476281632e-05,
+      "loss": 0.2199,
+      "step": 24509
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1768676640975267e-05,
+      "loss": 0.2258,
+      "step": 24510
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1768340805668901e-05,
+      "loss": 0.2561,
+      "step": 24511
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1768004970362534e-05,
+      "loss": 0.1593,
+      "step": 24512
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.176766913505617e-05,
+      "loss": 0.2117,
+      "step": 24513
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1767333299749803e-05,
+      "loss": 0.5008,
+      "step": 24514
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 1.1766997464443436e-05,
+      "loss": 0.3575,
+      "step": 24515
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1766661629137073e-05,
+      "loss": 0.4313,
+      "step": 24516
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1766325793830706e-05,
+      "loss": 0.3213,
+      "step": 24517
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.176598995852434e-05,
+      "loss": 0.3173,
+      "step": 24518
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1765654123217975e-05,
+      "loss": 0.4322,
+      "step": 24519
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.176531828791161e-05,
+      "loss": 0.2021,
+      "step": 24520
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1764982452605243e-05,
+      "loss": 0.3105,
+      "step": 24521
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1764646617298879e-05,
+      "loss": 0.1866,
+      "step": 24522
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1764310781992512e-05,
+      "loss": 0.3152,
+      "step": 24523
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1763974946686145e-05,
+      "loss": 0.2835,
+      "step": 24524
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1763639111379781e-05,
+      "loss": 0.303,
+      "step": 24525
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1763303276073414e-05,
+      "loss": 0.3557,
+      "step": 24526
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1762967440767049e-05,
+      "loss": 0.1548,
+      "step": 24527
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1762631605460683e-05,
+      "loss": 0.4725,
+      "step": 24528
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1762295770154316e-05,
+      "loss": 0.2119,
+      "step": 24529
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1761959934847951e-05,
+      "loss": 0.1288,
+      "step": 24530
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1761624099541586e-05,
+      "loss": 0.1408,
+      "step": 24531
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.176128826423522e-05,
+      "loss": 0.231,
+      "step": 24532
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1760952428928853e-05,
+      "loss": 0.1517,
+      "step": 24533
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.176061659362249e-05,
+      "loss": 0.3322,
+      "step": 24534
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1760280758316123e-05,
+      "loss": 0.2526,
+      "step": 24535
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1759944923009756e-05,
+      "loss": 0.4443,
+      "step": 24536
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1759609087703392e-05,
+      "loss": 0.3539,
+      "step": 24537
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1759273252397025e-05,
+      "loss": 0.1984,
+      "step": 24538
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.175893741709066e-05,
+      "loss": 0.1723,
+      "step": 24539
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1758601581784294e-05,
+      "loss": 0.2989,
+      "step": 24540
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1758265746477929e-05,
+      "loss": 0.3846,
+      "step": 24541
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1757929911171562e-05,
+      "loss": 0.2043,
+      "step": 24542
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1757594075865198e-05,
+      "loss": 0.471,
+      "step": 24543
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1757258240558831e-05,
+      "loss": 0.2727,
+      "step": 24544
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1756922405252464e-05,
+      "loss": 0.2827,
+      "step": 24545
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.17565865699461e-05,
+      "loss": 0.2254,
+      "step": 24546
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1756250734639733e-05,
+      "loss": 0.1799,
+      "step": 24547
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1755914899333368e-05,
+      "loss": 0.2092,
+      "step": 24548
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1755579064027003e-05,
+      "loss": 0.458,
+      "step": 24549
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1755243228720637e-05,
+      "loss": 0.6377,
+      "step": 24550
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.175490739341427e-05,
+      "loss": 0.2708,
+      "step": 24551
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1754571558107905e-05,
+      "loss": 0.2685,
+      "step": 24552
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.175423572280154e-05,
+      "loss": 0.1708,
+      "step": 24553
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1753899887495172e-05,
+      "loss": 0.1306,
+      "step": 24554
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1753564052188809e-05,
+      "loss": 0.3931,
+      "step": 24555
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1753228216882442e-05,
+      "loss": 0.2771,
+      "step": 24556
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1752892381576075e-05,
+      "loss": 0.2422,
+      "step": 24557
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1752556546269711e-05,
+      "loss": 0.2991,
+      "step": 24558
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1752220710963344e-05,
+      "loss": 0.3369,
+      "step": 24559
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1751884875656979e-05,
+      "loss": 0.3105,
+      "step": 24560
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1751549040350613e-05,
+      "loss": 0.2859,
+      "step": 24561
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1751213205044248e-05,
+      "loss": 0.4803,
+      "step": 24562
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1750877369737881e-05,
+      "loss": 0.4749,
+      "step": 24563
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1750541534431517e-05,
+      "loss": 0.3708,
+      "step": 24564
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.175020569912515e-05,
+      "loss": 0.1322,
+      "step": 24565
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1749869863818783e-05,
+      "loss": 0.3229,
+      "step": 24566
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.174953402851242e-05,
+      "loss": 0.2687,
+      "step": 24567
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1749198193206052e-05,
+      "loss": 0.2042,
+      "step": 24568
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1748862357899687e-05,
+      "loss": 0.2885,
+      "step": 24569
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1748526522593322e-05,
+      "loss": 0.2887,
+      "step": 24570
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1748190687286956e-05,
+      "loss": 0.2602,
+      "step": 24571
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.174785485198059e-05,
+      "loss": 0.3066,
+      "step": 24572
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1747519016674224e-05,
+      "loss": 0.3182,
+      "step": 24573
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1747183181367859e-05,
+      "loss": 0.3038,
+      "step": 24574
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1746847346061492e-05,
+      "loss": 0.5532,
+      "step": 24575
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1746511510755128e-05,
+      "loss": 0.3958,
+      "step": 24576
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1746175675448761e-05,
+      "loss": 0.3566,
+      "step": 24577
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1745839840142394e-05,
+      "loss": 0.1996,
+      "step": 24578
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.174550400483603e-05,
+      "loss": 0.3532,
+      "step": 24579
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1745168169529663e-05,
+      "loss": 0.3613,
+      "step": 24580
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1744832334223298e-05,
+      "loss": 0.3298,
+      "step": 24581
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1744496498916932e-05,
+      "loss": 0.2894,
+      "step": 24582
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1744160663610567e-05,
+      "loss": 0.2423,
+      "step": 24583
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.17438248283042e-05,
+      "loss": 0.1742,
+      "step": 24584
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1743488992997836e-05,
+      "loss": 0.2917,
+      "step": 24585
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.174315315769147e-05,
+      "loss": 0.4012,
+      "step": 24586
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1742817322385102e-05,
+      "loss": 0.3989,
+      "step": 24587
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1742481487078739e-05,
+      "loss": 0.3924,
+      "step": 24588
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1742145651772372e-05,
+      "loss": 0.2525,
+      "step": 24589
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1741809816466006e-05,
+      "loss": 0.2411,
+      "step": 24590
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.174147398115964e-05,
+      "loss": 0.3449,
+      "step": 24591
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1741138145853275e-05,
+      "loss": 0.2047,
+      "step": 24592
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1740802310546908e-05,
+      "loss": 0.4571,
+      "step": 24593
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1740466475240545e-05,
+      "loss": 0.3956,
+      "step": 24594
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1740130639934178e-05,
+      "loss": 0.1777,
+      "step": 24595
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.173979480462781e-05,
+      "loss": 0.2559,
+      "step": 24596
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1739458969321447e-05,
+      "loss": 0.2787,
+      "step": 24597
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.173912313401508e-05,
+      "loss": 0.299,
+      "step": 24598
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1738787298708713e-05,
+      "loss": 0.1638,
+      "step": 24599
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.173845146340235e-05,
+      "loss": 0.1926,
+      "step": 24600
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1738115628095982e-05,
+      "loss": 0.279,
+      "step": 24601
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1737779792789617e-05,
+      "loss": 0.2772,
+      "step": 24602
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1737443957483252e-05,
+      "loss": 0.3356,
+      "step": 24603
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1737108122176886e-05,
+      "loss": 0.3013,
+      "step": 24604
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1736772286870519e-05,
+      "loss": 0.2405,
+      "step": 24605
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1736436451564155e-05,
+      "loss": 0.2761,
+      "step": 24606
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1736100616257788e-05,
+      "loss": 0.3336,
+      "step": 24607
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1735764780951421e-05,
+      "loss": 0.2633,
+      "step": 24608
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1735428945645058e-05,
+      "loss": 0.1863,
+      "step": 24609
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.173509311033869e-05,
+      "loss": 0.2666,
+      "step": 24610
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1734757275032325e-05,
+      "loss": 0.2247,
+      "step": 24611
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.173442143972596e-05,
+      "loss": 0.2481,
+      "step": 24612
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1734085604419595e-05,
+      "loss": 0.3848,
+      "step": 24613
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1733749769113228e-05,
+      "loss": 0.3284,
+      "step": 24614
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1733413933806864e-05,
+      "loss": 0.3732,
+      "step": 24615
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1733078098500497e-05,
+      "loss": 0.2813,
+      "step": 24616
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.173274226319413e-05,
+      "loss": 0.204,
+      "step": 24617
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1732406427887766e-05,
+      "loss": 0.2448,
+      "step": 24618
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1732070592581399e-05,
+      "loss": 0.2374,
+      "step": 24619
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1731734757275032e-05,
+      "loss": 0.3204,
+      "step": 24620
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1731398921968668e-05,
+      "loss": 0.2598,
+      "step": 24621
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1731063086662301e-05,
+      "loss": 0.2115,
+      "step": 24622
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1730727251355936e-05,
+      "loss": 0.4031,
+      "step": 24623
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.173039141604957e-05,
+      "loss": 0.2308,
+      "step": 24624
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1730055580743205e-05,
+      "loss": 0.2561,
+      "step": 24625
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1729719745436838e-05,
+      "loss": 0.1867,
+      "step": 24626
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1729383910130475e-05,
+      "loss": 0.3561,
+      "step": 24627
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1729048074824108e-05,
+      "loss": 0.4277,
+      "step": 24628
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.172871223951774e-05,
+      "loss": 0.3114,
+      "step": 24629
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1728376404211377e-05,
+      "loss": 0.1396,
+      "step": 24630
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.172804056890501e-05,
+      "loss": 0.2696,
+      "step": 24631
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1727704733598644e-05,
+      "loss": 0.3643,
+      "step": 24632
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1727368898292279e-05,
+      "loss": 0.1963,
+      "step": 24633
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1727033062985914e-05,
+      "loss": 0.4834,
+      "step": 24634
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1726697227679547e-05,
+      "loss": 0.6227,
+      "step": 24635
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.172636139237318e-05,
+      "loss": 0.3844,
+      "step": 24636
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1726025557066816e-05,
+      "loss": 0.2652,
+      "step": 24637
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1725689721760449e-05,
+      "loss": 0.2934,
+      "step": 24638
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1725353886454084e-05,
+      "loss": 0.4238,
+      "step": 24639
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1725018051147718e-05,
+      "loss": 0.3157,
+      "step": 24640
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1724682215841351e-05,
+      "loss": 0.2616,
+      "step": 24641
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1724346380534986e-05,
+      "loss": 0.2487,
+      "step": 24642
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.172401054522862e-05,
+      "loss": 0.4197,
+      "step": 24643
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1723674709922255e-05,
+      "loss": 0.3209,
+      "step": 24644
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1723338874615888e-05,
+      "loss": 0.3474,
+      "step": 24645
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1723003039309524e-05,
+      "loss": 0.2617,
+      "step": 24646
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1722667204003157e-05,
+      "loss": 0.5228,
+      "step": 24647
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.172233136869679e-05,
+      "loss": 0.325,
+      "step": 24648
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1721995533390427e-05,
+      "loss": 0.1977,
+      "step": 24649
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.172165969808406e-05,
+      "loss": 0.3329,
+      "step": 24650
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1721323862777694e-05,
+      "loss": 0.2284,
+      "step": 24651
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1720988027471329e-05,
+      "loss": 0.2785,
+      "step": 24652
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1720652192164964e-05,
+      "loss": 0.2917,
+      "step": 24653
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1720316356858597e-05,
+      "loss": 0.288,
+      "step": 24654
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1719980521552233e-05,
+      "loss": 0.2212,
+      "step": 24655
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1719644686245866e-05,
+      "loss": 0.4452,
+      "step": 24656
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1719308850939499e-05,
+      "loss": 0.3439,
+      "step": 24657
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1718973015633135e-05,
+      "loss": 0.4154,
+      "step": 24658
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1718637180326768e-05,
+      "loss": 0.3105,
+      "step": 24659
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1718301345020403e-05,
+      "loss": 0.3315,
+      "step": 24660
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1717965509714037e-05,
+      "loss": 0.1935,
+      "step": 24661
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1717629674407672e-05,
+      "loss": 0.3691,
+      "step": 24662
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1717293839101305e-05,
+      "loss": 0.3233,
+      "step": 24663
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.171695800379494e-05,
+      "loss": 0.3493,
+      "step": 24664
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1716622168488574e-05,
+      "loss": 0.241,
+      "step": 24665
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1716286333182207e-05,
+      "loss": 0.3435,
+      "step": 24666
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1715950497875844e-05,
+      "loss": 0.1725,
+      "step": 24667
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1715614662569476e-05,
+      "loss": 0.2656,
+      "step": 24668
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.171527882726311e-05,
+      "loss": 0.2399,
+      "step": 24669
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1714942991956746e-05,
+      "loss": 0.359,
+      "step": 24670
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1714607156650379e-05,
+      "loss": 0.2377,
+      "step": 24671
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1714271321344013e-05,
+      "loss": 0.2304,
+      "step": 24672
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1713935486037648e-05,
+      "loss": 0.1387,
+      "step": 24673
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1713599650731283e-05,
+      "loss": 0.3352,
+      "step": 24674
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1713263815424916e-05,
+      "loss": 0.3256,
+      "step": 24675
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1712927980118552e-05,
+      "loss": 0.2579,
+      "step": 24676
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1712592144812185e-05,
+      "loss": 0.1561,
+      "step": 24677
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1712256309505818e-05,
+      "loss": 0.3961,
+      "step": 24678
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1711920474199454e-05,
+      "loss": 0.4676,
+      "step": 24679
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1711584638893087e-05,
+      "loss": 0.2838,
+      "step": 24680
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1711248803586722e-05,
+      "loss": 0.2515,
+      "step": 24681
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1710912968280356e-05,
+      "loss": 0.4741,
+      "step": 24682
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1710577132973991e-05,
+      "loss": 0.4858,
+      "step": 24683
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1710241297667624e-05,
+      "loss": 0.4641,
+      "step": 24684
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.170990546236126e-05,
+      "loss": 0.2437,
+      "step": 24685
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1709569627054893e-05,
+      "loss": 0.2101,
+      "step": 24686
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1709233791748526e-05,
+      "loss": 0.3982,
+      "step": 24687
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1708897956442163e-05,
+      "loss": 0.2174,
+      "step": 24688
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1708562121135796e-05,
+      "loss": 0.2973,
+      "step": 24689
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1708226285829429e-05,
+      "loss": 0.3404,
+      "step": 24690
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1707890450523065e-05,
+      "loss": 0.3377,
+      "step": 24691
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1707554615216698e-05,
+      "loss": 0.2921,
+      "step": 24692
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1707218779910333e-05,
+      "loss": 0.2095,
+      "step": 24693
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1706882944603967e-05,
+      "loss": 0.2388,
+      "step": 24694
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1706547109297602e-05,
+      "loss": 0.5093,
+      "step": 24695
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1706211273991235e-05,
+      "loss": 0.191,
+      "step": 24696
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1705875438684871e-05,
+      "loss": 0.3087,
+      "step": 24697
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1705539603378504e-05,
+      "loss": 0.188,
+      "step": 24698
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1705203768072137e-05,
+      "loss": 0.1951,
+      "step": 24699
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1704867932765773e-05,
+      "loss": 0.295,
+      "step": 24700
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1704532097459406e-05,
+      "loss": 0.2519,
+      "step": 24701
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1704196262153041e-05,
+      "loss": 0.3095,
+      "step": 24702
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1703860426846676e-05,
+      "loss": 0.2603,
+      "step": 24703
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.170352459154031e-05,
+      "loss": 0.3556,
+      "step": 24704
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1703188756233943e-05,
+      "loss": 0.2351,
+      "step": 24705
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.170285292092758e-05,
+      "loss": 0.2243,
+      "step": 24706
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1702517085621212e-05,
+      "loss": 0.3307,
+      "step": 24707
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1702181250314845e-05,
+      "loss": 0.318,
+      "step": 24708
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1701845415008482e-05,
+      "loss": 0.4174,
+      "step": 24709
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1701509579702115e-05,
+      "loss": 0.3095,
+      "step": 24710
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1701173744395748e-05,
+      "loss": 0.4476,
+      "step": 24711
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1700837909089384e-05,
+      "loss": 0.2008,
+      "step": 24712
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1700502073783017e-05,
+      "loss": 0.3501,
+      "step": 24713
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 1.1700166238476652e-05,
+      "loss": 0.2487,
+      "step": 24714
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1699830403170286e-05,
+      "loss": 0.2577,
+      "step": 24715
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1699494567863921e-05,
+      "loss": 0.2912,
+      "step": 24716
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1699158732557554e-05,
+      "loss": 0.2442,
+      "step": 24717
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.169882289725119e-05,
+      "loss": 0.1289,
+      "step": 24718
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1698487061944823e-05,
+      "loss": 0.2668,
+      "step": 24719
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1698151226638456e-05,
+      "loss": 0.4082,
+      "step": 24720
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1697815391332092e-05,
+      "loss": 0.4577,
+      "step": 24721
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1697479556025725e-05,
+      "loss": 0.2469,
+      "step": 24722
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.169714372071936e-05,
+      "loss": 0.3097,
+      "step": 24723
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1696807885412995e-05,
+      "loss": 0.1804,
+      "step": 24724
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.169647205010663e-05,
+      "loss": 0.3877,
+      "step": 24725
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1696136214800262e-05,
+      "loss": 0.2619,
+      "step": 24726
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1695800379493899e-05,
+      "loss": 0.1981,
+      "step": 24727
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1695464544187532e-05,
+      "loss": 0.3283,
+      "step": 24728
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1695128708881165e-05,
+      "loss": 0.2521,
+      "step": 24729
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1694792873574801e-05,
+      "loss": 0.1928,
+      "step": 24730
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1694457038268434e-05,
+      "loss": 0.2543,
+      "step": 24731
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1694121202962067e-05,
+      "loss": 0.4504,
+      "step": 24732
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1693785367655703e-05,
+      "loss": 0.2931,
+      "step": 24733
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1693449532349336e-05,
+      "loss": 0.2501,
+      "step": 24734
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.169311369704297e-05,
+      "loss": 0.336,
+      "step": 24735
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1692777861736605e-05,
+      "loss": 0.4646,
+      "step": 24736
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.169244202643024e-05,
+      "loss": 0.2869,
+      "step": 24737
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1692106191123873e-05,
+      "loss": 0.3406,
+      "step": 24738
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.169177035581751e-05,
+      "loss": 0.3143,
+      "step": 24739
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1691434520511142e-05,
+      "loss": 0.2663,
+      "step": 24740
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1691098685204775e-05,
+      "loss": 0.4306,
+      "step": 24741
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1690762849898412e-05,
+      "loss": 0.1636,
+      "step": 24742
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1690427014592045e-05,
+      "loss": 0.2766,
+      "step": 24743
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.169009117928568e-05,
+      "loss": 0.2884,
+      "step": 24744
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1689755343979314e-05,
+      "loss": 0.4569,
+      "step": 24745
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1689419508672948e-05,
+      "loss": 0.3168,
+      "step": 24746
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1689083673366581e-05,
+      "loss": 0.2925,
+      "step": 24747
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1688747838060218e-05,
+      "loss": 0.4014,
+      "step": 24748
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.168841200275385e-05,
+      "loss": 0.2801,
+      "step": 24749
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1688076167447484e-05,
+      "loss": 0.1937,
+      "step": 24750
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.168774033214112e-05,
+      "loss": 0.3542,
+      "step": 24751
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1687404496834753e-05,
+      "loss": 0.2675,
+      "step": 24752
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1687068661528388e-05,
+      "loss": 0.1569,
+      "step": 24753
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1686732826222022e-05,
+      "loss": 0.5252,
+      "step": 24754
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1686396990915655e-05,
+      "loss": 0.1268,
+      "step": 24755
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.168606115560929e-05,
+      "loss": 0.2429,
+      "step": 24756
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1685725320302925e-05,
+      "loss": 0.2567,
+      "step": 24757
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.168538948499656e-05,
+      "loss": 0.3712,
+      "step": 24758
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1685053649690192e-05,
+      "loss": 0.3438,
+      "step": 24759
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1684717814383828e-05,
+      "loss": 0.2857,
+      "step": 24760
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1684381979077461e-05,
+      "loss": 0.2079,
+      "step": 24761
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1684046143771094e-05,
+      "loss": 0.3334,
+      "step": 24762
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.168371030846473e-05,
+      "loss": 0.3004,
+      "step": 24763
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1683374473158364e-05,
+      "loss": 0.1468,
+      "step": 24764
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1683038637851998e-05,
+      "loss": 0.3425,
+      "step": 24765
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1682702802545633e-05,
+      "loss": 0.4202,
+      "step": 24766
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1682366967239268e-05,
+      "loss": 0.4314,
+      "step": 24767
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.16820311319329e-05,
+      "loss": 0.3133,
+      "step": 24768
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1681695296626537e-05,
+      "loss": 0.3088,
+      "step": 24769
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.168135946132017e-05,
+      "loss": 0.1763,
+      "step": 24770
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1681023626013803e-05,
+      "loss": 0.2283,
+      "step": 24771
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.168068779070744e-05,
+      "loss": 0.3763,
+      "step": 24772
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1680351955401072e-05,
+      "loss": 0.3411,
+      "step": 24773
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1680016120094707e-05,
+      "loss": 0.3627,
+      "step": 24774
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1679680284788341e-05,
+      "loss": 0.3382,
+      "step": 24775
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1679344449481974e-05,
+      "loss": 0.3167,
+      "step": 24776
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1679008614175609e-05,
+      "loss": 0.4936,
+      "step": 24777
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1678672778869244e-05,
+      "loss": 0.3248,
+      "step": 24778
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1678336943562878e-05,
+      "loss": 0.3197,
+      "step": 24779
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1678001108256511e-05,
+      "loss": 0.3274,
+      "step": 24780
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1677665272950148e-05,
+      "loss": 0.2699,
+      "step": 24781
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.167732943764378e-05,
+      "loss": 0.2433,
+      "step": 24782
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1676993602337414e-05,
+      "loss": 0.2973,
+      "step": 24783
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.167665776703105e-05,
+      "loss": 0.387,
+      "step": 24784
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1676321931724683e-05,
+      "loss": 0.3445,
+      "step": 24785
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1675986096418317e-05,
+      "loss": 0.1371,
+      "step": 24786
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1675650261111952e-05,
+      "loss": 0.2465,
+      "step": 24787
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1675314425805587e-05,
+      "loss": 0.3296,
+      "step": 24788
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.167497859049922e-05,
+      "loss": 0.2605,
+      "step": 24789
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1674642755192856e-05,
+      "loss": 0.3477,
+      "step": 24790
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1674306919886489e-05,
+      "loss": 0.2178,
+      "step": 24791
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1673971084580122e-05,
+      "loss": 0.3545,
+      "step": 24792
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1673635249273758e-05,
+      "loss": 0.1721,
+      "step": 24793
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1673299413967391e-05,
+      "loss": 0.3423,
+      "step": 24794
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1672963578661026e-05,
+      "loss": 0.2087,
+      "step": 24795
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.167262774335466e-05,
+      "loss": 0.4876,
+      "step": 24796
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1672291908048295e-05,
+      "loss": 0.2141,
+      "step": 24797
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1671956072741928e-05,
+      "loss": 0.2058,
+      "step": 24798
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1671620237435563e-05,
+      "loss": 0.326,
+      "step": 24799
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1671284402129197e-05,
+      "loss": 0.2931,
+      "step": 24800
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.167094856682283e-05,
+      "loss": 0.3833,
+      "step": 24801
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1670612731516467e-05,
+      "loss": 0.2586,
+      "step": 24802
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.16702768962101e-05,
+      "loss": 0.167,
+      "step": 24803
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1669941060903733e-05,
+      "loss": 0.272,
+      "step": 24804
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1669605225597369e-05,
+      "loss": 0.2349,
+      "step": 24805
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1669269390291002e-05,
+      "loss": 0.2736,
+      "step": 24806
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1668933554984637e-05,
+      "loss": 0.2215,
+      "step": 24807
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1668597719678271e-05,
+      "loss": 0.2985,
+      "step": 24808
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1668261884371906e-05,
+      "loss": 0.3773,
+      "step": 24809
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1667926049065539e-05,
+      "loss": 0.2375,
+      "step": 24810
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1667590213759175e-05,
+      "loss": 0.2867,
+      "step": 24811
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1667254378452808e-05,
+      "loss": 0.1964,
+      "step": 24812
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1666918543146441e-05,
+      "loss": 0.2972,
+      "step": 24813
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1666582707840076e-05,
+      "loss": 0.2094,
+      "step": 24814
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.166624687253371e-05,
+      "loss": 0.2703,
+      "step": 24815
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1665911037227345e-05,
+      "loss": 0.2466,
+      "step": 24816
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1665575201920978e-05,
+      "loss": 0.222,
+      "step": 24817
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1665239366614614e-05,
+      "loss": 0.1973,
+      "step": 24818
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1664903531308247e-05,
+      "loss": 0.1414,
+      "step": 24819
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.166456769600188e-05,
+      "loss": 0.2468,
+      "step": 24820
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1664231860695517e-05,
+      "loss": 0.3448,
+      "step": 24821
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.166389602538915e-05,
+      "loss": 0.1607,
+      "step": 24822
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1663560190082782e-05,
+      "loss": 0.2721,
+      "step": 24823
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1663224354776419e-05,
+      "loss": 0.2358,
+      "step": 24824
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1662888519470052e-05,
+      "loss": 0.4169,
+      "step": 24825
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1662552684163686e-05,
+      "loss": 0.2592,
+      "step": 24826
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1662216848857321e-05,
+      "loss": 0.2692,
+      "step": 24827
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1661881013550956e-05,
+      "loss": 0.3405,
+      "step": 24828
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1661545178244589e-05,
+      "loss": 0.2843,
+      "step": 24829
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1661209342938225e-05,
+      "loss": 0.2713,
+      "step": 24830
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1660873507631858e-05,
+      "loss": 0.2132,
+      "step": 24831
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1660537672325491e-05,
+      "loss": 0.324,
+      "step": 24832
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1660201837019127e-05,
+      "loss": 0.327,
+      "step": 24833
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.165986600171276e-05,
+      "loss": 0.2291,
+      "step": 24834
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1659530166406395e-05,
+      "loss": 0.3115,
+      "step": 24835
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.165919433110003e-05,
+      "loss": 0.346,
+      "step": 24836
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1658858495793664e-05,
+      "loss": 0.2574,
+      "step": 24837
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1658522660487297e-05,
+      "loss": 0.2757,
+      "step": 24838
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1658186825180933e-05,
+      "loss": 0.2647,
+      "step": 24839
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1657850989874566e-05,
+      "loss": 0.3045,
+      "step": 24840
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.16575151545682e-05,
+      "loss": 0.2377,
+      "step": 24841
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1657179319261836e-05,
+      "loss": 0.3347,
+      "step": 24842
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1656843483955469e-05,
+      "loss": 0.2208,
+      "step": 24843
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1656507648649103e-05,
+      "loss": 0.1705,
+      "step": 24844
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1656171813342738e-05,
+      "loss": 0.3002,
+      "step": 24845
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1655835978036371e-05,
+      "loss": 0.2739,
+      "step": 24846
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1655500142730006e-05,
+      "loss": 0.4409,
+      "step": 24847
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.165516430742364e-05,
+      "loss": 0.2947,
+      "step": 24848
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1654828472117275e-05,
+      "loss": 0.2649,
+      "step": 24849
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1654492636810908e-05,
+      "loss": 0.3443,
+      "step": 24850
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1654156801504544e-05,
+      "loss": 0.4418,
+      "step": 24851
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1653820966198177e-05,
+      "loss": 0.2899,
+      "step": 24852
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.165348513089181e-05,
+      "loss": 0.1329,
+      "step": 24853
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1653149295585446e-05,
+      "loss": 0.3411,
+      "step": 24854
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.165281346027908e-05,
+      "loss": 0.3236,
+      "step": 24855
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1652477624972714e-05,
+      "loss": 0.2215,
+      "step": 24856
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1652141789666349e-05,
+      "loss": 0.3087,
+      "step": 24857
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1651805954359983e-05,
+      "loss": 0.3482,
+      "step": 24858
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1651470119053616e-05,
+      "loss": 0.307,
+      "step": 24859
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1651134283747253e-05,
+      "loss": 0.2138,
+      "step": 24860
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1650798448440886e-05,
+      "loss": 0.2682,
+      "step": 24861
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1650462613134518e-05,
+      "loss": 0.2435,
+      "step": 24862
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1650126777828155e-05,
+      "loss": 0.1333,
+      "step": 24863
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1649790942521788e-05,
+      "loss": 0.3236,
+      "step": 24864
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1649455107215422e-05,
+      "loss": 0.2235,
+      "step": 24865
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1649119271909057e-05,
+      "loss": 0.3511,
+      "step": 24866
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.164878343660269e-05,
+      "loss": 0.3111,
+      "step": 24867
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1648447601296325e-05,
+      "loss": 0.2552,
+      "step": 24868
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.164811176598996e-05,
+      "loss": 0.3008,
+      "step": 24869
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1647775930683594e-05,
+      "loss": 0.4985,
+      "step": 24870
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1647440095377227e-05,
+      "loss": 0.1854,
+      "step": 24871
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1647104260070863e-05,
+      "loss": 0.2437,
+      "step": 24872
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1646768424764496e-05,
+      "loss": 0.2922,
+      "step": 24873
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1646432589458129e-05,
+      "loss": 0.4698,
+      "step": 24874
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1646096754151765e-05,
+      "loss": 0.3034,
+      "step": 24875
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1645760918845398e-05,
+      "loss": 0.5008,
+      "step": 24876
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1645425083539033e-05,
+      "loss": 0.216,
+      "step": 24877
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1645089248232668e-05,
+      "loss": 0.4301,
+      "step": 24878
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1644753412926302e-05,
+      "loss": 0.4139,
+      "step": 24879
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1644417577619935e-05,
+      "loss": 0.3324,
+      "step": 24880
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1644081742313572e-05,
+      "loss": 0.2827,
+      "step": 24881
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1643745907007205e-05,
+      "loss": 0.2617,
+      "step": 24882
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1643410071700838e-05,
+      "loss": 0.2641,
+      "step": 24883
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1643074236394474e-05,
+      "loss": 0.3004,
+      "step": 24884
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1642738401088107e-05,
+      "loss": 0.3352,
+      "step": 24885
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1642402565781742e-05,
+      "loss": 0.4633,
+      "step": 24886
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1642066730475376e-05,
+      "loss": 0.2313,
+      "step": 24887
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.164173089516901e-05,
+      "loss": 0.3357,
+      "step": 24888
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1641395059862644e-05,
+      "loss": 0.3303,
+      "step": 24889
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1641059224556278e-05,
+      "loss": 0.5083,
+      "step": 24890
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1640723389249913e-05,
+      "loss": 0.264,
+      "step": 24891
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1640387553943546e-05,
+      "loss": 0.3282,
+      "step": 24892
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1640051718637182e-05,
+      "loss": 0.3548,
+      "step": 24893
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1639715883330815e-05,
+      "loss": 0.28,
+      "step": 24894
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1639380048024448e-05,
+      "loss": 0.37,
+      "step": 24895
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1639044212718085e-05,
+      "loss": 0.4361,
+      "step": 24896
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1638708377411718e-05,
+      "loss": 0.4148,
+      "step": 24897
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1638372542105352e-05,
+      "loss": 0.4512,
+      "step": 24898
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1638036706798987e-05,
+      "loss": 0.4266,
+      "step": 24899
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1637700871492622e-05,
+      "loss": 0.2782,
+      "step": 24900
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1637365036186254e-05,
+      "loss": 0.3108,
+      "step": 24901
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.163702920087989e-05,
+      "loss": 0.3724,
+      "step": 24902
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1636693365573524e-05,
+      "loss": 0.2584,
+      "step": 24903
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1636357530267157e-05,
+      "loss": 0.3637,
+      "step": 24904
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1636021694960793e-05,
+      "loss": 0.4194,
+      "step": 24905
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1635685859654426e-05,
+      "loss": 0.3207,
+      "step": 24906
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.163535002434806e-05,
+      "loss": 0.3691,
+      "step": 24907
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1635014189041695e-05,
+      "loss": 0.168,
+      "step": 24908
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.163467835373533e-05,
+      "loss": 0.306,
+      "step": 24909
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1634342518428963e-05,
+      "loss": 0.2764,
+      "step": 24910
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.16340066831226e-05,
+      "loss": 0.184,
+      "step": 24911
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1633670847816232e-05,
+      "loss": 0.3416,
+      "step": 24912
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.1633335012509865e-05,
+      "loss": 0.3226,
+      "step": 24913
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1632999177203501e-05,
+      "loss": 0.2808,
+      "step": 24914
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1632663341897134e-05,
+      "loss": 0.3861,
+      "step": 24915
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1632327506590767e-05,
+      "loss": 0.3649,
+      "step": 24916
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1631991671284404e-05,
+      "loss": 0.6233,
+      "step": 24917
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1631655835978037e-05,
+      "loss": 0.3423,
+      "step": 24918
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1631320000671671e-05,
+      "loss": 0.3092,
+      "step": 24919
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1630984165365306e-05,
+      "loss": 0.2903,
+      "step": 24920
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.163064833005894e-05,
+      "loss": 0.211,
+      "step": 24921
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1630312494752574e-05,
+      "loss": 0.3639,
+      "step": 24922
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.162997665944621e-05,
+      "loss": 0.2068,
+      "step": 24923
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1629640824139843e-05,
+      "loss": 0.45,
+      "step": 24924
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1629304988833476e-05,
+      "loss": 0.2797,
+      "step": 24925
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1628969153527112e-05,
+      "loss": 0.338,
+      "step": 24926
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1628633318220745e-05,
+      "loss": 0.6659,
+      "step": 24927
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.162829748291438e-05,
+      "loss": 0.2084,
+      "step": 24928
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1627961647608014e-05,
+      "loss": 0.2901,
+      "step": 24929
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1627625812301649e-05,
+      "loss": 0.1867,
+      "step": 24930
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1627289976995282e-05,
+      "loss": 0.1541,
+      "step": 24931
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1626954141688918e-05,
+      "loss": 0.3402,
+      "step": 24932
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1626618306382551e-05,
+      "loss": 0.2841,
+      "step": 24933
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1626282471076184e-05,
+      "loss": 0.208,
+      "step": 24934
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.162594663576982e-05,
+      "loss": 0.3344,
+      "step": 24935
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1625610800463454e-05,
+      "loss": 0.24,
+      "step": 24936
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1625274965157087e-05,
+      "loss": 0.3426,
+      "step": 24937
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1624939129850723e-05,
+      "loss": 0.1626,
+      "step": 24938
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1624603294544356e-05,
+      "loss": 0.4206,
+      "step": 24939
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.162426745923799e-05,
+      "loss": 0.4059,
+      "step": 24940
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1623931623931625e-05,
+      "loss": 0.2254,
+      "step": 24941
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.162359578862526e-05,
+      "loss": 0.1589,
+      "step": 24942
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1623259953318893e-05,
+      "loss": 0.4773,
+      "step": 24943
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1622924118012529e-05,
+      "loss": 0.2253,
+      "step": 24944
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1622588282706162e-05,
+      "loss": 0.2229,
+      "step": 24945
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1622252447399795e-05,
+      "loss": 0.3686,
+      "step": 24946
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1621916612093431e-05,
+      "loss": 0.3938,
+      "step": 24947
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1621580776787064e-05,
+      "loss": 0.1747,
+      "step": 24948
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1621244941480699e-05,
+      "loss": 0.2474,
+      "step": 24949
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1620909106174334e-05,
+      "loss": 0.3538,
+      "step": 24950
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1620573270867968e-05,
+      "loss": 0.2322,
+      "step": 24951
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1620237435561601e-05,
+      "loss": 0.4001,
+      "step": 24952
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1619901600255237e-05,
+      "loss": 0.2707,
+      "step": 24953
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.161956576494887e-05,
+      "loss": 0.304,
+      "step": 24954
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1619229929642503e-05,
+      "loss": 0.2863,
+      "step": 24955
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.161889409433614e-05,
+      "loss": 0.4214,
+      "step": 24956
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1618558259029773e-05,
+      "loss": 0.2035,
+      "step": 24957
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1618222423723406e-05,
+      "loss": 0.4366,
+      "step": 24958
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1617886588417042e-05,
+      "loss": 0.3601,
+      "step": 24959
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1617550753110675e-05,
+      "loss": 0.2558,
+      "step": 24960
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.161721491780431e-05,
+      "loss": 0.2204,
+      "step": 24961
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1616879082497944e-05,
+      "loss": 0.2914,
+      "step": 24962
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1616543247191579e-05,
+      "loss": 0.3633,
+      "step": 24963
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1616207411885212e-05,
+      "loss": 0.2888,
+      "step": 24964
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1615871576578848e-05,
+      "loss": 0.2009,
+      "step": 24965
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1615535741272481e-05,
+      "loss": 0.3195,
+      "step": 24966
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1615199905966114e-05,
+      "loss": 0.2714,
+      "step": 24967
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.161486407065975e-05,
+      "loss": 0.1962,
+      "step": 24968
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1614528235353383e-05,
+      "loss": 0.2103,
+      "step": 24969
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1614192400047018e-05,
+      "loss": 0.2849,
+      "step": 24970
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1613856564740653e-05,
+      "loss": 0.3017,
+      "step": 24971
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1613520729434287e-05,
+      "loss": 0.3044,
+      "step": 24972
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.161318489412792e-05,
+      "loss": 0.2791,
+      "step": 24973
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1612849058821557e-05,
+      "loss": 0.4167,
+      "step": 24974
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.161251322351519e-05,
+      "loss": 0.2739,
+      "step": 24975
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1612177388208823e-05,
+      "loss": 0.1589,
+      "step": 24976
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1611841552902459e-05,
+      "loss": 0.4207,
+      "step": 24977
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1611505717596092e-05,
+      "loss": 0.2677,
+      "step": 24978
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1611169882289726e-05,
+      "loss": 0.2572,
+      "step": 24979
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1610834046983361e-05,
+      "loss": 0.1913,
+      "step": 24980
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1610498211676994e-05,
+      "loss": 0.2923,
+      "step": 24981
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1610162376370629e-05,
+      "loss": 0.3614,
+      "step": 24982
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1609826541064263e-05,
+      "loss": 0.4852,
+      "step": 24983
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1609490705757898e-05,
+      "loss": 0.3566,
+      "step": 24984
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1609154870451531e-05,
+      "loss": 0.186,
+      "step": 24985
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1608819035145167e-05,
+      "loss": 0.3221,
+      "step": 24986
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.16084831998388e-05,
+      "loss": 0.5474,
+      "step": 24987
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1608147364532433e-05,
+      "loss": 0.3034,
+      "step": 24988
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.160781152922607e-05,
+      "loss": 0.3467,
+      "step": 24989
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1607475693919703e-05,
+      "loss": 0.4546,
+      "step": 24990
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1607139858613337e-05,
+      "loss": 0.3768,
+      "step": 24991
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1606804023306972e-05,
+      "loss": 0.2705,
+      "step": 24992
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1606468188000606e-05,
+      "loss": 0.2931,
+      "step": 24993
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.160613235269424e-05,
+      "loss": 0.5764,
+      "step": 24994
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1605796517387872e-05,
+      "loss": 0.2774,
+      "step": 24995
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1605460682081509e-05,
+      "loss": 0.3256,
+      "step": 24996
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1605124846775142e-05,
+      "loss": 0.2525,
+      "step": 24997
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1604789011468776e-05,
+      "loss": 0.4517,
+      "step": 24998
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1604453176162411e-05,
+      "loss": 0.3567,
+      "step": 24999
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1604117340856046e-05,
+      "loss": 0.3462,
+      "step": 25000
+    },
+    {
+      "epoch": 1.26,
+      "eval_accuracy": 0.8792637324693834,
+      "eval_f1_microaverage": 0.8792637324693835,
+      "eval_loss": 0.3144929111003876,
+      "eval_precision_macroaverage": 0.7532877153730028,
+      "eval_precision_microaverage": 0.8792637324693834,
+      "eval_recall_macroaverage": 0.6680223123163342,
+      "eval_recall_microaverage": 0.8792637324693834,
+      "eval_runtime": 1173.9575,
+      "eval_samples_per_second": 231.898,
+      "eval_steps_per_second": 3.624,
+      "step": 25000
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1603781505549679e-05,
+      "loss": 0.2071,
+      "step": 25001
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1603445670243313e-05,
+      "loss": 0.2024,
+      "step": 25002
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1603109834936948e-05,
+      "loss": 0.2275,
+      "step": 25003
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.160277399963058e-05,
+      "loss": 0.4118,
+      "step": 25004
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1602438164324217e-05,
+      "loss": 0.2323,
+      "step": 25005
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.160210232901785e-05,
+      "loss": 0.2115,
+      "step": 25006
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1601766493711483e-05,
+      "loss": 0.2287,
+      "step": 25007
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.160143065840512e-05,
+      "loss": 0.3966,
+      "step": 25008
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1601094823098752e-05,
+      "loss": 0.4202,
+      "step": 25009
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1600758987792387e-05,
+      "loss": 0.5053,
+      "step": 25010
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1600423152486022e-05,
+      "loss": 0.361,
+      "step": 25011
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1600087317179656e-05,
+      "loss": 0.2976,
+      "step": 25012
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.159975148187329e-05,
+      "loss": 0.3198,
+      "step": 25013
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1599415646566926e-05,
+      "loss": 0.2776,
+      "step": 25014
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1599079811260559e-05,
+      "loss": 0.2454,
+      "step": 25015
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1598743975954191e-05,
+      "loss": 0.5326,
+      "step": 25016
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1598408140647828e-05,
+      "loss": 0.2935,
+      "step": 25017
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.159807230534146e-05,
+      "loss": 0.1746,
+      "step": 25018
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1597736470035095e-05,
+      "loss": 0.2591,
+      "step": 25019
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.159740063472873e-05,
+      "loss": 0.4514,
+      "step": 25020
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1597064799422365e-05,
+      "loss": 0.2386,
+      "step": 25021
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1596728964115998e-05,
+      "loss": 0.2022,
+      "step": 25022
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1596393128809634e-05,
+      "loss": 0.5115,
+      "step": 25023
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1596057293503267e-05,
+      "loss": 0.2274,
+      "step": 25024
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.15957214581969e-05,
+      "loss": 0.4808,
+      "step": 25025
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1595385622890536e-05,
+      "loss": 0.3686,
+      "step": 25026
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.159504978758417e-05,
+      "loss": 0.3975,
+      "step": 25027
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1594713952277802e-05,
+      "loss": 0.1884,
+      "step": 25028
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1594378116971439e-05,
+      "loss": 0.1505,
+      "step": 25029
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1594042281665071e-05,
+      "loss": 0.3416,
+      "step": 25030
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1593706446358706e-05,
+      "loss": 0.26,
+      "step": 25031
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.159337061105234e-05,
+      "loss": 0.4079,
+      "step": 25032
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1593034775745975e-05,
+      "loss": 0.3196,
+      "step": 25033
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1592698940439608e-05,
+      "loss": 0.3641,
+      "step": 25034
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1592363105133245e-05,
+      "loss": 0.2782,
+      "step": 25035
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1592027269826878e-05,
+      "loss": 0.1904,
+      "step": 25036
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.159169143452051e-05,
+      "loss": 0.3648,
+      "step": 25037
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1591355599214147e-05,
+      "loss": 0.3421,
+      "step": 25038
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.159101976390778e-05,
+      "loss": 0.2091,
+      "step": 25039
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1590683928601415e-05,
+      "loss": 0.2817,
+      "step": 25040
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.159034809329505e-05,
+      "loss": 0.347,
+      "step": 25041
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1590012257988684e-05,
+      "loss": 0.24,
+      "step": 25042
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1589676422682317e-05,
+      "loss": 0.2755,
+      "step": 25043
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1589340587375953e-05,
+      "loss": 0.3634,
+      "step": 25044
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1589004752069586e-05,
+      "loss": 0.2322,
+      "step": 25045
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1588668916763219e-05,
+      "loss": 0.3061,
+      "step": 25046
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1588333081456855e-05,
+      "loss": 0.2521,
+      "step": 25047
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1587997246150488e-05,
+      "loss": 0.2975,
+      "step": 25048
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1587661410844121e-05,
+      "loss": 0.3194,
+      "step": 25049
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1587325575537758e-05,
+      "loss": 0.2933,
+      "step": 25050
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.158698974023139e-05,
+      "loss": 0.1762,
+      "step": 25051
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1586653904925025e-05,
+      "loss": 0.2897,
+      "step": 25052
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.158631806961866e-05,
+      "loss": 0.1362,
+      "step": 25053
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1585982234312295e-05,
+      "loss": 0.297,
+      "step": 25054
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1585646399005927e-05,
+      "loss": 0.1303,
+      "step": 25055
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1585310563699564e-05,
+      "loss": 0.2127,
+      "step": 25056
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1584974728393197e-05,
+      "loss": 0.3899,
+      "step": 25057
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.158463889308683e-05,
+      "loss": 0.3529,
+      "step": 25058
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1584303057780466e-05,
+      "loss": 0.2144,
+      "step": 25059
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1583967222474099e-05,
+      "loss": 0.2438,
+      "step": 25060
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1583631387167734e-05,
+      "loss": 0.218,
+      "step": 25061
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1583295551861368e-05,
+      "loss": 0.4744,
+      "step": 25062
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1582959716555003e-05,
+      "loss": 0.2427,
+      "step": 25063
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1582623881248636e-05,
+      "loss": 0.2057,
+      "step": 25064
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1582288045942272e-05,
+      "loss": 0.3981,
+      "step": 25065
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1581952210635905e-05,
+      "loss": 0.3198,
+      "step": 25066
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1581616375329538e-05,
+      "loss": 0.3098,
+      "step": 25067
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1581280540023175e-05,
+      "loss": 0.246,
+      "step": 25068
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1580944704716807e-05,
+      "loss": 0.3515,
+      "step": 25069
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1580608869410442e-05,
+      "loss": 0.3676,
+      "step": 25070
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1580273034104077e-05,
+      "loss": 0.1791,
+      "step": 25071
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.157993719879771e-05,
+      "loss": 0.3192,
+      "step": 25072
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1579601363491344e-05,
+      "loss": 0.3049,
+      "step": 25073
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1579265528184979e-05,
+      "loss": 0.2092,
+      "step": 25074
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1578929692878614e-05,
+      "loss": 0.3145,
+      "step": 25075
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1578593857572247e-05,
+      "loss": 0.3023,
+      "step": 25076
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1578258022265883e-05,
+      "loss": 0.266,
+      "step": 25077
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1577922186959516e-05,
+      "loss": 0.4091,
+      "step": 25078
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1577586351653149e-05,
+      "loss": 0.333,
+      "step": 25079
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1577250516346785e-05,
+      "loss": 0.3884,
+      "step": 25080
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1576914681040418e-05,
+      "loss": 0.3027,
+      "step": 25081
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1576578845734053e-05,
+      "loss": 0.3013,
+      "step": 25082
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1576243010427687e-05,
+      "loss": 0.2451,
+      "step": 25083
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1575907175121322e-05,
+      "loss": 0.4262,
+      "step": 25084
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1575571339814955e-05,
+      "loss": 0.3047,
+      "step": 25085
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1575235504508591e-05,
+      "loss": 0.2112,
+      "step": 25086
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1574899669202224e-05,
+      "loss": 0.2712,
+      "step": 25087
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1574563833895857e-05,
+      "loss": 0.2608,
+      "step": 25088
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1574227998589494e-05,
+      "loss": 0.4667,
+      "step": 25089
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1573892163283127e-05,
+      "loss": 0.2417,
+      "step": 25090
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1573556327976761e-05,
+      "loss": 0.2099,
+      "step": 25091
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1573220492670396e-05,
+      "loss": 0.37,
+      "step": 25092
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1572884657364029e-05,
+      "loss": 0.1634,
+      "step": 25093
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1572548822057663e-05,
+      "loss": 0.2754,
+      "step": 25094
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1572212986751298e-05,
+      "loss": 0.1728,
+      "step": 25095
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1571877151444933e-05,
+      "loss": 0.3065,
+      "step": 25096
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1571541316138566e-05,
+      "loss": 0.3552,
+      "step": 25097
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1571205480832202e-05,
+      "loss": 0.2793,
+      "step": 25098
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1570869645525835e-05,
+      "loss": 0.1557,
+      "step": 25099
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1570533810219468e-05,
+      "loss": 0.1597,
+      "step": 25100
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1570197974913104e-05,
+      "loss": 0.2855,
+      "step": 25101
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1569862139606737e-05,
+      "loss": 0.3505,
+      "step": 25102
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1569526304300372e-05,
+      "loss": 0.3906,
+      "step": 25103
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1569190468994007e-05,
+      "loss": 0.2333,
+      "step": 25104
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1568854633687641e-05,
+      "loss": 0.3656,
+      "step": 25105
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1568518798381274e-05,
+      "loss": 0.5342,
+      "step": 25106
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.156818296307491e-05,
+      "loss": 0.1639,
+      "step": 25107
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1567847127768543e-05,
+      "loss": 0.4422,
+      "step": 25108
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1567511292462176e-05,
+      "loss": 0.4214,
+      "step": 25109
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1567175457155813e-05,
+      "loss": 0.2621,
+      "step": 25110
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 1.1566839621849446e-05,
+      "loss": 0.5394,
+      "step": 25111
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.156650378654308e-05,
+      "loss": 0.2731,
+      "step": 25112
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1566167951236715e-05,
+      "loss": 0.3898,
+      "step": 25113
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.156583211593035e-05,
+      "loss": 0.3247,
+      "step": 25114
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1565496280623983e-05,
+      "loss": 0.3006,
+      "step": 25115
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1565160445317617e-05,
+      "loss": 0.3074,
+      "step": 25116
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1564824610011252e-05,
+      "loss": 0.2325,
+      "step": 25117
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1564488774704885e-05,
+      "loss": 0.3617,
+      "step": 25118
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1564152939398521e-05,
+      "loss": 0.2747,
+      "step": 25119
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1563817104092154e-05,
+      "loss": 0.3519,
+      "step": 25120
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1563481268785787e-05,
+      "loss": 0.3252,
+      "step": 25121
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1563145433479423e-05,
+      "loss": 0.139,
+      "step": 25122
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1562809598173056e-05,
+      "loss": 0.1978,
+      "step": 25123
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1562473762866691e-05,
+      "loss": 0.3289,
+      "step": 25124
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1562137927560326e-05,
+      "loss": 0.2324,
+      "step": 25125
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.156180209225396e-05,
+      "loss": 0.2692,
+      "step": 25126
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1561466256947593e-05,
+      "loss": 0.3201,
+      "step": 25127
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.156113042164123e-05,
+      "loss": 0.2402,
+      "step": 25128
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1560794586334863e-05,
+      "loss": 0.3439,
+      "step": 25129
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1560458751028496e-05,
+      "loss": 0.3564,
+      "step": 25130
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1560122915722132e-05,
+      "loss": 0.3837,
+      "step": 25131
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1559787080415765e-05,
+      "loss": 0.2499,
+      "step": 25132
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.15594512451094e-05,
+      "loss": 0.3029,
+      "step": 25133
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1559115409803034e-05,
+      "loss": 0.3972,
+      "step": 25134
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1558779574496669e-05,
+      "loss": 0.3513,
+      "step": 25135
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1558443739190302e-05,
+      "loss": 0.3983,
+      "step": 25136
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1558107903883936e-05,
+      "loss": 0.1612,
+      "step": 25137
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1557772068577571e-05,
+      "loss": 0.5282,
+      "step": 25138
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1557436233271204e-05,
+      "loss": 0.2762,
+      "step": 25139
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.155710039796484e-05,
+      "loss": 0.2363,
+      "step": 25140
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1556764562658473e-05,
+      "loss": 0.1816,
+      "step": 25141
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1556428727352106e-05,
+      "loss": 0.3279,
+      "step": 25142
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1556092892045743e-05,
+      "loss": 0.3757,
+      "step": 25143
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1555757056739376e-05,
+      "loss": 0.1786,
+      "step": 25144
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.155542122143301e-05,
+      "loss": 0.2545,
+      "step": 25145
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1555085386126645e-05,
+      "loss": 0.3543,
+      "step": 25146
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.155474955082028e-05,
+      "loss": 0.2977,
+      "step": 25147
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1554413715513912e-05,
+      "loss": 0.214,
+      "step": 25148
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1554077880207549e-05,
+      "loss": 0.4024,
+      "step": 25149
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1553742044901182e-05,
+      "loss": 0.4104,
+      "step": 25150
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1553406209594815e-05,
+      "loss": 0.2301,
+      "step": 25151
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1553070374288451e-05,
+      "loss": 0.1435,
+      "step": 25152
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1552734538982084e-05,
+      "loss": 0.2637,
+      "step": 25153
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1552398703675719e-05,
+      "loss": 0.308,
+      "step": 25154
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1552062868369353e-05,
+      "loss": 0.3048,
+      "step": 25155
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1551727033062988e-05,
+      "loss": 0.2082,
+      "step": 25156
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1551391197756621e-05,
+      "loss": 0.2805,
+      "step": 25157
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1551055362450257e-05,
+      "loss": 0.2439,
+      "step": 25158
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.155071952714389e-05,
+      "loss": 0.1493,
+      "step": 25159
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1550383691837523e-05,
+      "loss": 0.4032,
+      "step": 25160
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.155004785653116e-05,
+      "loss": 0.2772,
+      "step": 25161
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1549712021224792e-05,
+      "loss": 0.2631,
+      "step": 25162
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1549376185918425e-05,
+      "loss": 0.2801,
+      "step": 25163
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1549040350612062e-05,
+      "loss": 0.282,
+      "step": 25164
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1548704515305695e-05,
+      "loss": 0.188,
+      "step": 25165
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.154836867999933e-05,
+      "loss": 0.2214,
+      "step": 25166
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1548032844692964e-05,
+      "loss": 0.2069,
+      "step": 25167
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1547697009386599e-05,
+      "loss": 0.1341,
+      "step": 25168
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1547361174080232e-05,
+      "loss": 0.3633,
+      "step": 25169
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1547025338773868e-05,
+      "loss": 0.1367,
+      "step": 25170
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.15466895034675e-05,
+      "loss": 0.1633,
+      "step": 25171
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1546353668161134e-05,
+      "loss": 0.3294,
+      "step": 25172
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.154601783285477e-05,
+      "loss": 0.2933,
+      "step": 25173
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1545681997548403e-05,
+      "loss": 0.1668,
+      "step": 25174
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1545346162242038e-05,
+      "loss": 0.3671,
+      "step": 25175
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.154501032693567e-05,
+      "loss": 0.2086,
+      "step": 25176
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1544674491629307e-05,
+      "loss": 0.24,
+      "step": 25177
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.154433865632294e-05,
+      "loss": 0.3577,
+      "step": 25178
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1544002821016573e-05,
+      "loss": 0.2471,
+      "step": 25179
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.154366698571021e-05,
+      "loss": 0.4071,
+      "step": 25180
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1543331150403842e-05,
+      "loss": 0.2777,
+      "step": 25181
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1542995315097477e-05,
+      "loss": 0.3664,
+      "step": 25182
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1542659479791112e-05,
+      "loss": 0.226,
+      "step": 25183
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1542323644484744e-05,
+      "loss": 0.2433,
+      "step": 25184
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1541987809178379e-05,
+      "loss": 0.3491,
+      "step": 25185
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1541651973872014e-05,
+      "loss": 0.2895,
+      "step": 25186
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1541316138565648e-05,
+      "loss": 0.3425,
+      "step": 25187
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1540980303259281e-05,
+      "loss": 0.3786,
+      "step": 25188
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1540644467952918e-05,
+      "loss": 0.146,
+      "step": 25189
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.154030863264655e-05,
+      "loss": 0.5026,
+      "step": 25190
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1539972797340184e-05,
+      "loss": 0.1651,
+      "step": 25191
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.153963696203382e-05,
+      "loss": 0.1925,
+      "step": 25192
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1539301126727453e-05,
+      "loss": 0.3445,
+      "step": 25193
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1538965291421088e-05,
+      "loss": 0.3032,
+      "step": 25194
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1538629456114722e-05,
+      "loss": 0.2604,
+      "step": 25195
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1538293620808357e-05,
+      "loss": 0.2849,
+      "step": 25196
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.153795778550199e-05,
+      "loss": 0.3299,
+      "step": 25197
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1537621950195626e-05,
+      "loss": 0.2075,
+      "step": 25198
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1537286114889259e-05,
+      "loss": 0.2058,
+      "step": 25199
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1536950279582892e-05,
+      "loss": 0.3739,
+      "step": 25200
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1536614444276528e-05,
+      "loss": 0.4635,
+      "step": 25201
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1536278608970161e-05,
+      "loss": 0.3027,
+      "step": 25202
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1535942773663796e-05,
+      "loss": 0.2955,
+      "step": 25203
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.153560693835743e-05,
+      "loss": 0.2728,
+      "step": 25204
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1535271103051065e-05,
+      "loss": 0.449,
+      "step": 25205
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1534935267744698e-05,
+      "loss": 0.2678,
+      "step": 25206
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1534599432438333e-05,
+      "loss": 0.231,
+      "step": 25207
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1534263597131968e-05,
+      "loss": 0.2437,
+      "step": 25208
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.15339277618256e-05,
+      "loss": 0.3429,
+      "step": 25209
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1533591926519237e-05,
+      "loss": 0.316,
+      "step": 25210
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.153325609121287e-05,
+      "loss": 0.4315,
+      "step": 25211
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1532920255906503e-05,
+      "loss": 0.2656,
+      "step": 25212
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1532584420600139e-05,
+      "loss": 0.2199,
+      "step": 25213
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1532248585293772e-05,
+      "loss": 0.256,
+      "step": 25214
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1531912749987407e-05,
+      "loss": 0.2784,
+      "step": 25215
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1531576914681041e-05,
+      "loss": 0.2863,
+      "step": 25216
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1531241079374676e-05,
+      "loss": 0.15,
+      "step": 25217
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1530905244068309e-05,
+      "loss": 0.2779,
+      "step": 25218
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1530569408761945e-05,
+      "loss": 0.269,
+      "step": 25219
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1530233573455578e-05,
+      "loss": 0.3271,
+      "step": 25220
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1529897738149211e-05,
+      "loss": 0.2676,
+      "step": 25221
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1529561902842848e-05,
+      "loss": 0.2313,
+      "step": 25222
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.152922606753648e-05,
+      "loss": 0.3245,
+      "step": 25223
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1528890232230115e-05,
+      "loss": 0.2542,
+      "step": 25224
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.152855439692375e-05,
+      "loss": 0.3661,
+      "step": 25225
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1528218561617384e-05,
+      "loss": 0.2048,
+      "step": 25226
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1527882726311017e-05,
+      "loss": 0.401,
+      "step": 25227
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1527546891004652e-05,
+      "loss": 0.2632,
+      "step": 25228
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1527211055698287e-05,
+      "loss": 0.1763,
+      "step": 25229
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.152687522039192e-05,
+      "loss": 0.2396,
+      "step": 25230
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1526539385085556e-05,
+      "loss": 0.4473,
+      "step": 25231
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1526203549779189e-05,
+      "loss": 0.3438,
+      "step": 25232
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1525867714472822e-05,
+      "loss": 0.2836,
+      "step": 25233
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1525531879166458e-05,
+      "loss": 0.2144,
+      "step": 25234
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1525196043860091e-05,
+      "loss": 0.3256,
+      "step": 25235
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1524860208553726e-05,
+      "loss": 0.4671,
+      "step": 25236
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.152452437324736e-05,
+      "loss": 0.3221,
+      "step": 25237
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1524188537940995e-05,
+      "loss": 0.4633,
+      "step": 25238
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1523852702634628e-05,
+      "loss": 0.162,
+      "step": 25239
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1523516867328264e-05,
+      "loss": 0.3329,
+      "step": 25240
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1523181032021897e-05,
+      "loss": 0.3345,
+      "step": 25241
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.152284519671553e-05,
+      "loss": 0.4011,
+      "step": 25242
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1522509361409167e-05,
+      "loss": 0.2094,
+      "step": 25243
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.15221735261028e-05,
+      "loss": 0.1783,
+      "step": 25244
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1521837690796434e-05,
+      "loss": 0.2779,
+      "step": 25245
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1521501855490069e-05,
+      "loss": 0.3189,
+      "step": 25246
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1521166020183704e-05,
+      "loss": 0.4209,
+      "step": 25247
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1520830184877337e-05,
+      "loss": 0.3803,
+      "step": 25248
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1520494349570973e-05,
+      "loss": 0.5348,
+      "step": 25249
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1520158514264606e-05,
+      "loss": 0.3113,
+      "step": 25250
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1519822678958239e-05,
+      "loss": 0.1689,
+      "step": 25251
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1519486843651875e-05,
+      "loss": 0.306,
+      "step": 25252
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1519151008345508e-05,
+      "loss": 0.2596,
+      "step": 25253
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1518815173039141e-05,
+      "loss": 0.3427,
+      "step": 25254
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1518479337732777e-05,
+      "loss": 0.2194,
+      "step": 25255
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.151814350242641e-05,
+      "loss": 0.1549,
+      "step": 25256
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1517807667120045e-05,
+      "loss": 0.3236,
+      "step": 25257
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.151747183181368e-05,
+      "loss": 0.2711,
+      "step": 25258
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1517135996507314e-05,
+      "loss": 0.2859,
+      "step": 25259
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1516800161200947e-05,
+      "loss": 0.1913,
+      "step": 25260
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1516464325894584e-05,
+      "loss": 0.2091,
+      "step": 25261
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1516128490588216e-05,
+      "loss": 0.3184,
+      "step": 25262
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.151579265528185e-05,
+      "loss": 0.3198,
+      "step": 25263
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1515456819975486e-05,
+      "loss": 0.238,
+      "step": 25264
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1515120984669119e-05,
+      "loss": 0.0979,
+      "step": 25265
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1514785149362753e-05,
+      "loss": 0.1658,
+      "step": 25266
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1514449314056388e-05,
+      "loss": 0.3288,
+      "step": 25267
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1514113478750023e-05,
+      "loss": 0.2429,
+      "step": 25268
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1513777643443656e-05,
+      "loss": 0.2529,
+      "step": 25269
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1513441808137292e-05,
+      "loss": 0.5148,
+      "step": 25270
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1513105972830925e-05,
+      "loss": 0.3248,
+      "step": 25271
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1512770137524558e-05,
+      "loss": 0.4204,
+      "step": 25272
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1512434302218194e-05,
+      "loss": 0.3432,
+      "step": 25273
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1512098466911827e-05,
+      "loss": 0.2136,
+      "step": 25274
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.151176263160546e-05,
+      "loss": 0.2428,
+      "step": 25275
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1511426796299096e-05,
+      "loss": 0.2635,
+      "step": 25276
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.151109096099273e-05,
+      "loss": 0.2699,
+      "step": 25277
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1510755125686364e-05,
+      "loss": 0.2775,
+      "step": 25278
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1510419290379999e-05,
+      "loss": 0.25,
+      "step": 25279
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1510083455073633e-05,
+      "loss": 0.2693,
+      "step": 25280
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1509747619767266e-05,
+      "loss": 0.3605,
+      "step": 25281
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1509411784460903e-05,
+      "loss": 0.2806,
+      "step": 25282
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1509075949154536e-05,
+      "loss": 0.288,
+      "step": 25283
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1508740113848169e-05,
+      "loss": 0.243,
+      "step": 25284
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1508404278541805e-05,
+      "loss": 0.3799,
+      "step": 25285
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1508068443235438e-05,
+      "loss": 0.3824,
+      "step": 25286
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1507732607929073e-05,
+      "loss": 0.3,
+      "step": 25287
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1507396772622707e-05,
+      "loss": 0.2606,
+      "step": 25288
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1507060937316342e-05,
+      "loss": 0.3459,
+      "step": 25289
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1506725102009975e-05,
+      "loss": 0.2417,
+      "step": 25290
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1506389266703611e-05,
+      "loss": 0.2319,
+      "step": 25291
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1506053431397244e-05,
+      "loss": 0.1947,
+      "step": 25292
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1505717596090877e-05,
+      "loss": 0.2537,
+      "step": 25293
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1505381760784513e-05,
+      "loss": 0.2245,
+      "step": 25294
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1505045925478146e-05,
+      "loss": 0.4144,
+      "step": 25295
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.150471009017178e-05,
+      "loss": 0.2119,
+      "step": 25296
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1504374254865416e-05,
+      "loss": 0.2222,
+      "step": 25297
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1504038419559049e-05,
+      "loss": 0.2453,
+      "step": 25298
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1503702584252683e-05,
+      "loss": 0.2958,
+      "step": 25299
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1503366748946318e-05,
+      "loss": 0.2756,
+      "step": 25300
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1503030913639952e-05,
+      "loss": 0.3923,
+      "step": 25301
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1502695078333585e-05,
+      "loss": 0.3352,
+      "step": 25302
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1502359243027222e-05,
+      "loss": 0.1504,
+      "step": 25303
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1502023407720855e-05,
+      "loss": 0.2602,
+      "step": 25304
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1501687572414488e-05,
+      "loss": 0.2416,
+      "step": 25305
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1501351737108124e-05,
+      "loss": 0.2259,
+      "step": 25306
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1501015901801757e-05,
+      "loss": 0.4801,
+      "step": 25307
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1500680066495392e-05,
+      "loss": 0.35,
+      "step": 25308
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1500344231189026e-05,
+      "loss": 0.2493,
+      "step": 25309
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 1.1500008395882661e-05,
+      "loss": 0.2338,
+      "step": 25310
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1499672560576294e-05,
+      "loss": 0.2287,
+      "step": 25311
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.149933672526993e-05,
+      "loss": 0.4457,
+      "step": 25312
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1499000889963563e-05,
+      "loss": 0.435,
+      "step": 25313
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1498665054657196e-05,
+      "loss": 0.297,
+      "step": 25314
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1498329219350832e-05,
+      "loss": 0.3311,
+      "step": 25315
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1497993384044465e-05,
+      "loss": 0.2253,
+      "step": 25316
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.14976575487381e-05,
+      "loss": 0.3157,
+      "step": 25317
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1497321713431735e-05,
+      "loss": 0.2849,
+      "step": 25318
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1496985878125368e-05,
+      "loss": 0.1313,
+      "step": 25319
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1496650042819002e-05,
+      "loss": 0.2574,
+      "step": 25320
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1496314207512637e-05,
+      "loss": 0.2903,
+      "step": 25321
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1495978372206272e-05,
+      "loss": 0.3586,
+      "step": 25322
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1495642536899905e-05,
+      "loss": 0.285,
+      "step": 25323
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1495306701593541e-05,
+      "loss": 0.3504,
+      "step": 25324
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1494970866287174e-05,
+      "loss": 0.3876,
+      "step": 25325
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1494635030980807e-05,
+      "loss": 0.3772,
+      "step": 25326
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1494299195674443e-05,
+      "loss": 0.3779,
+      "step": 25327
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1493963360368076e-05,
+      "loss": 0.3201,
+      "step": 25328
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.149362752506171e-05,
+      "loss": 0.3373,
+      "step": 25329
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1493291689755345e-05,
+      "loss": 0.2839,
+      "step": 25330
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.149295585444898e-05,
+      "loss": 0.3328,
+      "step": 25331
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1492620019142613e-05,
+      "loss": 0.2317,
+      "step": 25332
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.149228418383625e-05,
+      "loss": 0.3075,
+      "step": 25333
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1491948348529882e-05,
+      "loss": 0.3285,
+      "step": 25334
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1491612513223515e-05,
+      "loss": 0.2577,
+      "step": 25335
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1491276677917152e-05,
+      "loss": 0.215,
+      "step": 25336
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1490940842610785e-05,
+      "loss": 0.2994,
+      "step": 25337
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.149060500730442e-05,
+      "loss": 0.3494,
+      "step": 25338
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1490269171998054e-05,
+      "loss": 0.3135,
+      "step": 25339
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1489933336691688e-05,
+      "loss": 0.1612,
+      "step": 25340
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1489597501385321e-05,
+      "loss": 0.4213,
+      "step": 25341
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1489261666078956e-05,
+      "loss": 0.5051,
+      "step": 25342
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.148892583077259e-05,
+      "loss": 0.2984,
+      "step": 25343
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1488589995466224e-05,
+      "loss": 0.3101,
+      "step": 25344
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.148825416015986e-05,
+      "loss": 0.2872,
+      "step": 25345
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1487918324853493e-05,
+      "loss": 0.2373,
+      "step": 25346
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1487582489547126e-05,
+      "loss": 0.1478,
+      "step": 25347
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1487246654240762e-05,
+      "loss": 0.2601,
+      "step": 25348
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1486910818934395e-05,
+      "loss": 0.3181,
+      "step": 25349
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.148657498362803e-05,
+      "loss": 0.4733,
+      "step": 25350
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1486239148321665e-05,
+      "loss": 0.2611,
+      "step": 25351
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.14859033130153e-05,
+      "loss": 0.2687,
+      "step": 25352
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1485567477708932e-05,
+      "loss": 0.4308,
+      "step": 25353
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1485231642402568e-05,
+      "loss": 0.3408,
+      "step": 25354
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1484895807096201e-05,
+      "loss": 0.332,
+      "step": 25355
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1484559971789834e-05,
+      "loss": 0.3961,
+      "step": 25356
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1484224136483469e-05,
+      "loss": 0.3396,
+      "step": 25357
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1483888301177104e-05,
+      "loss": 0.2912,
+      "step": 25358
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1483552465870738e-05,
+      "loss": 0.254,
+      "step": 25359
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1483216630564371e-05,
+      "loss": 0.2406,
+      "step": 25360
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1482880795258008e-05,
+      "loss": 0.2018,
+      "step": 25361
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.148254495995164e-05,
+      "loss": 0.381,
+      "step": 25362
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1482209124645274e-05,
+      "loss": 0.2886,
+      "step": 25363
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.148187328933891e-05,
+      "loss": 0.1973,
+      "step": 25364
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1481537454032543e-05,
+      "loss": 0.3266,
+      "step": 25365
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1481201618726176e-05,
+      "loss": 0.2644,
+      "step": 25366
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1480865783419812e-05,
+      "loss": 0.4034,
+      "step": 25367
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1480529948113445e-05,
+      "loss": 0.3581,
+      "step": 25368
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.148019411280708e-05,
+      "loss": 0.3511,
+      "step": 25369
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1479858277500714e-05,
+      "loss": 0.2162,
+      "step": 25370
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1479522442194349e-05,
+      "loss": 0.2202,
+      "step": 25371
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1479186606887982e-05,
+      "loss": 0.2013,
+      "step": 25372
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1478850771581618e-05,
+      "loss": 0.4914,
+      "step": 25373
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1478514936275251e-05,
+      "loss": 0.289,
+      "step": 25374
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1478179100968884e-05,
+      "loss": 0.2827,
+      "step": 25375
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.147784326566252e-05,
+      "loss": 0.2172,
+      "step": 25376
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1477507430356153e-05,
+      "loss": 0.2693,
+      "step": 25377
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1477171595049788e-05,
+      "loss": 0.3184,
+      "step": 25378
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1476835759743423e-05,
+      "loss": 0.2579,
+      "step": 25379
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1476499924437057e-05,
+      "loss": 0.3872,
+      "step": 25380
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.147616408913069e-05,
+      "loss": 0.3046,
+      "step": 25381
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1475828253824327e-05,
+      "loss": 0.1326,
+      "step": 25382
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.147549241851796e-05,
+      "loss": 0.3359,
+      "step": 25383
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1475156583211593e-05,
+      "loss": 0.2216,
+      "step": 25384
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1474820747905229e-05,
+      "loss": 0.3144,
+      "step": 25385
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1474484912598862e-05,
+      "loss": 0.2952,
+      "step": 25386
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1474149077292495e-05,
+      "loss": 0.2373,
+      "step": 25387
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1473813241986131e-05,
+      "loss": 0.3616,
+      "step": 25388
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1473477406679764e-05,
+      "loss": 0.3898,
+      "step": 25389
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1473141571373399e-05,
+      "loss": 0.2335,
+      "step": 25390
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1472805736067033e-05,
+      "loss": 0.3072,
+      "step": 25391
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1472469900760668e-05,
+      "loss": 0.2701,
+      "step": 25392
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1472134065454301e-05,
+      "loss": 0.1954,
+      "step": 25393
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1471798230147937e-05,
+      "loss": 0.3987,
+      "step": 25394
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.147146239484157e-05,
+      "loss": 0.4413,
+      "step": 25395
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1471126559535203e-05,
+      "loss": 0.3569,
+      "step": 25396
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.147079072422884e-05,
+      "loss": 0.3979,
+      "step": 25397
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1470454888922473e-05,
+      "loss": 0.3502,
+      "step": 25398
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1470119053616107e-05,
+      "loss": 0.2985,
+      "step": 25399
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1469783218309742e-05,
+      "loss": 0.2111,
+      "step": 25400
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1469447383003377e-05,
+      "loss": 0.377,
+      "step": 25401
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.146911154769701e-05,
+      "loss": 0.3479,
+      "step": 25402
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1468775712390646e-05,
+      "loss": 0.6117,
+      "step": 25403
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1468439877084279e-05,
+      "loss": 0.2746,
+      "step": 25404
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1468104041777912e-05,
+      "loss": 0.2705,
+      "step": 25405
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1467768206471548e-05,
+      "loss": 0.268,
+      "step": 25406
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1467432371165181e-05,
+      "loss": 0.4019,
+      "step": 25407
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1467096535858816e-05,
+      "loss": 0.4895,
+      "step": 25408
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.146676070055245e-05,
+      "loss": 0.1911,
+      "step": 25409
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1466424865246083e-05,
+      "loss": 0.3258,
+      "step": 25410
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1466089029939718e-05,
+      "loss": 0.2947,
+      "step": 25411
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1465753194633353e-05,
+      "loss": 0.3206,
+      "step": 25412
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1465417359326987e-05,
+      "loss": 0.2594,
+      "step": 25413
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.146508152402062e-05,
+      "loss": 0.1581,
+      "step": 25414
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1464745688714257e-05,
+      "loss": 0.2827,
+      "step": 25415
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.146440985340789e-05,
+      "loss": 0.1795,
+      "step": 25416
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1464074018101522e-05,
+      "loss": 0.2619,
+      "step": 25417
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1463738182795159e-05,
+      "loss": 0.2148,
+      "step": 25418
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1463402347488792e-05,
+      "loss": 0.3484,
+      "step": 25419
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1463066512182426e-05,
+      "loss": 0.2161,
+      "step": 25420
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1462730676876061e-05,
+      "loss": 0.154,
+      "step": 25421
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1462394841569696e-05,
+      "loss": 0.304,
+      "step": 25422
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1462059006263329e-05,
+      "loss": 0.3238,
+      "step": 25423
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1461723170956965e-05,
+      "loss": 0.1565,
+      "step": 25424
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1461387335650598e-05,
+      "loss": 0.2838,
+      "step": 25425
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1461051500344231e-05,
+      "loss": 0.328,
+      "step": 25426
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1460715665037867e-05,
+      "loss": 0.1548,
+      "step": 25427
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.14603798297315e-05,
+      "loss": 0.2355,
+      "step": 25428
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1460043994425135e-05,
+      "loss": 0.5187,
+      "step": 25429
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.145970815911877e-05,
+      "loss": 0.2353,
+      "step": 25430
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1459372323812402e-05,
+      "loss": 0.2766,
+      "step": 25431
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1459036488506037e-05,
+      "loss": 0.2922,
+      "step": 25432
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1458700653199672e-05,
+      "loss": 0.1836,
+      "step": 25433
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1458364817893306e-05,
+      "loss": 0.3083,
+      "step": 25434
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.145802898258694e-05,
+      "loss": 0.2194,
+      "step": 25435
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1457693147280576e-05,
+      "loss": 0.2505,
+      "step": 25436
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1457357311974209e-05,
+      "loss": 0.3174,
+      "step": 25437
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1457021476667842e-05,
+      "loss": 0.4467,
+      "step": 25438
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1456685641361478e-05,
+      "loss": 0.174,
+      "step": 25439
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1456349806055111e-05,
+      "loss": 0.2689,
+      "step": 25440
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1456013970748746e-05,
+      "loss": 0.2673,
+      "step": 25441
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.145567813544238e-05,
+      "loss": 0.3484,
+      "step": 25442
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1455342300136015e-05,
+      "loss": 0.4695,
+      "step": 25443
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1455006464829648e-05,
+      "loss": 0.3172,
+      "step": 25444
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1454670629523284e-05,
+      "loss": 0.273,
+      "step": 25445
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1454334794216917e-05,
+      "loss": 0.252,
+      "step": 25446
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.145399895891055e-05,
+      "loss": 0.2966,
+      "step": 25447
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1453663123604186e-05,
+      "loss": 0.2931,
+      "step": 25448
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.145332728829782e-05,
+      "loss": 0.2383,
+      "step": 25449
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1452991452991454e-05,
+      "loss": 0.3065,
+      "step": 25450
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1452655617685089e-05,
+      "loss": 0.6328,
+      "step": 25451
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1452319782378723e-05,
+      "loss": 0.3806,
+      "step": 25452
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1451983947072356e-05,
+      "loss": 0.2322,
+      "step": 25453
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1451648111765991e-05,
+      "loss": 0.3151,
+      "step": 25454
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1451312276459626e-05,
+      "loss": 0.2631,
+      "step": 25455
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1450976441153258e-05,
+      "loss": 0.4206,
+      "step": 25456
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1450640605846895e-05,
+      "loss": 0.3368,
+      "step": 25457
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1450304770540528e-05,
+      "loss": 0.4669,
+      "step": 25458
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.144996893523416e-05,
+      "loss": 0.3905,
+      "step": 25459
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1449633099927797e-05,
+      "loss": 0.3413,
+      "step": 25460
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.144929726462143e-05,
+      "loss": 0.2978,
+      "step": 25461
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1448961429315065e-05,
+      "loss": 0.3107,
+      "step": 25462
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.14486255940087e-05,
+      "loss": 0.2183,
+      "step": 25463
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1448289758702334e-05,
+      "loss": 0.2839,
+      "step": 25464
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1447953923395967e-05,
+      "loss": 0.2791,
+      "step": 25465
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1447618088089603e-05,
+      "loss": 0.2597,
+      "step": 25466
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1447282252783236e-05,
+      "loss": 0.374,
+      "step": 25467
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1446946417476869e-05,
+      "loss": 0.3383,
+      "step": 25468
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1446610582170505e-05,
+      "loss": 0.3688,
+      "step": 25469
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1446274746864138e-05,
+      "loss": 0.2878,
+      "step": 25470
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1445938911557773e-05,
+      "loss": 0.166,
+      "step": 25471
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1445603076251408e-05,
+      "loss": 0.2828,
+      "step": 25472
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1445267240945042e-05,
+      "loss": 0.3829,
+      "step": 25473
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1444931405638675e-05,
+      "loss": 0.3188,
+      "step": 25474
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1444595570332312e-05,
+      "loss": 0.3385,
+      "step": 25475
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1444259735025945e-05,
+      "loss": 0.4332,
+      "step": 25476
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1443923899719578e-05,
+      "loss": 0.1926,
+      "step": 25477
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1443588064413214e-05,
+      "loss": 0.3045,
+      "step": 25478
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1443252229106847e-05,
+      "loss": 0.3293,
+      "step": 25479
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.144291639380048e-05,
+      "loss": 0.3197,
+      "step": 25480
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1442580558494116e-05,
+      "loss": 0.2655,
+      "step": 25481
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1442244723187749e-05,
+      "loss": 0.3299,
+      "step": 25482
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1441908887881384e-05,
+      "loss": 0.188,
+      "step": 25483
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1441573052575018e-05,
+      "loss": 0.2326,
+      "step": 25484
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1441237217268653e-05,
+      "loss": 0.3218,
+      "step": 25485
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1440901381962286e-05,
+      "loss": 0.2539,
+      "step": 25486
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1440565546655922e-05,
+      "loss": 0.4158,
+      "step": 25487
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1440229711349555e-05,
+      "loss": 0.1999,
+      "step": 25488
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1439893876043188e-05,
+      "loss": 0.3605,
+      "step": 25489
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1439558040736825e-05,
+      "loss": 0.4163,
+      "step": 25490
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1439222205430458e-05,
+      "loss": 0.3078,
+      "step": 25491
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1438886370124092e-05,
+      "loss": 0.4305,
+      "step": 25492
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1438550534817727e-05,
+      "loss": 0.1726,
+      "step": 25493
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1438214699511362e-05,
+      "loss": 0.3514,
+      "step": 25494
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1437878864204994e-05,
+      "loss": 0.1967,
+      "step": 25495
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.143754302889863e-05,
+      "loss": 0.2778,
+      "step": 25496
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1437207193592264e-05,
+      "loss": 0.1828,
+      "step": 25497
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1436871358285897e-05,
+      "loss": 0.3072,
+      "step": 25498
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1436535522979533e-05,
+      "loss": 0.1348,
+      "step": 25499
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1436199687673166e-05,
+      "loss": 0.1602,
+      "step": 25500
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1435863852366799e-05,
+      "loss": 0.1698,
+      "step": 25501
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1435528017060435e-05,
+      "loss": 0.312,
+      "step": 25502
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1435192181754068e-05,
+      "loss": 0.2585,
+      "step": 25503
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1434856346447703e-05,
+      "loss": 0.2983,
+      "step": 25504
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1434520511141338e-05,
+      "loss": 0.1931,
+      "step": 25505
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1434184675834972e-05,
+      "loss": 0.2836,
+      "step": 25506
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1433848840528605e-05,
+      "loss": 0.4674,
+      "step": 25507
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.1433513005222241e-05,
+      "loss": 0.236,
+      "step": 25508
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1433177169915874e-05,
+      "loss": 0.4775,
+      "step": 25509
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1432841334609507e-05,
+      "loss": 0.3489,
+      "step": 25510
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1432505499303144e-05,
+      "loss": 0.3438,
+      "step": 25511
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1432169663996777e-05,
+      "loss": 0.3456,
+      "step": 25512
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1431833828690411e-05,
+      "loss": 0.3602,
+      "step": 25513
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1431497993384046e-05,
+      "loss": 0.1886,
+      "step": 25514
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.143116215807768e-05,
+      "loss": 0.4807,
+      "step": 25515
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1430826322771314e-05,
+      "loss": 0.4193,
+      "step": 25516
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.143049048746495e-05,
+      "loss": 0.2976,
+      "step": 25517
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1430154652158583e-05,
+      "loss": 0.3057,
+      "step": 25518
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1429818816852216e-05,
+      "loss": 0.2807,
+      "step": 25519
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1429482981545852e-05,
+      "loss": 0.3884,
+      "step": 25520
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1429147146239485e-05,
+      "loss": 0.3548,
+      "step": 25521
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1428811310933118e-05,
+      "loss": 0.2286,
+      "step": 25522
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1428475475626754e-05,
+      "loss": 0.2824,
+      "step": 25523
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1428139640320387e-05,
+      "loss": 0.5256,
+      "step": 25524
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1427803805014022e-05,
+      "loss": 0.3495,
+      "step": 25525
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1427467969707657e-05,
+      "loss": 0.3143,
+      "step": 25526
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1427132134401291e-05,
+      "loss": 0.3855,
+      "step": 25527
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1426796299094924e-05,
+      "loss": 0.2617,
+      "step": 25528
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.142646046378856e-05,
+      "loss": 0.2735,
+      "step": 25529
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1426124628482194e-05,
+      "loss": 0.2786,
+      "step": 25530
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1425788793175827e-05,
+      "loss": 0.3801,
+      "step": 25531
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1425452957869463e-05,
+      "loss": 0.3729,
+      "step": 25532
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1425117122563096e-05,
+      "loss": 0.4626,
+      "step": 25533
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.142478128725673e-05,
+      "loss": 0.3223,
+      "step": 25534
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1424445451950365e-05,
+      "loss": 0.3393,
+      "step": 25535
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1424109616644e-05,
+      "loss": 0.2202,
+      "step": 25536
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1423773781337633e-05,
+      "loss": 0.4434,
+      "step": 25537
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1423437946031266e-05,
+      "loss": 0.2255,
+      "step": 25538
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1423102110724902e-05,
+      "loss": 0.2392,
+      "step": 25539
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1422766275418535e-05,
+      "loss": 0.3534,
+      "step": 25540
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.142243044011217e-05,
+      "loss": 0.4257,
+      "step": 25541
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1422094604805804e-05,
+      "loss": 0.2302,
+      "step": 25542
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1421758769499439e-05,
+      "loss": 0.3138,
+      "step": 25543
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1421422934193072e-05,
+      "loss": 0.2398,
+      "step": 25544
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1421087098886706e-05,
+      "loss": 0.2403,
+      "step": 25545
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1420751263580341e-05,
+      "loss": 0.1831,
+      "step": 25546
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1420415428273974e-05,
+      "loss": 0.269,
+      "step": 25547
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.142007959296761e-05,
+      "loss": 0.2452,
+      "step": 25548
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1419743757661243e-05,
+      "loss": 0.3122,
+      "step": 25549
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1419407922354876e-05,
+      "loss": 0.2742,
+      "step": 25550
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1419072087048513e-05,
+      "loss": 0.1913,
+      "step": 25551
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1418736251742146e-05,
+      "loss": 0.1484,
+      "step": 25552
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.141840041643578e-05,
+      "loss": 0.4067,
+      "step": 25553
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1418064581129415e-05,
+      "loss": 0.2132,
+      "step": 25554
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.141772874582305e-05,
+      "loss": 0.3569,
+      "step": 25555
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1417392910516683e-05,
+      "loss": 0.2342,
+      "step": 25556
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1417057075210319e-05,
+      "loss": 0.4219,
+      "step": 25557
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1416721239903952e-05,
+      "loss": 0.2835,
+      "step": 25558
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1416385404597585e-05,
+      "loss": 0.3146,
+      "step": 25559
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1416049569291221e-05,
+      "loss": 0.3371,
+      "step": 25560
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1415713733984854e-05,
+      "loss": 0.2358,
+      "step": 25561
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1415377898678489e-05,
+      "loss": 0.2158,
+      "step": 25562
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1415042063372123e-05,
+      "loss": 0.2047,
+      "step": 25563
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1414706228065758e-05,
+      "loss": 0.2124,
+      "step": 25564
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1414370392759391e-05,
+      "loss": 0.1501,
+      "step": 25565
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1414034557453026e-05,
+      "loss": 0.2842,
+      "step": 25566
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.141369872214666e-05,
+      "loss": 0.2413,
+      "step": 25567
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1413362886840293e-05,
+      "loss": 0.2924,
+      "step": 25568
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.141302705153393e-05,
+      "loss": 0.459,
+      "step": 25569
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1412691216227563e-05,
+      "loss": 0.3307,
+      "step": 25570
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1412355380921195e-05,
+      "loss": 0.2185,
+      "step": 25571
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1412019545614832e-05,
+      "loss": 0.3101,
+      "step": 25572
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1411683710308465e-05,
+      "loss": 0.2241,
+      "step": 25573
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.14113478750021e-05,
+      "loss": 0.435,
+      "step": 25574
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1411012039695734e-05,
+      "loss": 0.3081,
+      "step": 25575
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1410676204389369e-05,
+      "loss": 0.2252,
+      "step": 25576
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1410340369083002e-05,
+      "loss": 0.384,
+      "step": 25577
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1410004533776638e-05,
+      "loss": 0.3109,
+      "step": 25578
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1409668698470271e-05,
+      "loss": 0.2483,
+      "step": 25579
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1409332863163904e-05,
+      "loss": 0.2424,
+      "step": 25580
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.140899702785754e-05,
+      "loss": 0.2107,
+      "step": 25581
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1408661192551173e-05,
+      "loss": 0.3781,
+      "step": 25582
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1408325357244808e-05,
+      "loss": 0.2107,
+      "step": 25583
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1407989521938442e-05,
+      "loss": 0.3014,
+      "step": 25584
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1407653686632077e-05,
+      "loss": 0.2249,
+      "step": 25585
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.140731785132571e-05,
+      "loss": 0.3429,
+      "step": 25586
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1406982016019346e-05,
+      "loss": 0.2334,
+      "step": 25587
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.140664618071298e-05,
+      "loss": 0.3294,
+      "step": 25588
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1406310345406612e-05,
+      "loss": 0.2675,
+      "step": 25589
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1405974510100249e-05,
+      "loss": 0.2959,
+      "step": 25590
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1405638674793882e-05,
+      "loss": 0.1688,
+      "step": 25591
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1405302839487515e-05,
+      "loss": 0.3016,
+      "step": 25592
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1404967004181151e-05,
+      "loss": 0.2567,
+      "step": 25593
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1404631168874784e-05,
+      "loss": 0.35,
+      "step": 25594
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1404295333568419e-05,
+      "loss": 0.363,
+      "step": 25595
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1403959498262053e-05,
+      "loss": 0.4251,
+      "step": 25596
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1403623662955688e-05,
+      "loss": 0.2303,
+      "step": 25597
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.140328782764932e-05,
+      "loss": 0.3411,
+      "step": 25598
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1402951992342957e-05,
+      "loss": 0.3043,
+      "step": 25599
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.140261615703659e-05,
+      "loss": 0.2404,
+      "step": 25600
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1402280321730223e-05,
+      "loss": 0.2835,
+      "step": 25601
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.140194448642386e-05,
+      "loss": 0.2649,
+      "step": 25602
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1401608651117492e-05,
+      "loss": 0.3731,
+      "step": 25603
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1401272815811127e-05,
+      "loss": 0.2937,
+      "step": 25604
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1400936980504762e-05,
+      "loss": 0.2926,
+      "step": 25605
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1400601145198396e-05,
+      "loss": 0.292,
+      "step": 25606
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.140026530989203e-05,
+      "loss": 0.2522,
+      "step": 25607
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1399929474585666e-05,
+      "loss": 0.3666,
+      "step": 25608
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1399593639279299e-05,
+      "loss": 0.263,
+      "step": 25609
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1399257803972931e-05,
+      "loss": 0.4439,
+      "step": 25610
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1398921968666568e-05,
+      "loss": 0.4465,
+      "step": 25611
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.13985861333602e-05,
+      "loss": 0.2584,
+      "step": 25612
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1398250298053834e-05,
+      "loss": 0.3336,
+      "step": 25613
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.139791446274747e-05,
+      "loss": 0.2646,
+      "step": 25614
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1397578627441103e-05,
+      "loss": 0.3546,
+      "step": 25615
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1397242792134738e-05,
+      "loss": 0.4717,
+      "step": 25616
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1396906956828372e-05,
+      "loss": 0.2449,
+      "step": 25617
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1396571121522007e-05,
+      "loss": 0.2006,
+      "step": 25618
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.139623528621564e-05,
+      "loss": 0.3812,
+      "step": 25619
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1395899450909276e-05,
+      "loss": 0.3095,
+      "step": 25620
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.139556361560291e-05,
+      "loss": 0.3132,
+      "step": 25621
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1395227780296542e-05,
+      "loss": 0.3619,
+      "step": 25622
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1394891944990178e-05,
+      "loss": 0.1901,
+      "step": 25623
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1394556109683811e-05,
+      "loss": 0.2354,
+      "step": 25624
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1394220274377446e-05,
+      "loss": 0.3035,
+      "step": 25625
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.139388443907108e-05,
+      "loss": 0.2796,
+      "step": 25626
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1393548603764715e-05,
+      "loss": 0.3967,
+      "step": 25627
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1393212768458348e-05,
+      "loss": 0.2642,
+      "step": 25628
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1392876933151985e-05,
+      "loss": 0.29,
+      "step": 25629
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1392541097845618e-05,
+      "loss": 0.3289,
+      "step": 25630
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.139220526253925e-05,
+      "loss": 0.2044,
+      "step": 25631
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1391869427232887e-05,
+      "loss": 0.3146,
+      "step": 25632
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.139153359192652e-05,
+      "loss": 0.399,
+      "step": 25633
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1391197756620155e-05,
+      "loss": 0.3546,
+      "step": 25634
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.139086192131379e-05,
+      "loss": 0.3503,
+      "step": 25635
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1390526086007422e-05,
+      "loss": 0.3664,
+      "step": 25636
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1390190250701057e-05,
+      "loss": 0.1607,
+      "step": 25637
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1389854415394691e-05,
+      "loss": 0.2448,
+      "step": 25638
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1389518580088326e-05,
+      "loss": 0.2541,
+      "step": 25639
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1389182744781959e-05,
+      "loss": 0.3686,
+      "step": 25640
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1388846909475595e-05,
+      "loss": 0.3804,
+      "step": 25641
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1388511074169228e-05,
+      "loss": 0.3284,
+      "step": 25642
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1388175238862861e-05,
+      "loss": 0.3235,
+      "step": 25643
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1387839403556498e-05,
+      "loss": 0.2594,
+      "step": 25644
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.138750356825013e-05,
+      "loss": 0.1846,
+      "step": 25645
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1387167732943765e-05,
+      "loss": 0.2845,
+      "step": 25646
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.13868318976374e-05,
+      "loss": 0.2846,
+      "step": 25647
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1386496062331035e-05,
+      "loss": 0.3516,
+      "step": 25648
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1386160227024667e-05,
+      "loss": 0.348,
+      "step": 25649
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1385824391718304e-05,
+      "loss": 0.275,
+      "step": 25650
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1385488556411937e-05,
+      "loss": 0.3167,
+      "step": 25651
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.138515272110557e-05,
+      "loss": 0.2449,
+      "step": 25652
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1384816885799206e-05,
+      "loss": 0.1992,
+      "step": 25653
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1384481050492839e-05,
+      "loss": 0.2465,
+      "step": 25654
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1384145215186474e-05,
+      "loss": 0.4224,
+      "step": 25655
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1383809379880108e-05,
+      "loss": 0.2499,
+      "step": 25656
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1383473544573741e-05,
+      "loss": 0.2324,
+      "step": 25657
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1383137709267376e-05,
+      "loss": 0.3732,
+      "step": 25658
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.138280187396101e-05,
+      "loss": 0.2742,
+      "step": 25659
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1382466038654645e-05,
+      "loss": 0.4844,
+      "step": 25660
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1382130203348278e-05,
+      "loss": 0.317,
+      "step": 25661
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1381794368041914e-05,
+      "loss": 0.1697,
+      "step": 25662
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1381458532735547e-05,
+      "loss": 0.3968,
+      "step": 25663
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.138112269742918e-05,
+      "loss": 0.2665,
+      "step": 25664
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1380786862122817e-05,
+      "loss": 0.326,
+      "step": 25665
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.138045102681645e-05,
+      "loss": 0.2548,
+      "step": 25666
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1380115191510084e-05,
+      "loss": 0.2781,
+      "step": 25667
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1379779356203719e-05,
+      "loss": 0.2478,
+      "step": 25668
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1379443520897354e-05,
+      "loss": 0.5993,
+      "step": 25669
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1379107685590987e-05,
+      "loss": 0.3559,
+      "step": 25670
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1378771850284623e-05,
+      "loss": 0.438,
+      "step": 25671
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1378436014978256e-05,
+      "loss": 0.3008,
+      "step": 25672
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1378100179671889e-05,
+      "loss": 0.2767,
+      "step": 25673
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1377764344365525e-05,
+      "loss": 0.427,
+      "step": 25674
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1377428509059158e-05,
+      "loss": 0.2192,
+      "step": 25675
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1377092673752793e-05,
+      "loss": 0.1512,
+      "step": 25676
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1376756838446427e-05,
+      "loss": 0.1584,
+      "step": 25677
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1376421003140062e-05,
+      "loss": 0.2865,
+      "step": 25678
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1376085167833695e-05,
+      "loss": 0.2841,
+      "step": 25679
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.137574933252733e-05,
+      "loss": 0.1639,
+      "step": 25680
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1375413497220964e-05,
+      "loss": 0.3997,
+      "step": 25681
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1375077661914597e-05,
+      "loss": 0.3658,
+      "step": 25682
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1374741826608234e-05,
+      "loss": 0.2741,
+      "step": 25683
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1374405991301867e-05,
+      "loss": 0.2003,
+      "step": 25684
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.13740701559955e-05,
+      "loss": 0.3511,
+      "step": 25685
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1373734320689136e-05,
+      "loss": 0.2685,
+      "step": 25686
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1373398485382769e-05,
+      "loss": 0.3115,
+      "step": 25687
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1373062650076403e-05,
+      "loss": 0.3365,
+      "step": 25688
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1372726814770038e-05,
+      "loss": 0.2495,
+      "step": 25689
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1372390979463673e-05,
+      "loss": 0.3241,
+      "step": 25690
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1372055144157306e-05,
+      "loss": 0.1522,
+      "step": 25691
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1371719308850942e-05,
+      "loss": 0.3221,
+      "step": 25692
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1371383473544575e-05,
+      "loss": 0.3916,
+      "step": 25693
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1371047638238208e-05,
+      "loss": 0.2039,
+      "step": 25694
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1370711802931844e-05,
+      "loss": 0.3433,
+      "step": 25695
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1370375967625477e-05,
+      "loss": 0.2666,
+      "step": 25696
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1370040132319112e-05,
+      "loss": 0.2073,
+      "step": 25697
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1369704297012747e-05,
+      "loss": 0.4467,
+      "step": 25698
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1369368461706381e-05,
+      "loss": 0.4136,
+      "step": 25699
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1369032626400014e-05,
+      "loss": 0.2697,
+      "step": 25700
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1368696791093649e-05,
+      "loss": 0.2895,
+      "step": 25701
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1368360955787283e-05,
+      "loss": 0.3345,
+      "step": 25702
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1368025120480916e-05,
+      "loss": 0.4044,
+      "step": 25703
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1367689285174553e-05,
+      "loss": 0.3874,
+      "step": 25704
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1367353449868186e-05,
+      "loss": 0.2703,
+      "step": 25705
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1367017614561819e-05,
+      "loss": 0.2569,
+      "step": 25706
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 1.1366681779255455e-05,
+      "loss": 0.4051,
+      "step": 25707
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1366345943949088e-05,
+      "loss": 0.489,
+      "step": 25708
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1366010108642723e-05,
+      "loss": 0.295,
+      "step": 25709
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1365674273336357e-05,
+      "loss": 0.3063,
+      "step": 25710
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1365338438029992e-05,
+      "loss": 0.2982,
+      "step": 25711
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1365002602723625e-05,
+      "loss": 0.2508,
+      "step": 25712
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1364666767417261e-05,
+      "loss": 0.3081,
+      "step": 25713
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1364330932110894e-05,
+      "loss": 0.3068,
+      "step": 25714
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1363995096804527e-05,
+      "loss": 0.3567,
+      "step": 25715
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1363659261498163e-05,
+      "loss": 0.3264,
+      "step": 25716
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1363323426191796e-05,
+      "loss": 0.2957,
+      "step": 25717
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1362987590885431e-05,
+      "loss": 0.3761,
+      "step": 25718
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1362651755579064e-05,
+      "loss": 0.2976,
+      "step": 25719
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.13623159202727e-05,
+      "loss": 0.2274,
+      "step": 25720
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1361980084966333e-05,
+      "loss": 0.2496,
+      "step": 25721
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1361644249659966e-05,
+      "loss": 0.3383,
+      "step": 25722
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1361308414353603e-05,
+      "loss": 0.3616,
+      "step": 25723
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1360972579047236e-05,
+      "loss": 0.1995,
+      "step": 25724
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.136063674374087e-05,
+      "loss": 0.4033,
+      "step": 25725
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1360300908434505e-05,
+      "loss": 0.3279,
+      "step": 25726
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1359965073128138e-05,
+      "loss": 0.2789,
+      "step": 25727
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1359629237821772e-05,
+      "loss": 0.2465,
+      "step": 25728
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1359293402515407e-05,
+      "loss": 0.2299,
+      "step": 25729
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1358957567209042e-05,
+      "loss": 0.2884,
+      "step": 25730
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1358621731902675e-05,
+      "loss": 0.212,
+      "step": 25731
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1358285896596311e-05,
+      "loss": 0.2674,
+      "step": 25732
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1357950061289944e-05,
+      "loss": 0.2028,
+      "step": 25733
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1357614225983577e-05,
+      "loss": 0.3044,
+      "step": 25734
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1357278390677213e-05,
+      "loss": 0.3199,
+      "step": 25735
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1356942555370846e-05,
+      "loss": 0.2805,
+      "step": 25736
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1356606720064481e-05,
+      "loss": 0.3488,
+      "step": 25737
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1356270884758116e-05,
+      "loss": 0.2056,
+      "step": 25738
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.135593504945175e-05,
+      "loss": 0.165,
+      "step": 25739
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1355599214145383e-05,
+      "loss": 0.4265,
+      "step": 25740
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.135526337883902e-05,
+      "loss": 0.2814,
+      "step": 25741
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1354927543532652e-05,
+      "loss": 0.3289,
+      "step": 25742
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1354591708226285e-05,
+      "loss": 0.2561,
+      "step": 25743
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1354255872919922e-05,
+      "loss": 0.1987,
+      "step": 25744
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1353920037613555e-05,
+      "loss": 0.302,
+      "step": 25745
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.135358420230719e-05,
+      "loss": 0.3464,
+      "step": 25746
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1353248367000824e-05,
+      "loss": 0.2835,
+      "step": 25747
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1352912531694457e-05,
+      "loss": 0.1119,
+      "step": 25748
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1352576696388092e-05,
+      "loss": 0.4145,
+      "step": 25749
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1352240861081726e-05,
+      "loss": 0.4269,
+      "step": 25750
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1351905025775361e-05,
+      "loss": 0.2483,
+      "step": 25751
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1351569190468994e-05,
+      "loss": 0.1979,
+      "step": 25752
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.135123335516263e-05,
+      "loss": 0.2249,
+      "step": 25753
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1350897519856263e-05,
+      "loss": 0.1814,
+      "step": 25754
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1350561684549896e-05,
+      "loss": 0.2715,
+      "step": 25755
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1350225849243532e-05,
+      "loss": 0.1332,
+      "step": 25756
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1349890013937165e-05,
+      "loss": 0.2635,
+      "step": 25757
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.13495541786308e-05,
+      "loss": 0.411,
+      "step": 25758
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1349218343324435e-05,
+      "loss": 0.2346,
+      "step": 25759
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.134888250801807e-05,
+      "loss": 0.2771,
+      "step": 25760
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1348546672711702e-05,
+      "loss": 0.1911,
+      "step": 25761
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1348210837405339e-05,
+      "loss": 0.2824,
+      "step": 25762
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1347875002098972e-05,
+      "loss": 0.2388,
+      "step": 25763
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1347539166792604e-05,
+      "loss": 0.3802,
+      "step": 25764
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.134720333148624e-05,
+      "loss": 0.2339,
+      "step": 25765
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1346867496179874e-05,
+      "loss": 0.181,
+      "step": 25766
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1346531660873508e-05,
+      "loss": 0.4022,
+      "step": 25767
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1346195825567143e-05,
+      "loss": 0.3474,
+      "step": 25768
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1345859990260778e-05,
+      "loss": 0.2437,
+      "step": 25769
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.134552415495441e-05,
+      "loss": 0.2036,
+      "step": 25770
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1345188319648045e-05,
+      "loss": 0.236,
+      "step": 25771
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.134485248434168e-05,
+      "loss": 0.4436,
+      "step": 25772
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1344516649035313e-05,
+      "loss": 0.3673,
+      "step": 25773
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.134418081372895e-05,
+      "loss": 0.2939,
+      "step": 25774
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1343844978422582e-05,
+      "loss": 0.2517,
+      "step": 25775
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1343509143116215e-05,
+      "loss": 0.3041,
+      "step": 25776
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1343173307809852e-05,
+      "loss": 0.3076,
+      "step": 25777
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1342837472503484e-05,
+      "loss": 0.4656,
+      "step": 25778
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1342501637197119e-05,
+      "loss": 0.3994,
+      "step": 25779
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1342165801890754e-05,
+      "loss": 0.416,
+      "step": 25780
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1341829966584388e-05,
+      "loss": 0.6109,
+      "step": 25781
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1341494131278021e-05,
+      "loss": 0.2642,
+      "step": 25782
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1341158295971658e-05,
+      "loss": 0.3085,
+      "step": 25783
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.134082246066529e-05,
+      "loss": 0.3272,
+      "step": 25784
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1340486625358924e-05,
+      "loss": 0.5785,
+      "step": 25785
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.134015079005256e-05,
+      "loss": 0.2685,
+      "step": 25786
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1339814954746193e-05,
+      "loss": 0.309,
+      "step": 25787
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1339479119439828e-05,
+      "loss": 0.2476,
+      "step": 25788
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1339143284133462e-05,
+      "loss": 0.3967,
+      "step": 25789
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1338807448827097e-05,
+      "loss": 0.2049,
+      "step": 25790
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.133847161352073e-05,
+      "loss": 0.1713,
+      "step": 25791
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1338135778214364e-05,
+      "loss": 0.1902,
+      "step": 25792
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1337799942907999e-05,
+      "loss": 0.2771,
+      "step": 25793
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1337464107601632e-05,
+      "loss": 0.3117,
+      "step": 25794
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1337128272295268e-05,
+      "loss": 0.5098,
+      "step": 25795
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1336792436988901e-05,
+      "loss": 0.3694,
+      "step": 25796
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1336456601682534e-05,
+      "loss": 0.1368,
+      "step": 25797
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.133612076637617e-05,
+      "loss": 0.1853,
+      "step": 25798
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1335784931069804e-05,
+      "loss": 0.413,
+      "step": 25799
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1335449095763438e-05,
+      "loss": 0.397,
+      "step": 25800
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1335113260457073e-05,
+      "loss": 0.5047,
+      "step": 25801
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1334777425150708e-05,
+      "loss": 0.1861,
+      "step": 25802
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.133444158984434e-05,
+      "loss": 0.3492,
+      "step": 25803
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1334105754537977e-05,
+      "loss": 0.3409,
+      "step": 25804
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.133376991923161e-05,
+      "loss": 0.2432,
+      "step": 25805
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1333434083925243e-05,
+      "loss": 0.2412,
+      "step": 25806
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1333098248618879e-05,
+      "loss": 0.2941,
+      "step": 25807
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1332762413312512e-05,
+      "loss": 0.3636,
+      "step": 25808
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1332426578006147e-05,
+      "loss": 0.3074,
+      "step": 25809
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1332090742699781e-05,
+      "loss": 0.2913,
+      "step": 25810
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1331754907393416e-05,
+      "loss": 0.3153,
+      "step": 25811
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1331419072087049e-05,
+      "loss": 0.2479,
+      "step": 25812
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1331083236780685e-05,
+      "loss": 0.2664,
+      "step": 25813
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1330747401474318e-05,
+      "loss": 0.3277,
+      "step": 25814
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1330411566167951e-05,
+      "loss": 0.416,
+      "step": 25815
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1330075730861588e-05,
+      "loss": 0.4759,
+      "step": 25816
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.132973989555522e-05,
+      "loss": 0.3643,
+      "step": 25817
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1329404060248853e-05,
+      "loss": 0.2939,
+      "step": 25818
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.132906822494249e-05,
+      "loss": 0.2271,
+      "step": 25819
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1328732389636123e-05,
+      "loss": 0.376,
+      "step": 25820
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1328396554329757e-05,
+      "loss": 0.2342,
+      "step": 25821
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1328060719023392e-05,
+      "loss": 0.4569,
+      "step": 25822
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1327724883717027e-05,
+      "loss": 0.1917,
+      "step": 25823
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.132738904841066e-05,
+      "loss": 0.3968,
+      "step": 25824
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1327053213104296e-05,
+      "loss": 0.2171,
+      "step": 25825
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1326717377797929e-05,
+      "loss": 0.2556,
+      "step": 25826
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1326381542491562e-05,
+      "loss": 0.2137,
+      "step": 25827
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1326045707185198e-05,
+      "loss": 0.3154,
+      "step": 25828
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1325709871878831e-05,
+      "loss": 0.2402,
+      "step": 25829
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1325374036572466e-05,
+      "loss": 0.3015,
+      "step": 25830
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.13250382012661e-05,
+      "loss": 0.3657,
+      "step": 25831
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1324702365959735e-05,
+      "loss": 0.2832,
+      "step": 25832
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1324366530653368e-05,
+      "loss": 0.3216,
+      "step": 25833
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1324030695347004e-05,
+      "loss": 0.3199,
+      "step": 25834
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1323694860040637e-05,
+      "loss": 0.2199,
+      "step": 25835
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.132335902473427e-05,
+      "loss": 0.203,
+      "step": 25836
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1323023189427907e-05,
+      "loss": 0.2483,
+      "step": 25837
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.132268735412154e-05,
+      "loss": 0.1687,
+      "step": 25838
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1322351518815173e-05,
+      "loss": 0.3914,
+      "step": 25839
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1322015683508809e-05,
+      "loss": 0.2289,
+      "step": 25840
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1321679848202442e-05,
+      "loss": 0.3792,
+      "step": 25841
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1321344012896076e-05,
+      "loss": 0.2265,
+      "step": 25842
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1321008177589711e-05,
+      "loss": 0.2497,
+      "step": 25843
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1320672342283346e-05,
+      "loss": 0.2559,
+      "step": 25844
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1320336506976979e-05,
+      "loss": 0.1172,
+      "step": 25845
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1320000671670615e-05,
+      "loss": 0.2144,
+      "step": 25846
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1319664836364248e-05,
+      "loss": 0.2381,
+      "step": 25847
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1319329001057881e-05,
+      "loss": 0.2324,
+      "step": 25848
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1318993165751517e-05,
+      "loss": 0.1315,
+      "step": 25849
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.131865733044515e-05,
+      "loss": 0.1284,
+      "step": 25850
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1318321495138785e-05,
+      "loss": 0.3402,
+      "step": 25851
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.131798565983242e-05,
+      "loss": 0.2831,
+      "step": 25852
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1317649824526054e-05,
+      "loss": 0.34,
+      "step": 25853
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1317313989219687e-05,
+      "loss": 0.3029,
+      "step": 25854
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1316978153913324e-05,
+      "loss": 0.3145,
+      "step": 25855
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1316642318606956e-05,
+      "loss": 0.2067,
+      "step": 25856
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.131630648330059e-05,
+      "loss": 0.3247,
+      "step": 25857
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1315970647994226e-05,
+      "loss": 0.3072,
+      "step": 25858
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1315634812687859e-05,
+      "loss": 0.2457,
+      "step": 25859
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1315298977381493e-05,
+      "loss": 0.3718,
+      "step": 25860
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1314963142075128e-05,
+      "loss": 0.3502,
+      "step": 25861
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1314627306768761e-05,
+      "loss": 0.3379,
+      "step": 25862
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1314291471462396e-05,
+      "loss": 0.261,
+      "step": 25863
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.131395563615603e-05,
+      "loss": 0.271,
+      "step": 25864
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1313619800849665e-05,
+      "loss": 0.3067,
+      "step": 25865
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1313283965543298e-05,
+      "loss": 0.4278,
+      "step": 25866
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1312948130236934e-05,
+      "loss": 0.2321,
+      "step": 25867
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1312612294930567e-05,
+      "loss": 0.2934,
+      "step": 25868
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.13122764596242e-05,
+      "loss": 0.1704,
+      "step": 25869
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1311940624317836e-05,
+      "loss": 0.2848,
+      "step": 25870
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.131160478901147e-05,
+      "loss": 0.2603,
+      "step": 25871
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1311268953705104e-05,
+      "loss": 0.2113,
+      "step": 25872
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1310933118398739e-05,
+      "loss": 0.2082,
+      "step": 25873
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1310597283092373e-05,
+      "loss": 0.2207,
+      "step": 25874
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1310261447786006e-05,
+      "loss": 0.3019,
+      "step": 25875
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1309925612479643e-05,
+      "loss": 0.186,
+      "step": 25876
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1309589777173276e-05,
+      "loss": 0.4459,
+      "step": 25877
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1309253941866909e-05,
+      "loss": 0.1806,
+      "step": 25878
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1308918106560545e-05,
+      "loss": 0.566,
+      "step": 25879
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1308582271254178e-05,
+      "loss": 0.3888,
+      "step": 25880
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1308246435947812e-05,
+      "loss": 0.1945,
+      "step": 25881
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1307910600641447e-05,
+      "loss": 0.2653,
+      "step": 25882
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.130757476533508e-05,
+      "loss": 0.4514,
+      "step": 25883
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1307238930028715e-05,
+      "loss": 0.2501,
+      "step": 25884
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.130690309472235e-05,
+      "loss": 0.3516,
+      "step": 25885
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1306567259415984e-05,
+      "loss": 0.4115,
+      "step": 25886
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1306231424109617e-05,
+      "loss": 0.2079,
+      "step": 25887
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1305895588803253e-05,
+      "loss": 0.2492,
+      "step": 25888
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1305559753496886e-05,
+      "loss": 0.4641,
+      "step": 25889
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.130522391819052e-05,
+      "loss": 0.2288,
+      "step": 25890
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1304888082884156e-05,
+      "loss": 0.2288,
+      "step": 25891
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1304552247577789e-05,
+      "loss": 0.3393,
+      "step": 25892
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1304216412271423e-05,
+      "loss": 0.2785,
+      "step": 25893
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1303880576965058e-05,
+      "loss": 0.4651,
+      "step": 25894
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1303544741658692e-05,
+      "loss": 0.3043,
+      "step": 25895
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1303208906352325e-05,
+      "loss": 0.2366,
+      "step": 25896
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1302873071045958e-05,
+      "loss": 0.2404,
+      "step": 25897
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1302537235739595e-05,
+      "loss": 0.309,
+      "step": 25898
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1302201400433228e-05,
+      "loss": 0.2805,
+      "step": 25899
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1301865565126862e-05,
+      "loss": 0.414,
+      "step": 25900
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1301529729820497e-05,
+      "loss": 0.1469,
+      "step": 25901
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1301193894514132e-05,
+      "loss": 0.3503,
+      "step": 25902
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1300858059207765e-05,
+      "loss": 0.1313,
+      "step": 25903
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1300522223901401e-05,
+      "loss": 0.3517,
+      "step": 25904
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 1.1300186388595034e-05,
+      "loss": 0.361,
+      "step": 25905
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1299850553288667e-05,
+      "loss": 0.3512,
+      "step": 25906
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1299514717982303e-05,
+      "loss": 0.5018,
+      "step": 25907
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1299178882675936e-05,
+      "loss": 0.2906,
+      "step": 25908
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1298843047369569e-05,
+      "loss": 0.1921,
+      "step": 25909
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1298507212063205e-05,
+      "loss": 0.2099,
+      "step": 25910
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1298171376756838e-05,
+      "loss": 0.3713,
+      "step": 25911
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1297835541450473e-05,
+      "loss": 0.4615,
+      "step": 25912
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1297499706144108e-05,
+      "loss": 0.196,
+      "step": 25913
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1297163870837742e-05,
+      "loss": 0.2969,
+      "step": 25914
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1296828035531375e-05,
+      "loss": 0.195,
+      "step": 25915
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1296492200225012e-05,
+      "loss": 0.1327,
+      "step": 25916
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1296156364918645e-05,
+      "loss": 0.0894,
+      "step": 25917
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1295820529612278e-05,
+      "loss": 0.3322,
+      "step": 25918
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1295484694305914e-05,
+      "loss": 0.2654,
+      "step": 25919
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1295148858999547e-05,
+      "loss": 0.2421,
+      "step": 25920
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1294813023693181e-05,
+      "loss": 0.3095,
+      "step": 25921
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1294477188386816e-05,
+      "loss": 0.4154,
+      "step": 25922
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.129414135308045e-05,
+      "loss": 0.233,
+      "step": 25923
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1293805517774084e-05,
+      "loss": 0.3893,
+      "step": 25924
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.129346968246772e-05,
+      "loss": 0.258,
+      "step": 25925
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1293133847161353e-05,
+      "loss": 0.2072,
+      "step": 25926
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1292798011854986e-05,
+      "loss": 0.4382,
+      "step": 25927
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1292462176548622e-05,
+      "loss": 0.2081,
+      "step": 25928
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1292126341242255e-05,
+      "loss": 0.406,
+      "step": 25929
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1291790505935888e-05,
+      "loss": 0.3089,
+      "step": 25930
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1291454670629525e-05,
+      "loss": 0.348,
+      "step": 25931
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1291118835323157e-05,
+      "loss": 0.2027,
+      "step": 25932
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1290783000016792e-05,
+      "loss": 0.2465,
+      "step": 25933
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1290447164710427e-05,
+      "loss": 0.4935,
+      "step": 25934
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1290111329404061e-05,
+      "loss": 0.1983,
+      "step": 25935
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1289775494097694e-05,
+      "loss": 0.2686,
+      "step": 25936
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.128943965879133e-05,
+      "loss": 0.1349,
+      "step": 25937
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1289103823484964e-05,
+      "loss": 0.4639,
+      "step": 25938
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1288767988178597e-05,
+      "loss": 0.1543,
+      "step": 25939
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1288432152872233e-05,
+      "loss": 0.2022,
+      "step": 25940
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1288096317565866e-05,
+      "loss": 0.4105,
+      "step": 25941
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.12877604822595e-05,
+      "loss": 0.2553,
+      "step": 25942
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1287424646953135e-05,
+      "loss": 0.3698,
+      "step": 25943
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.128708881164677e-05,
+      "loss": 0.3551,
+      "step": 25944
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1286752976340403e-05,
+      "loss": 0.4007,
+      "step": 25945
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.128641714103404e-05,
+      "loss": 0.1426,
+      "step": 25946
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1286081305727672e-05,
+      "loss": 0.2965,
+      "step": 25947
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1285745470421305e-05,
+      "loss": 0.2707,
+      "step": 25948
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1285409635114941e-05,
+      "loss": 0.3188,
+      "step": 25949
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1285073799808574e-05,
+      "loss": 0.249,
+      "step": 25950
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1284737964502207e-05,
+      "loss": 0.2557,
+      "step": 25951
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1284402129195844e-05,
+      "loss": 0.2528,
+      "step": 25952
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1284066293889477e-05,
+      "loss": 0.2604,
+      "step": 25953
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1283730458583111e-05,
+      "loss": 0.3324,
+      "step": 25954
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1283394623276746e-05,
+      "loss": 0.3516,
+      "step": 25955
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.128305878797038e-05,
+      "loss": 0.3765,
+      "step": 25956
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1282722952664014e-05,
+      "loss": 0.4046,
+      "step": 25957
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.128238711735765e-05,
+      "loss": 0.3488,
+      "step": 25958
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1282051282051283e-05,
+      "loss": 0.186,
+      "step": 25959
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1281715446744916e-05,
+      "loss": 0.213,
+      "step": 25960
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1281379611438552e-05,
+      "loss": 0.4156,
+      "step": 25961
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1281043776132185e-05,
+      "loss": 0.2397,
+      "step": 25962
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.128070794082582e-05,
+      "loss": 0.3962,
+      "step": 25963
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1280372105519454e-05,
+      "loss": 0.4429,
+      "step": 25964
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1280036270213089e-05,
+      "loss": 0.1818,
+      "step": 25965
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1279700434906722e-05,
+      "loss": 0.2348,
+      "step": 25966
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1279364599600358e-05,
+      "loss": 0.1483,
+      "step": 25967
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1279028764293991e-05,
+      "loss": 0.2468,
+      "step": 25968
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1278692928987624e-05,
+      "loss": 0.1612,
+      "step": 25969
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.127835709368126e-05,
+      "loss": 0.3581,
+      "step": 25970
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1278021258374893e-05,
+      "loss": 0.2824,
+      "step": 25971
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1277685423068528e-05,
+      "loss": 0.1503,
+      "step": 25972
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1277349587762163e-05,
+      "loss": 0.2932,
+      "step": 25973
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1277013752455796e-05,
+      "loss": 0.2373,
+      "step": 25974
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.127667791714943e-05,
+      "loss": 0.2348,
+      "step": 25975
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1276342081843065e-05,
+      "loss": 0.2997,
+      "step": 25976
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.12760062465367e-05,
+      "loss": 0.3186,
+      "step": 25977
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1275670411230333e-05,
+      "loss": 0.2379,
+      "step": 25978
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1275334575923969e-05,
+      "loss": 0.1436,
+      "step": 25979
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1274998740617602e-05,
+      "loss": 0.2512,
+      "step": 25980
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1274662905311235e-05,
+      "loss": 0.1594,
+      "step": 25981
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1274327070004871e-05,
+      "loss": 0.4009,
+      "step": 25982
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1273991234698504e-05,
+      "loss": 0.3995,
+      "step": 25983
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1273655399392139e-05,
+      "loss": 0.3039,
+      "step": 25984
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1273319564085773e-05,
+      "loss": 0.3359,
+      "step": 25985
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1272983728779408e-05,
+      "loss": 0.3492,
+      "step": 25986
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1272647893473041e-05,
+      "loss": 0.4541,
+      "step": 25987
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1272312058166677e-05,
+      "loss": 0.5418,
+      "step": 25988
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.127197622286031e-05,
+      "loss": 0.363,
+      "step": 25989
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1271640387553943e-05,
+      "loss": 0.3058,
+      "step": 25990
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.127130455224758e-05,
+      "loss": 0.226,
+      "step": 25991
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1270968716941213e-05,
+      "loss": 0.1997,
+      "step": 25992
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1270632881634847e-05,
+      "loss": 0.2083,
+      "step": 25993
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1270297046328482e-05,
+      "loss": 0.1018,
+      "step": 25994
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1269961211022117e-05,
+      "loss": 0.3909,
+      "step": 25995
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.126962537571575e-05,
+      "loss": 0.2719,
+      "step": 25996
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1269289540409384e-05,
+      "loss": 0.2195,
+      "step": 25997
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1268953705103019e-05,
+      "loss": 0.2465,
+      "step": 25998
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1268617869796652e-05,
+      "loss": 0.2836,
+      "step": 25999
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1268282034490288e-05,
+      "loss": 0.3242,
+      "step": 26000
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1267946199183921e-05,
+      "loss": 0.4277,
+      "step": 26001
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1267610363877554e-05,
+      "loss": 0.4392,
+      "step": 26002
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.126727452857119e-05,
+      "loss": 0.3063,
+      "step": 26003
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1266938693264823e-05,
+      "loss": 0.3145,
+      "step": 26004
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1266602857958458e-05,
+      "loss": 0.3076,
+      "step": 26005
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1266267022652093e-05,
+      "loss": 0.0913,
+      "step": 26006
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1265931187345727e-05,
+      "loss": 0.3184,
+      "step": 26007
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.126559535203936e-05,
+      "loss": 0.239,
+      "step": 26008
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1265259516732997e-05,
+      "loss": 0.2089,
+      "step": 26009
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.126492368142663e-05,
+      "loss": 0.4534,
+      "step": 26010
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1264587846120262e-05,
+      "loss": 0.2466,
+      "step": 26011
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1264252010813899e-05,
+      "loss": 0.3933,
+      "step": 26012
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1263916175507532e-05,
+      "loss": 0.3175,
+      "step": 26013
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1263580340201166e-05,
+      "loss": 0.4033,
+      "step": 26014
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1263244504894801e-05,
+      "loss": 0.2905,
+      "step": 26015
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1262908669588436e-05,
+      "loss": 0.1544,
+      "step": 26016
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1262572834282069e-05,
+      "loss": 0.1536,
+      "step": 26017
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1262236998975703e-05,
+      "loss": 0.2235,
+      "step": 26018
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1261901163669338e-05,
+      "loss": 0.3654,
+      "step": 26019
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1261565328362971e-05,
+      "loss": 0.2344,
+      "step": 26020
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1261229493056607e-05,
+      "loss": 0.3539,
+      "step": 26021
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.126089365775024e-05,
+      "loss": 0.266,
+      "step": 26022
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1260557822443873e-05,
+      "loss": 0.2685,
+      "step": 26023
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.126022198713751e-05,
+      "loss": 0.4396,
+      "step": 26024
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1259886151831142e-05,
+      "loss": 0.3868,
+      "step": 26025
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1259550316524777e-05,
+      "loss": 0.4443,
+      "step": 26026
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1259214481218412e-05,
+      "loss": 0.1911,
+      "step": 26027
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1258878645912046e-05,
+      "loss": 0.3318,
+      "step": 26028
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.125854281060568e-05,
+      "loss": 0.195,
+      "step": 26029
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1258206975299316e-05,
+      "loss": 0.394,
+      "step": 26030
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1257871139992949e-05,
+      "loss": 0.2975,
+      "step": 26031
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1257535304686582e-05,
+      "loss": 0.2908,
+      "step": 26032
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1257199469380218e-05,
+      "loss": 0.3831,
+      "step": 26033
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1256863634073851e-05,
+      "loss": 0.3085,
+      "step": 26034
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1256527798767486e-05,
+      "loss": 0.2013,
+      "step": 26035
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.125619196346112e-05,
+      "loss": 0.4358,
+      "step": 26036
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1255856128154755e-05,
+      "loss": 0.2204,
+      "step": 26037
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1255520292848388e-05,
+      "loss": 0.3558,
+      "step": 26038
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1255184457542024e-05,
+      "loss": 0.2904,
+      "step": 26039
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1254848622235657e-05,
+      "loss": 0.2293,
+      "step": 26040
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.125451278692929e-05,
+      "loss": 0.4176,
+      "step": 26041
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1254176951622926e-05,
+      "loss": 0.3478,
+      "step": 26042
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.125384111631656e-05,
+      "loss": 0.3393,
+      "step": 26043
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1253505281010192e-05,
+      "loss": 0.3003,
+      "step": 26044
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1253169445703829e-05,
+      "loss": 0.2871,
+      "step": 26045
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1252833610397462e-05,
+      "loss": 0.2758,
+      "step": 26046
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1252497775091096e-05,
+      "loss": 0.3154,
+      "step": 26047
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.125216193978473e-05,
+      "loss": 0.2305,
+      "step": 26048
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1251826104478365e-05,
+      "loss": 0.2208,
+      "step": 26049
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1251490269171998e-05,
+      "loss": 0.3877,
+      "step": 26050
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1251154433865635e-05,
+      "loss": 0.3499,
+      "step": 26051
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1250818598559268e-05,
+      "loss": 0.2484,
+      "step": 26052
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.12504827632529e-05,
+      "loss": 0.3719,
+      "step": 26053
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1250146927946537e-05,
+      "loss": 0.3192,
+      "step": 26054
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.124981109264017e-05,
+      "loss": 0.1686,
+      "step": 26055
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1249475257333805e-05,
+      "loss": 0.3855,
+      "step": 26056
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.124913942202744e-05,
+      "loss": 0.3257,
+      "step": 26057
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1248803586721074e-05,
+      "loss": 0.1499,
+      "step": 26058
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1248467751414707e-05,
+      "loss": 0.2401,
+      "step": 26059
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1248131916108343e-05,
+      "loss": 0.3874,
+      "step": 26060
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1247796080801976e-05,
+      "loss": 0.2861,
+      "step": 26061
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1247460245495609e-05,
+      "loss": 0.4218,
+      "step": 26062
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1247124410189245e-05,
+      "loss": 0.423,
+      "step": 26063
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1246788574882878e-05,
+      "loss": 0.2362,
+      "step": 26064
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1246452739576511e-05,
+      "loss": 0.2344,
+      "step": 26065
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1246116904270148e-05,
+      "loss": 0.2636,
+      "step": 26066
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.124578106896378e-05,
+      "loss": 0.2051,
+      "step": 26067
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1245445233657415e-05,
+      "loss": 0.3001,
+      "step": 26068
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.124510939835105e-05,
+      "loss": 0.2516,
+      "step": 26069
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1244773563044685e-05,
+      "loss": 0.246,
+      "step": 26070
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1244437727738318e-05,
+      "loss": 0.146,
+      "step": 26071
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1244101892431954e-05,
+      "loss": 0.3536,
+      "step": 26072
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1243766057125587e-05,
+      "loss": 0.3115,
+      "step": 26073
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.124343022181922e-05,
+      "loss": 0.1723,
+      "step": 26074
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1243094386512856e-05,
+      "loss": 0.4173,
+      "step": 26075
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1242758551206489e-05,
+      "loss": 0.2671,
+      "step": 26076
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1242422715900124e-05,
+      "loss": 0.1287,
+      "step": 26077
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1242086880593757e-05,
+      "loss": 0.3617,
+      "step": 26078
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1241751045287393e-05,
+      "loss": 0.3347,
+      "step": 26079
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1241415209981026e-05,
+      "loss": 0.3955,
+      "step": 26080
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1241079374674659e-05,
+      "loss": 0.3203,
+      "step": 26081
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1240743539368295e-05,
+      "loss": 0.4308,
+      "step": 26082
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1240407704061928e-05,
+      "loss": 0.3037,
+      "step": 26083
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1240071868755563e-05,
+      "loss": 0.168,
+      "step": 26084
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1239736033449198e-05,
+      "loss": 0.3027,
+      "step": 26085
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.123940019814283e-05,
+      "loss": 0.356,
+      "step": 26086
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1239064362836465e-05,
+      "loss": 0.1603,
+      "step": 26087
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.12387285275301e-05,
+      "loss": 0.1669,
+      "step": 26088
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1238392692223734e-05,
+      "loss": 0.3696,
+      "step": 26089
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1238056856917367e-05,
+      "loss": 0.2163,
+      "step": 26090
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1237721021611004e-05,
+      "loss": 0.3492,
+      "step": 26091
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1237385186304637e-05,
+      "loss": 0.2769,
+      "step": 26092
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.123704935099827e-05,
+      "loss": 0.2353,
+      "step": 26093
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1236713515691906e-05,
+      "loss": 0.3521,
+      "step": 26094
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1236377680385539e-05,
+      "loss": 0.3953,
+      "step": 26095
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1236041845079174e-05,
+      "loss": 0.1571,
+      "step": 26096
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1235706009772808e-05,
+      "loss": 0.3626,
+      "step": 26097
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1235370174466443e-05,
+      "loss": 0.2667,
+      "step": 26098
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1235034339160076e-05,
+      "loss": 0.3398,
+      "step": 26099
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1234698503853712e-05,
+      "loss": 0.2835,
+      "step": 26100
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1234362668547345e-05,
+      "loss": 0.315,
+      "step": 26101
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1234026833240978e-05,
+      "loss": 0.3403,
+      "step": 26102
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1233690997934614e-05,
+      "loss": 0.2441,
+      "step": 26103
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.1233355162628247e-05,
+      "loss": 0.1881,
+      "step": 26104
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1233019327321882e-05,
+      "loss": 0.2433,
+      "step": 26105
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1232683492015517e-05,
+      "loss": 0.3854,
+      "step": 26106
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1232347656709151e-05,
+      "loss": 0.1945,
+      "step": 26107
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1232011821402784e-05,
+      "loss": 0.3802,
+      "step": 26108
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1231675986096419e-05,
+      "loss": 0.2594,
+      "step": 26109
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1231340150790054e-05,
+      "loss": 0.1432,
+      "step": 26110
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1231004315483687e-05,
+      "loss": 0.3621,
+      "step": 26111
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1230668480177323e-05,
+      "loss": 0.3929,
+      "step": 26112
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1230332644870956e-05,
+      "loss": 0.3134,
+      "step": 26113
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1229996809564589e-05,
+      "loss": 0.1463,
+      "step": 26114
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1229660974258225e-05,
+      "loss": 0.2405,
+      "step": 26115
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1229325138951858e-05,
+      "loss": 0.3999,
+      "step": 26116
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1228989303645493e-05,
+      "loss": 0.3242,
+      "step": 26117
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1228653468339127e-05,
+      "loss": 0.2042,
+      "step": 26118
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1228317633032762e-05,
+      "loss": 0.1564,
+      "step": 26119
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1227981797726395e-05,
+      "loss": 0.3822,
+      "step": 26120
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1227645962420031e-05,
+      "loss": 0.2963,
+      "step": 26121
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1227310127113664e-05,
+      "loss": 0.1577,
+      "step": 26122
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1226974291807297e-05,
+      "loss": 0.2953,
+      "step": 26123
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1226638456500934e-05,
+      "loss": 0.4916,
+      "step": 26124
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1226302621194567e-05,
+      "loss": 0.3434,
+      "step": 26125
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1225966785888201e-05,
+      "loss": 0.1678,
+      "step": 26126
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1225630950581836e-05,
+      "loss": 0.3504,
+      "step": 26127
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.122529511527547e-05,
+      "loss": 0.3608,
+      "step": 26128
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1224959279969103e-05,
+      "loss": 0.3996,
+      "step": 26129
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.122462344466274e-05,
+      "loss": 0.3578,
+      "step": 26130
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1224287609356373e-05,
+      "loss": 0.2473,
+      "step": 26131
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1223951774050006e-05,
+      "loss": 0.2813,
+      "step": 26132
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1223615938743642e-05,
+      "loss": 0.3671,
+      "step": 26133
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1223280103437275e-05,
+      "loss": 0.3541,
+      "step": 26134
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1222944268130908e-05,
+      "loss": 0.3376,
+      "step": 26135
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1222608432824544e-05,
+      "loss": 0.2718,
+      "step": 26136
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1222272597518177e-05,
+      "loss": 0.3985,
+      "step": 26137
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1221936762211812e-05,
+      "loss": 0.3723,
+      "step": 26138
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1221600926905446e-05,
+      "loss": 0.2137,
+      "step": 26139
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1221265091599081e-05,
+      "loss": 0.3105,
+      "step": 26140
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1220929256292714e-05,
+      "loss": 0.3848,
+      "step": 26141
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.122059342098635e-05,
+      "loss": 0.2751,
+      "step": 26142
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1220257585679983e-05,
+      "loss": 0.3685,
+      "step": 26143
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1219921750373616e-05,
+      "loss": 0.2405,
+      "step": 26144
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1219585915067253e-05,
+      "loss": 0.2298,
+      "step": 26145
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1219250079760886e-05,
+      "loss": 0.1807,
+      "step": 26146
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.121891424445452e-05,
+      "loss": 0.4243,
+      "step": 26147
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1218578409148155e-05,
+      "loss": 0.1719,
+      "step": 26148
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.121824257384179e-05,
+      "loss": 0.2675,
+      "step": 26149
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1217906738535423e-05,
+      "loss": 0.3684,
+      "step": 26150
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1217570903229059e-05,
+      "loss": 0.2131,
+      "step": 26151
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1217235067922692e-05,
+      "loss": 0.2079,
+      "step": 26152
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1216899232616325e-05,
+      "loss": 0.388,
+      "step": 26153
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1216563397309961e-05,
+      "loss": 0.1968,
+      "step": 26154
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1216227562003594e-05,
+      "loss": 0.3776,
+      "step": 26155
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1215891726697227e-05,
+      "loss": 0.2801,
+      "step": 26156
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1215555891390863e-05,
+      "loss": 0.2449,
+      "step": 26157
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1215220056084496e-05,
+      "loss": 0.2977,
+      "step": 26158
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1214884220778131e-05,
+      "loss": 0.4094,
+      "step": 26159
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1214548385471766e-05,
+      "loss": 0.2738,
+      "step": 26160
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.12142125501654e-05,
+      "loss": 0.2391,
+      "step": 26161
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1213876714859033e-05,
+      "loss": 0.3791,
+      "step": 26162
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.121354087955267e-05,
+      "loss": 0.2251,
+      "step": 26163
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1213205044246303e-05,
+      "loss": 0.4151,
+      "step": 26164
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1212869208939935e-05,
+      "loss": 0.2638,
+      "step": 26165
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1212533373633572e-05,
+      "loss": 0.3983,
+      "step": 26166
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1212197538327205e-05,
+      "loss": 0.3487,
+      "step": 26167
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.121186170302084e-05,
+      "loss": 0.3573,
+      "step": 26168
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1211525867714474e-05,
+      "loss": 0.308,
+      "step": 26169
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1211190032408109e-05,
+      "loss": 0.2337,
+      "step": 26170
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1210854197101742e-05,
+      "loss": 0.3358,
+      "step": 26171
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1210518361795378e-05,
+      "loss": 0.2157,
+      "step": 26172
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1210182526489011e-05,
+      "loss": 0.2126,
+      "step": 26173
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1209846691182644e-05,
+      "loss": 0.3412,
+      "step": 26174
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.120951085587628e-05,
+      "loss": 0.2943,
+      "step": 26175
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1209175020569913e-05,
+      "loss": 0.2958,
+      "step": 26176
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1208839185263546e-05,
+      "loss": 0.2125,
+      "step": 26177
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1208503349957182e-05,
+      "loss": 0.3989,
+      "step": 26178
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1208167514650815e-05,
+      "loss": 0.4547,
+      "step": 26179
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.120783167934445e-05,
+      "loss": 0.2426,
+      "step": 26180
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1207495844038085e-05,
+      "loss": 0.1956,
+      "step": 26181
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.120716000873172e-05,
+      "loss": 0.2479,
+      "step": 26182
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1206824173425352e-05,
+      "loss": 0.1814,
+      "step": 26183
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1206488338118989e-05,
+      "loss": 0.4143,
+      "step": 26184
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1206152502812622e-05,
+      "loss": 0.3403,
+      "step": 26185
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1205816667506255e-05,
+      "loss": 0.3932,
+      "step": 26186
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1205480832199891e-05,
+      "loss": 0.3985,
+      "step": 26187
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1205144996893524e-05,
+      "loss": 0.392,
+      "step": 26188
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1204809161587159e-05,
+      "loss": 0.2289,
+      "step": 26189
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1204473326280793e-05,
+      "loss": 0.372,
+      "step": 26190
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1204137490974428e-05,
+      "loss": 0.2823,
+      "step": 26191
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.120380165566806e-05,
+      "loss": 0.338,
+      "step": 26192
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1203465820361697e-05,
+      "loss": 0.2988,
+      "step": 26193
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.120312998505533e-05,
+      "loss": 0.4443,
+      "step": 26194
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1202794149748963e-05,
+      "loss": 0.4103,
+      "step": 26195
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.12024583144426e-05,
+      "loss": 0.1793,
+      "step": 26196
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1202122479136232e-05,
+      "loss": 0.2776,
+      "step": 26197
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1201786643829867e-05,
+      "loss": 0.348,
+      "step": 26198
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1201450808523502e-05,
+      "loss": 0.2244,
+      "step": 26199
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1201114973217135e-05,
+      "loss": 0.3918,
+      "step": 26200
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.120077913791077e-05,
+      "loss": 0.2114,
+      "step": 26201
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1200443302604404e-05,
+      "loss": 0.3591,
+      "step": 26202
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1200107467298039e-05,
+      "loss": 0.3537,
+      "step": 26203
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1199771631991671e-05,
+      "loss": 0.3279,
+      "step": 26204
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1199435796685308e-05,
+      "loss": 0.4197,
+      "step": 26205
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.119909996137894e-05,
+      "loss": 0.3071,
+      "step": 26206
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1198764126072574e-05,
+      "loss": 0.2761,
+      "step": 26207
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.119842829076621e-05,
+      "loss": 0.4247,
+      "step": 26208
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1198092455459843e-05,
+      "loss": 0.2959,
+      "step": 26209
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1197756620153478e-05,
+      "loss": 0.3601,
+      "step": 26210
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1197420784847112e-05,
+      "loss": 0.2548,
+      "step": 26211
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1197084949540747e-05,
+      "loss": 0.3509,
+      "step": 26212
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.119674911423438e-05,
+      "loss": 0.4837,
+      "step": 26213
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1196413278928016e-05,
+      "loss": 0.2652,
+      "step": 26214
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.119607744362165e-05,
+      "loss": 0.23,
+      "step": 26215
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1195741608315282e-05,
+      "loss": 0.3679,
+      "step": 26216
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1195405773008918e-05,
+      "loss": 0.1317,
+      "step": 26217
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1195069937702551e-05,
+      "loss": 0.3324,
+      "step": 26218
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1194734102396186e-05,
+      "loss": 0.2352,
+      "step": 26219
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.119439826708982e-05,
+      "loss": 0.245,
+      "step": 26220
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1194062431783454e-05,
+      "loss": 0.4451,
+      "step": 26221
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1193726596477088e-05,
+      "loss": 0.1714,
+      "step": 26222
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1193390761170723e-05,
+      "loss": 0.2414,
+      "step": 26223
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1193054925864358e-05,
+      "loss": 0.4279,
+      "step": 26224
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.119271909055799e-05,
+      "loss": 0.2159,
+      "step": 26225
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1192383255251627e-05,
+      "loss": 0.3369,
+      "step": 26226
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.119204741994526e-05,
+      "loss": 0.3076,
+      "step": 26227
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1191711584638893e-05,
+      "loss": 0.2436,
+      "step": 26228
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.119137574933253e-05,
+      "loss": 0.4541,
+      "step": 26229
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1191039914026162e-05,
+      "loss": 0.2277,
+      "step": 26230
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1190704078719797e-05,
+      "loss": 0.2909,
+      "step": 26231
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1190368243413431e-05,
+      "loss": 0.1875,
+      "step": 26232
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1190032408107066e-05,
+      "loss": 0.3122,
+      "step": 26233
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1189696572800699e-05,
+      "loss": 0.2326,
+      "step": 26234
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1189360737494335e-05,
+      "loss": 0.2927,
+      "step": 26235
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1189024902187968e-05,
+      "loss": 0.2168,
+      "step": 26236
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1188689066881601e-05,
+      "loss": 0.2385,
+      "step": 26237
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1188353231575238e-05,
+      "loss": 0.1264,
+      "step": 26238
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.118801739626887e-05,
+      "loss": 0.2915,
+      "step": 26239
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1187681560962505e-05,
+      "loss": 0.3133,
+      "step": 26240
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.118734572565614e-05,
+      "loss": 0.2955,
+      "step": 26241
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1187009890349775e-05,
+      "loss": 0.2333,
+      "step": 26242
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1186674055043407e-05,
+      "loss": 0.2411,
+      "step": 26243
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1186338219737042e-05,
+      "loss": 0.2951,
+      "step": 26244
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1186002384430677e-05,
+      "loss": 0.3186,
+      "step": 26245
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.118566654912431e-05,
+      "loss": 0.1948,
+      "step": 26246
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1185330713817946e-05,
+      "loss": 0.5524,
+      "step": 26247
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1184994878511579e-05,
+      "loss": 0.2399,
+      "step": 26248
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1184659043205212e-05,
+      "loss": 0.382,
+      "step": 26249
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1184323207898848e-05,
+      "loss": 0.3748,
+      "step": 26250
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1183987372592481e-05,
+      "loss": 0.3109,
+      "step": 26251
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1183651537286116e-05,
+      "loss": 0.2462,
+      "step": 26252
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.118331570197975e-05,
+      "loss": 0.3858,
+      "step": 26253
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1182979866673385e-05,
+      "loss": 0.4378,
+      "step": 26254
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1182644031367018e-05,
+      "loss": 0.2316,
+      "step": 26255
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1182308196060654e-05,
+      "loss": 0.303,
+      "step": 26256
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1181972360754287e-05,
+      "loss": 0.2345,
+      "step": 26257
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.118163652544792e-05,
+      "loss": 0.3674,
+      "step": 26258
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1181300690141555e-05,
+      "loss": 0.2223,
+      "step": 26259
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.118096485483519e-05,
+      "loss": 0.347,
+      "step": 26260
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1180629019528824e-05,
+      "loss": 0.4055,
+      "step": 26261
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1180293184222457e-05,
+      "loss": 0.1983,
+      "step": 26262
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1179957348916094e-05,
+      "loss": 0.3137,
+      "step": 26263
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1179621513609727e-05,
+      "loss": 0.4437,
+      "step": 26264
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.117928567830336e-05,
+      "loss": 0.3821,
+      "step": 26265
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1178949842996996e-05,
+      "loss": 0.3593,
+      "step": 26266
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1178614007690629e-05,
+      "loss": 0.2752,
+      "step": 26267
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1178278172384262e-05,
+      "loss": 0.5211,
+      "step": 26268
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1177942337077898e-05,
+      "loss": 0.2027,
+      "step": 26269
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1177606501771531e-05,
+      "loss": 0.3197,
+      "step": 26270
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1177270666465166e-05,
+      "loss": 0.1827,
+      "step": 26271
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.11769348311588e-05,
+      "loss": 0.3387,
+      "step": 26272
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1176598995852435e-05,
+      "loss": 0.2464,
+      "step": 26273
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1176263160546068e-05,
+      "loss": 0.4809,
+      "step": 26274
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1175927325239704e-05,
+      "loss": 0.2677,
+      "step": 26275
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1175591489933337e-05,
+      "loss": 0.3327,
+      "step": 26276
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.117525565462697e-05,
+      "loss": 0.4747,
+      "step": 26277
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1174919819320607e-05,
+      "loss": 0.2863,
+      "step": 26278
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.117458398401424e-05,
+      "loss": 0.372,
+      "step": 26279
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1174248148707874e-05,
+      "loss": 0.394,
+      "step": 26280
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1173912313401509e-05,
+      "loss": 0.2247,
+      "step": 26281
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1173576478095143e-05,
+      "loss": 0.264,
+      "step": 26282
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1173240642788776e-05,
+      "loss": 0.5605,
+      "step": 26283
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1172904807482413e-05,
+      "loss": 0.2441,
+      "step": 26284
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1172568972176046e-05,
+      "loss": 0.4692,
+      "step": 26285
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1172233136869679e-05,
+      "loss": 0.2384,
+      "step": 26286
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1171897301563315e-05,
+      "loss": 0.3233,
+      "step": 26287
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1171561466256948e-05,
+      "loss": 0.41,
+      "step": 26288
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1171225630950583e-05,
+      "loss": 0.2413,
+      "step": 26289
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1170889795644217e-05,
+      "loss": 0.3684,
+      "step": 26290
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.117055396033785e-05,
+      "loss": 0.5543,
+      "step": 26291
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1170218125031485e-05,
+      "loss": 0.401,
+      "step": 26292
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.116988228972512e-05,
+      "loss": 0.1823,
+      "step": 26293
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1169546454418754e-05,
+      "loss": 0.3881,
+      "step": 26294
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1169210619112387e-05,
+      "loss": 0.2183,
+      "step": 26295
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1168874783806023e-05,
+      "loss": 0.3357,
+      "step": 26296
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1168538948499656e-05,
+      "loss": 0.3132,
+      "step": 26297
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.116820311319329e-05,
+      "loss": 0.2872,
+      "step": 26298
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1167867277886926e-05,
+      "loss": 0.2426,
+      "step": 26299
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1167531442580559e-05,
+      "loss": 0.3148,
+      "step": 26300
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1167195607274193e-05,
+      "loss": 0.2339,
+      "step": 26301
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 1.1166859771967828e-05,
+      "loss": 0.3224,
+      "step": 26302
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1166523936661463e-05,
+      "loss": 0.482,
+      "step": 26303
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1166188101355096e-05,
+      "loss": 0.3459,
+      "step": 26304
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1165852266048732e-05,
+      "loss": 0.4005,
+      "step": 26305
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1165516430742365e-05,
+      "loss": 0.1929,
+      "step": 26306
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1165180595435998e-05,
+      "loss": 0.2772,
+      "step": 26307
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1164844760129634e-05,
+      "loss": 0.3364,
+      "step": 26308
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1164508924823267e-05,
+      "loss": 0.3336,
+      "step": 26309
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1164173089516902e-05,
+      "loss": 0.2683,
+      "step": 26310
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1163837254210536e-05,
+      "loss": 0.2205,
+      "step": 26311
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.116350141890417e-05,
+      "loss": 0.5142,
+      "step": 26312
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1163165583597804e-05,
+      "loss": 0.2992,
+      "step": 26313
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1162829748291439e-05,
+      "loss": 0.3119,
+      "step": 26314
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1162493912985073e-05,
+      "loss": 0.3524,
+      "step": 26315
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1162158077678706e-05,
+      "loss": 0.203,
+      "step": 26316
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1161822242372343e-05,
+      "loss": 0.2142,
+      "step": 26317
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1161486407065976e-05,
+      "loss": 0.2406,
+      "step": 26318
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1161150571759608e-05,
+      "loss": 0.2037,
+      "step": 26319
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1160814736453245e-05,
+      "loss": 0.3081,
+      "step": 26320
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1160478901146878e-05,
+      "loss": 0.4329,
+      "step": 26321
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1160143065840512e-05,
+      "loss": 0.3006,
+      "step": 26322
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1159807230534147e-05,
+      "loss": 0.2415,
+      "step": 26323
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1159471395227782e-05,
+      "loss": 0.3469,
+      "step": 26324
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1159135559921415e-05,
+      "loss": 0.277,
+      "step": 26325
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1158799724615051e-05,
+      "loss": 0.3111,
+      "step": 26326
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1158463889308684e-05,
+      "loss": 0.264,
+      "step": 26327
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1158128054002317e-05,
+      "loss": 0.1422,
+      "step": 26328
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1157792218695953e-05,
+      "loss": 0.3552,
+      "step": 26329
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1157456383389586e-05,
+      "loss": 0.231,
+      "step": 26330
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1157120548083221e-05,
+      "loss": 0.3021,
+      "step": 26331
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1156784712776856e-05,
+      "loss": 0.2046,
+      "step": 26332
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.115644887747049e-05,
+      "loss": 0.2649,
+      "step": 26333
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1156113042164123e-05,
+      "loss": 0.3949,
+      "step": 26334
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1155777206857758e-05,
+      "loss": 0.2403,
+      "step": 26335
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1155441371551392e-05,
+      "loss": 0.2184,
+      "step": 26336
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1155105536245025e-05,
+      "loss": 0.2946,
+      "step": 26337
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1154769700938662e-05,
+      "loss": 0.2118,
+      "step": 26338
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1154433865632295e-05,
+      "loss": 0.2207,
+      "step": 26339
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1154098030325928e-05,
+      "loss": 0.3535,
+      "step": 26340
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1153762195019564e-05,
+      "loss": 0.3525,
+      "step": 26341
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1153426359713197e-05,
+      "loss": 0.3488,
+      "step": 26342
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1153090524406832e-05,
+      "loss": 0.289,
+      "step": 26343
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1152754689100466e-05,
+      "loss": 0.4155,
+      "step": 26344
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.11524188537941e-05,
+      "loss": 0.2941,
+      "step": 26345
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1152083018487734e-05,
+      "loss": 0.2946,
+      "step": 26346
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.115174718318137e-05,
+      "loss": 0.1683,
+      "step": 26347
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1151411347875003e-05,
+      "loss": 0.3036,
+      "step": 26348
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1151075512568636e-05,
+      "loss": 0.2766,
+      "step": 26349
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1150739677262272e-05,
+      "loss": 0.2493,
+      "step": 26350
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1150403841955905e-05,
+      "loss": 0.2848,
+      "step": 26351
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.115006800664954e-05,
+      "loss": 0.3242,
+      "step": 26352
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1149732171343175e-05,
+      "loss": 0.3064,
+      "step": 26353
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.114939633603681e-05,
+      "loss": 0.203,
+      "step": 26354
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1149060500730442e-05,
+      "loss": 0.1874,
+      "step": 26355
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1148724665424077e-05,
+      "loss": 0.3846,
+      "step": 26356
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1148388830117712e-05,
+      "loss": 0.2105,
+      "step": 26357
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1148052994811344e-05,
+      "loss": 0.2646,
+      "step": 26358
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.114771715950498e-05,
+      "loss": 0.5314,
+      "step": 26359
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1147381324198614e-05,
+      "loss": 0.2965,
+      "step": 26360
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1147045488892247e-05,
+      "loss": 0.3054,
+      "step": 26361
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1146709653585883e-05,
+      "loss": 0.2884,
+      "step": 26362
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1146373818279516e-05,
+      "loss": 0.1924,
+      "step": 26363
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.114603798297315e-05,
+      "loss": 0.2368,
+      "step": 26364
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1145702147666785e-05,
+      "loss": 0.1851,
+      "step": 26365
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.114536631236042e-05,
+      "loss": 0.2081,
+      "step": 26366
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1145030477054053e-05,
+      "loss": 0.3384,
+      "step": 26367
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.114469464174769e-05,
+      "loss": 0.2099,
+      "step": 26368
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1144358806441322e-05,
+      "loss": 0.3479,
+      "step": 26369
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1144022971134955e-05,
+      "loss": 0.3916,
+      "step": 26370
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1143687135828592e-05,
+      "loss": 0.3393,
+      "step": 26371
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1143351300522224e-05,
+      "loss": 0.3101,
+      "step": 26372
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1143015465215859e-05,
+      "loss": 0.2968,
+      "step": 26373
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1142679629909494e-05,
+      "loss": 0.3975,
+      "step": 26374
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1142343794603128e-05,
+      "loss": 0.2601,
+      "step": 26375
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1142007959296761e-05,
+      "loss": 0.2947,
+      "step": 26376
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1141672123990398e-05,
+      "loss": 0.1724,
+      "step": 26377
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.114133628868403e-05,
+      "loss": 0.3468,
+      "step": 26378
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1141000453377664e-05,
+      "loss": 0.2176,
+      "step": 26379
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.11406646180713e-05,
+      "loss": 0.2854,
+      "step": 26380
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1140328782764933e-05,
+      "loss": 0.2909,
+      "step": 26381
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1139992947458566e-05,
+      "loss": 0.3362,
+      "step": 26382
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1139657112152202e-05,
+      "loss": 0.256,
+      "step": 26383
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1139321276845835e-05,
+      "loss": 0.3399,
+      "step": 26384
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.113898544153947e-05,
+      "loss": 0.2646,
+      "step": 26385
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1138649606233104e-05,
+      "loss": 0.279,
+      "step": 26386
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1138313770926739e-05,
+      "loss": 0.3109,
+      "step": 26387
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1137977935620372e-05,
+      "loss": 0.2754,
+      "step": 26388
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1137642100314008e-05,
+      "loss": 0.1531,
+      "step": 26389
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1137306265007641e-05,
+      "loss": 0.2687,
+      "step": 26390
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1136970429701274e-05,
+      "loss": 0.3279,
+      "step": 26391
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.113663459439491e-05,
+      "loss": 0.4438,
+      "step": 26392
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1136298759088544e-05,
+      "loss": 0.3154,
+      "step": 26393
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1135962923782178e-05,
+      "loss": 0.314,
+      "step": 26394
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1135627088475813e-05,
+      "loss": 0.3915,
+      "step": 26395
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1135291253169448e-05,
+      "loss": 0.2434,
+      "step": 26396
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.113495541786308e-05,
+      "loss": 0.2624,
+      "step": 26397
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1134619582556717e-05,
+      "loss": 0.4042,
+      "step": 26398
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.113428374725035e-05,
+      "loss": 0.3756,
+      "step": 26399
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1133947911943983e-05,
+      "loss": 0.4219,
+      "step": 26400
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1133612076637619e-05,
+      "loss": 0.2359,
+      "step": 26401
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1133276241331252e-05,
+      "loss": 0.1991,
+      "step": 26402
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1132940406024885e-05,
+      "loss": 0.3264,
+      "step": 26403
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1132604570718521e-05,
+      "loss": 0.2402,
+      "step": 26404
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1132268735412154e-05,
+      "loss": 0.3024,
+      "step": 26405
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1131932900105789e-05,
+      "loss": 0.2494,
+      "step": 26406
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1131597064799424e-05,
+      "loss": 0.2528,
+      "step": 26407
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1131261229493058e-05,
+      "loss": 0.3175,
+      "step": 26408
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1130925394186691e-05,
+      "loss": 0.333,
+      "step": 26409
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1130589558880328e-05,
+      "loss": 0.2446,
+      "step": 26410
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.113025372357396e-05,
+      "loss": 0.1916,
+      "step": 26411
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1129917888267593e-05,
+      "loss": 0.3512,
+      "step": 26412
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.112958205296123e-05,
+      "loss": 0.206,
+      "step": 26413
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1129246217654863e-05,
+      "loss": 0.2758,
+      "step": 26414
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1128910382348497e-05,
+      "loss": 0.3066,
+      "step": 26415
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1128574547042132e-05,
+      "loss": 0.1755,
+      "step": 26416
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1128238711735767e-05,
+      "loss": 0.2816,
+      "step": 26417
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.11279028764294e-05,
+      "loss": 0.3089,
+      "step": 26418
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1127567041123036e-05,
+      "loss": 0.2915,
+      "step": 26419
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1127231205816669e-05,
+      "loss": 0.405,
+      "step": 26420
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1126895370510302e-05,
+      "loss": 0.2924,
+      "step": 26421
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1126559535203938e-05,
+      "loss": 0.2171,
+      "step": 26422
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1126223699897571e-05,
+      "loss": 0.4905,
+      "step": 26423
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1125887864591206e-05,
+      "loss": 0.3146,
+      "step": 26424
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.112555202928484e-05,
+      "loss": 0.2104,
+      "step": 26425
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1125216193978473e-05,
+      "loss": 0.3463,
+      "step": 26426
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1124880358672108e-05,
+      "loss": 0.2296,
+      "step": 26427
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1124544523365743e-05,
+      "loss": 0.1576,
+      "step": 26428
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1124208688059377e-05,
+      "loss": 0.1995,
+      "step": 26429
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.112387285275301e-05,
+      "loss": 0.2155,
+      "step": 26430
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1123537017446647e-05,
+      "loss": 0.3337,
+      "step": 26431
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.112320118214028e-05,
+      "loss": 0.3312,
+      "step": 26432
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1122865346833913e-05,
+      "loss": 0.2634,
+      "step": 26433
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1122529511527549e-05,
+      "loss": 0.2152,
+      "step": 26434
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1122193676221182e-05,
+      "loss": 0.4254,
+      "step": 26435
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1121857840914816e-05,
+      "loss": 0.3486,
+      "step": 26436
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1121522005608451e-05,
+      "loss": 0.2039,
+      "step": 26437
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1121186170302086e-05,
+      "loss": 0.2695,
+      "step": 26438
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1120850334995719e-05,
+      "loss": 0.2756,
+      "step": 26439
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1120514499689352e-05,
+      "loss": 0.349,
+      "step": 26440
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1120178664382988e-05,
+      "loss": 0.3035,
+      "step": 26441
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1119842829076621e-05,
+      "loss": 0.3116,
+      "step": 26442
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1119506993770256e-05,
+      "loss": 0.2742,
+      "step": 26443
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.111917115846389e-05,
+      "loss": 0.3495,
+      "step": 26444
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1118835323157525e-05,
+      "loss": 0.3984,
+      "step": 26445
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1118499487851158e-05,
+      "loss": 0.2065,
+      "step": 26446
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1118163652544793e-05,
+      "loss": 0.272,
+      "step": 26447
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1117827817238427e-05,
+      "loss": 0.2916,
+      "step": 26448
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.111749198193206e-05,
+      "loss": 0.2778,
+      "step": 26449
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1117156146625696e-05,
+      "loss": 0.3778,
+      "step": 26450
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.111682031131933e-05,
+      "loss": 0.1748,
+      "step": 26451
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1116484476012962e-05,
+      "loss": 0.3049,
+      "step": 26452
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1116148640706599e-05,
+      "loss": 0.3722,
+      "step": 26453
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1115812805400232e-05,
+      "loss": 0.4037,
+      "step": 26454
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1115476970093866e-05,
+      "loss": 0.3616,
+      "step": 26455
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1115141134787501e-05,
+      "loss": 0.3844,
+      "step": 26456
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1114805299481136e-05,
+      "loss": 0.2206,
+      "step": 26457
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1114469464174769e-05,
+      "loss": 0.3251,
+      "step": 26458
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1114133628868405e-05,
+      "loss": 0.2342,
+      "step": 26459
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1113797793562038e-05,
+      "loss": 0.2681,
+      "step": 26460
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.111346195825567e-05,
+      "loss": 0.3439,
+      "step": 26461
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1113126122949307e-05,
+      "loss": 0.3525,
+      "step": 26462
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.111279028764294e-05,
+      "loss": 0.2922,
+      "step": 26463
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1112454452336575e-05,
+      "loss": 0.3439,
+      "step": 26464
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.111211861703021e-05,
+      "loss": 0.3304,
+      "step": 26465
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1111782781723844e-05,
+      "loss": 0.2264,
+      "step": 26466
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1111446946417477e-05,
+      "loss": 0.2595,
+      "step": 26467
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1111111111111113e-05,
+      "loss": 0.3443,
+      "step": 26468
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1110775275804746e-05,
+      "loss": 0.1899,
+      "step": 26469
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.111043944049838e-05,
+      "loss": 0.4122,
+      "step": 26470
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1110103605192016e-05,
+      "loss": 0.2876,
+      "step": 26471
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1109767769885649e-05,
+      "loss": 0.5182,
+      "step": 26472
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1109431934579281e-05,
+      "loss": 0.2422,
+      "step": 26473
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1109096099272918e-05,
+      "loss": 0.3249,
+      "step": 26474
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.110876026396655e-05,
+      "loss": 0.1412,
+      "step": 26475
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1108424428660185e-05,
+      "loss": 0.3594,
+      "step": 26476
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.110808859335382e-05,
+      "loss": 0.2067,
+      "step": 26477
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1107752758047455e-05,
+      "loss": 0.6551,
+      "step": 26478
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1107416922741088e-05,
+      "loss": 0.4172,
+      "step": 26479
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1107081087434724e-05,
+      "loss": 0.4576,
+      "step": 26480
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1106745252128357e-05,
+      "loss": 0.2056,
+      "step": 26481
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.110640941682199e-05,
+      "loss": 0.3292,
+      "step": 26482
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1106073581515626e-05,
+      "loss": 0.2245,
+      "step": 26483
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.110573774620926e-05,
+      "loss": 0.3936,
+      "step": 26484
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1105401910902894e-05,
+      "loss": 0.2741,
+      "step": 26485
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1105066075596529e-05,
+      "loss": 0.2748,
+      "step": 26486
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1104730240290163e-05,
+      "loss": 0.269,
+      "step": 26487
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1104394404983796e-05,
+      "loss": 0.4427,
+      "step": 26488
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1104058569677432e-05,
+      "loss": 0.1554,
+      "step": 26489
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1103722734371065e-05,
+      "loss": 0.2692,
+      "step": 26490
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1103386899064698e-05,
+      "loss": 0.3446,
+      "step": 26491
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1103051063758335e-05,
+      "loss": 0.2659,
+      "step": 26492
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1102715228451968e-05,
+      "loss": 0.3293,
+      "step": 26493
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.11023793931456e-05,
+      "loss": 0.3145,
+      "step": 26494
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1102043557839237e-05,
+      "loss": 0.2421,
+      "step": 26495
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.110170772253287e-05,
+      "loss": 0.4182,
+      "step": 26496
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1101371887226505e-05,
+      "loss": 0.4928,
+      "step": 26497
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.110103605192014e-05,
+      "loss": 0.259,
+      "step": 26498
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1100700216613774e-05,
+      "loss": 0.328,
+      "step": 26499
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1100364381307407e-05,
+      "loss": 0.1857,
+      "step": 26500
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.1100028546001043e-05,
+      "loss": 0.3369,
+      "step": 26501
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1099692710694676e-05,
+      "loss": 0.2278,
+      "step": 26502
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1099356875388309e-05,
+      "loss": 0.2342,
+      "step": 26503
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1099021040081945e-05,
+      "loss": 0.3038,
+      "step": 26504
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1098685204775578e-05,
+      "loss": 0.1859,
+      "step": 26505
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1098349369469213e-05,
+      "loss": 0.1829,
+      "step": 26506
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1098013534162848e-05,
+      "loss": 0.2636,
+      "step": 26507
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1097677698856482e-05,
+      "loss": 0.4219,
+      "step": 26508
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1097341863550115e-05,
+      "loss": 0.285,
+      "step": 26509
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1097006028243752e-05,
+      "loss": 0.3674,
+      "step": 26510
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1096670192937385e-05,
+      "loss": 0.1641,
+      "step": 26511
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1096334357631017e-05,
+      "loss": 0.179,
+      "step": 26512
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1095998522324654e-05,
+      "loss": 0.241,
+      "step": 26513
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1095662687018287e-05,
+      "loss": 0.3968,
+      "step": 26514
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.109532685171192e-05,
+      "loss": 0.3943,
+      "step": 26515
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1094991016405556e-05,
+      "loss": 0.3684,
+      "step": 26516
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1094655181099189e-05,
+      "loss": 0.3003,
+      "step": 26517
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1094319345792824e-05,
+      "loss": 0.3484,
+      "step": 26518
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1093983510486458e-05,
+      "loss": 0.2447,
+      "step": 26519
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1093647675180093e-05,
+      "loss": 0.3123,
+      "step": 26520
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1093311839873726e-05,
+      "loss": 0.4351,
+      "step": 26521
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1092976004567362e-05,
+      "loss": 0.4159,
+      "step": 26522
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1092640169260995e-05,
+      "loss": 0.2614,
+      "step": 26523
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1092304333954628e-05,
+      "loss": 0.3426,
+      "step": 26524
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1091968498648265e-05,
+      "loss": 0.2679,
+      "step": 26525
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1091632663341897e-05,
+      "loss": 0.3294,
+      "step": 26526
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1091296828035532e-05,
+      "loss": 0.1964,
+      "step": 26527
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1090960992729167e-05,
+      "loss": 0.2192,
+      "step": 26528
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1090625157422801e-05,
+      "loss": 0.4063,
+      "step": 26529
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1090289322116434e-05,
+      "loss": 0.2821,
+      "step": 26530
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.108995348681007e-05,
+      "loss": 0.3646,
+      "step": 26531
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1089617651503704e-05,
+      "loss": 0.3666,
+      "step": 26532
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1089281816197337e-05,
+      "loss": 0.3194,
+      "step": 26533
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1088945980890973e-05,
+      "loss": 0.2627,
+      "step": 26534
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1088610145584606e-05,
+      "loss": 0.4143,
+      "step": 26535
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.108827431027824e-05,
+      "loss": 0.2939,
+      "step": 26536
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1087938474971875e-05,
+      "loss": 0.2824,
+      "step": 26537
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1087602639665508e-05,
+      "loss": 0.1629,
+      "step": 26538
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1087266804359143e-05,
+      "loss": 0.4194,
+      "step": 26539
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1086930969052777e-05,
+      "loss": 0.1582,
+      "step": 26540
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1086595133746412e-05,
+      "loss": 0.2501,
+      "step": 26541
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1086259298440045e-05,
+      "loss": 0.2955,
+      "step": 26542
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1085923463133681e-05,
+      "loss": 0.4197,
+      "step": 26543
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1085587627827314e-05,
+      "loss": 0.2433,
+      "step": 26544
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1085251792520947e-05,
+      "loss": 0.3828,
+      "step": 26545
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1084915957214584e-05,
+      "loss": 0.2252,
+      "step": 26546
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1084580121908217e-05,
+      "loss": 0.2718,
+      "step": 26547
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1084244286601851e-05,
+      "loss": 0.2447,
+      "step": 26548
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1083908451295486e-05,
+      "loss": 0.3708,
+      "step": 26549
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.108357261598912e-05,
+      "loss": 0.2407,
+      "step": 26550
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1083236780682753e-05,
+      "loss": 0.2923,
+      "step": 26551
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.108290094537639e-05,
+      "loss": 0.2149,
+      "step": 26552
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1082565110070023e-05,
+      "loss": 0.2246,
+      "step": 26553
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1082229274763656e-05,
+      "loss": 0.1686,
+      "step": 26554
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1081893439457292e-05,
+      "loss": 0.2398,
+      "step": 26555
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1081557604150925e-05,
+      "loss": 0.2584,
+      "step": 26556
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.108122176884456e-05,
+      "loss": 0.2327,
+      "step": 26557
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1080885933538194e-05,
+      "loss": 0.1797,
+      "step": 26558
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1080550098231829e-05,
+      "loss": 0.2902,
+      "step": 26559
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1080214262925462e-05,
+      "loss": 0.2736,
+      "step": 26560
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1079878427619097e-05,
+      "loss": 0.4253,
+      "step": 26561
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1079542592312731e-05,
+      "loss": 0.2845,
+      "step": 26562
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1079206757006364e-05,
+      "loss": 0.3603,
+      "step": 26563
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.10788709217e-05,
+      "loss": 0.3003,
+      "step": 26564
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1078535086393633e-05,
+      "loss": 0.3812,
+      "step": 26565
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1078199251087266e-05,
+      "loss": 0.3168,
+      "step": 26566
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1077863415780903e-05,
+      "loss": 0.2406,
+      "step": 26567
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1077527580474536e-05,
+      "loss": 0.3176,
+      "step": 26568
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.107719174516817e-05,
+      "loss": 0.3774,
+      "step": 26569
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1076855909861805e-05,
+      "loss": 0.2647,
+      "step": 26570
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.107652007455544e-05,
+      "loss": 0.2722,
+      "step": 26571
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1076184239249073e-05,
+      "loss": 0.4009,
+      "step": 26572
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1075848403942709e-05,
+      "loss": 0.3244,
+      "step": 26573
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1075512568636342e-05,
+      "loss": 0.3826,
+      "step": 26574
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1075176733329975e-05,
+      "loss": 0.213,
+      "step": 26575
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1074840898023611e-05,
+      "loss": 0.2419,
+      "step": 26576
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1074505062717244e-05,
+      "loss": 0.3511,
+      "step": 26577
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1074169227410879e-05,
+      "loss": 0.1769,
+      "step": 26578
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1073833392104513e-05,
+      "loss": 0.4013,
+      "step": 26579
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1073497556798148e-05,
+      "loss": 0.4513,
+      "step": 26580
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1073161721491781e-05,
+      "loss": 0.3241,
+      "step": 26581
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1072825886185416e-05,
+      "loss": 0.3382,
+      "step": 26582
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.107249005087905e-05,
+      "loss": 0.2683,
+      "step": 26583
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1072154215572683e-05,
+      "loss": 0.12,
+      "step": 26584
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.107181838026632e-05,
+      "loss": 0.3013,
+      "step": 26585
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1071482544959953e-05,
+      "loss": 0.2436,
+      "step": 26586
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1071146709653586e-05,
+      "loss": 0.1955,
+      "step": 26587
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1070810874347222e-05,
+      "loss": 0.2966,
+      "step": 26588
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1070475039040855e-05,
+      "loss": 0.2479,
+      "step": 26589
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.107013920373449e-05,
+      "loss": 0.3406,
+      "step": 26590
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1069803368428124e-05,
+      "loss": 0.3239,
+      "step": 26591
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1069467533121759e-05,
+      "loss": 0.2407,
+      "step": 26592
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1069131697815392e-05,
+      "loss": 0.3188,
+      "step": 26593
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1068795862509028e-05,
+      "loss": 0.1958,
+      "step": 26594
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1068460027202661e-05,
+      "loss": 0.2878,
+      "step": 26595
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1068124191896294e-05,
+      "loss": 0.3454,
+      "step": 26596
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.106778835658993e-05,
+      "loss": 0.2631,
+      "step": 26597
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1067452521283563e-05,
+      "loss": 0.4321,
+      "step": 26598
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1067116685977198e-05,
+      "loss": 0.2941,
+      "step": 26599
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1066780850670833e-05,
+      "loss": 0.5184,
+      "step": 26600
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1066445015364467e-05,
+      "loss": 0.4736,
+      "step": 26601
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.10661091800581e-05,
+      "loss": 0.3675,
+      "step": 26602
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1065773344751737e-05,
+      "loss": 0.3622,
+      "step": 26603
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.106543750944537e-05,
+      "loss": 0.3693,
+      "step": 26604
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1065101674139002e-05,
+      "loss": 0.1535,
+      "step": 26605
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1064765838832639e-05,
+      "loss": 0.2668,
+      "step": 26606
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1064430003526272e-05,
+      "loss": 0.3971,
+      "step": 26607
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1064094168219905e-05,
+      "loss": 0.2409,
+      "step": 26608
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1063758332913541e-05,
+      "loss": 0.29,
+      "step": 26609
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1063422497607174e-05,
+      "loss": 0.3142,
+      "step": 26610
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1063086662300809e-05,
+      "loss": 0.1846,
+      "step": 26611
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1062750826994443e-05,
+      "loss": 0.5121,
+      "step": 26612
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1062414991688078e-05,
+      "loss": 0.3166,
+      "step": 26613
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1062079156381711e-05,
+      "loss": 0.2656,
+      "step": 26614
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1061743321075347e-05,
+      "loss": 0.374,
+      "step": 26615
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.106140748576898e-05,
+      "loss": 0.2543,
+      "step": 26616
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1061071650462613e-05,
+      "loss": 0.3639,
+      "step": 26617
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.106073581515625e-05,
+      "loss": 0.3609,
+      "step": 26618
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1060399979849882e-05,
+      "loss": 0.3746,
+      "step": 26619
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1060064144543517e-05,
+      "loss": 0.1525,
+      "step": 26620
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.105972830923715e-05,
+      "loss": 0.3334,
+      "step": 26621
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1059392473930786e-05,
+      "loss": 0.2587,
+      "step": 26622
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.105905663862442e-05,
+      "loss": 0.3308,
+      "step": 26623
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1058720803318052e-05,
+      "loss": 0.2244,
+      "step": 26624
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1058384968011689e-05,
+      "loss": 0.188,
+      "step": 26625
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1058049132705322e-05,
+      "loss": 0.196,
+      "step": 26626
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1057713297398956e-05,
+      "loss": 0.2516,
+      "step": 26627
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1057377462092591e-05,
+      "loss": 0.3995,
+      "step": 26628
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1057041626786224e-05,
+      "loss": 0.2711,
+      "step": 26629
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1056705791479858e-05,
+      "loss": 0.3881,
+      "step": 26630
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1056369956173493e-05,
+      "loss": 0.2798,
+      "step": 26631
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1056034120867128e-05,
+      "loss": 0.1967,
+      "step": 26632
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.105569828556076e-05,
+      "loss": 0.2126,
+      "step": 26633
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1055362450254397e-05,
+      "loss": 0.4146,
+      "step": 26634
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.105502661494803e-05,
+      "loss": 0.3043,
+      "step": 26635
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1054690779641663e-05,
+      "loss": 0.3754,
+      "step": 26636
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.10543549443353e-05,
+      "loss": 0.3322,
+      "step": 26637
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1054019109028932e-05,
+      "loss": 0.4061,
+      "step": 26638
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1053683273722567e-05,
+      "loss": 0.3915,
+      "step": 26639
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1053347438416202e-05,
+      "loss": 0.3263,
+      "step": 26640
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1053011603109836e-05,
+      "loss": 0.3178,
+      "step": 26641
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1052675767803469e-05,
+      "loss": 0.3542,
+      "step": 26642
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1052339932497105e-05,
+      "loss": 0.336,
+      "step": 26643
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1052004097190738e-05,
+      "loss": 0.2522,
+      "step": 26644
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1051668261884371e-05,
+      "loss": 0.1936,
+      "step": 26645
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1051332426578008e-05,
+      "loss": 0.2975,
+      "step": 26646
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.105099659127164e-05,
+      "loss": 0.2799,
+      "step": 26647
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1050660755965275e-05,
+      "loss": 0.2181,
+      "step": 26648
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.105032492065891e-05,
+      "loss": 0.2631,
+      "step": 26649
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1049989085352545e-05,
+      "loss": 0.4256,
+      "step": 26650
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1049653250046178e-05,
+      "loss": 0.185,
+      "step": 26651
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1049317414739812e-05,
+      "loss": 0.217,
+      "step": 26652
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1048981579433447e-05,
+      "loss": 0.1852,
+      "step": 26653
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.104864574412708e-05,
+      "loss": 0.3802,
+      "step": 26654
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1048309908820716e-05,
+      "loss": 0.39,
+      "step": 26655
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1047974073514349e-05,
+      "loss": 0.3704,
+      "step": 26656
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1047638238207982e-05,
+      "loss": 0.2846,
+      "step": 26657
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1047302402901618e-05,
+      "loss": 0.2698,
+      "step": 26658
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1046966567595251e-05,
+      "loss": 0.2549,
+      "step": 26659
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1046630732288886e-05,
+      "loss": 0.2369,
+      "step": 26660
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.104629489698252e-05,
+      "loss": 0.2164,
+      "step": 26661
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1045959061676155e-05,
+      "loss": 0.3407,
+      "step": 26662
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1045623226369788e-05,
+      "loss": 0.358,
+      "step": 26663
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1045287391063425e-05,
+      "loss": 0.2655,
+      "step": 26664
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1044951555757058e-05,
+      "loss": 0.2668,
+      "step": 26665
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.104461572045069e-05,
+      "loss": 0.2023,
+      "step": 26666
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1044279885144327e-05,
+      "loss": 0.383,
+      "step": 26667
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.104394404983796e-05,
+      "loss": 0.3358,
+      "step": 26668
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1043608214531594e-05,
+      "loss": 0.3385,
+      "step": 26669
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1043272379225229e-05,
+      "loss": 0.354,
+      "step": 26670
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1042936543918864e-05,
+      "loss": 0.1422,
+      "step": 26671
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1042600708612497e-05,
+      "loss": 0.3167,
+      "step": 26672
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1042264873306131e-05,
+      "loss": 0.2605,
+      "step": 26673
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1041929037999766e-05,
+      "loss": 0.1969,
+      "step": 26674
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1041593202693399e-05,
+      "loss": 0.4115,
+      "step": 26675
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1041257367387035e-05,
+      "loss": 0.4381,
+      "step": 26676
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1040921532080668e-05,
+      "loss": 0.2841,
+      "step": 26677
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1040585696774301e-05,
+      "loss": 0.3058,
+      "step": 26678
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1040249861467938e-05,
+      "loss": 0.1926,
+      "step": 26679
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.103991402616157e-05,
+      "loss": 0.3015,
+      "step": 26680
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1039578190855205e-05,
+      "loss": 0.5033,
+      "step": 26681
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.103924235554884e-05,
+      "loss": 0.2007,
+      "step": 26682
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1038906520242474e-05,
+      "loss": 0.2415,
+      "step": 26683
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1038570684936107e-05,
+      "loss": 0.3761,
+      "step": 26684
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1038234849629744e-05,
+      "loss": 0.2668,
+      "step": 26685
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1037899014323377e-05,
+      "loss": 0.2219,
+      "step": 26686
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.103756317901701e-05,
+      "loss": 0.2181,
+      "step": 26687
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1037227343710646e-05,
+      "loss": 0.4137,
+      "step": 26688
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1036891508404279e-05,
+      "loss": 0.2125,
+      "step": 26689
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1036555673097914e-05,
+      "loss": 0.401,
+      "step": 26690
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1036219837791548e-05,
+      "loss": 0.2094,
+      "step": 26691
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1035884002485183e-05,
+      "loss": 0.1635,
+      "step": 26692
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1035548167178816e-05,
+      "loss": 0.1159,
+      "step": 26693
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1035212331872452e-05,
+      "loss": 0.2102,
+      "step": 26694
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1034876496566085e-05,
+      "loss": 0.4147,
+      "step": 26695
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1034540661259718e-05,
+      "loss": 0.2931,
+      "step": 26696
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1034204825953354e-05,
+      "loss": 0.4032,
+      "step": 26697
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.1033868990646987e-05,
+      "loss": 0.2675,
+      "step": 26698
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 1.103353315534062e-05,
+      "loss": 0.2391,
+      "step": 26699
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1033197320034257e-05,
+      "loss": 0.5682,
+      "step": 26700
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.103286148472789e-05,
+      "loss": 0.2205,
+      "step": 26701
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1032525649421524e-05,
+      "loss": 0.3422,
+      "step": 26702
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1032189814115159e-05,
+      "loss": 0.3109,
+      "step": 26703
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1031853978808794e-05,
+      "loss": 0.3306,
+      "step": 26704
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1031518143502427e-05,
+      "loss": 0.359,
+      "step": 26705
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1031182308196063e-05,
+      "loss": 0.3334,
+      "step": 26706
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1030846472889696e-05,
+      "loss": 0.3969,
+      "step": 26707
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1030510637583329e-05,
+      "loss": 0.3289,
+      "step": 26708
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1030174802276965e-05,
+      "loss": 0.3057,
+      "step": 26709
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1029838966970598e-05,
+      "loss": 0.3012,
+      "step": 26710
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1029503131664233e-05,
+      "loss": 0.2913,
+      "step": 26711
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1029167296357867e-05,
+      "loss": 0.4502,
+      "step": 26712
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1028831461051502e-05,
+      "loss": 0.3381,
+      "step": 26713
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1028495625745135e-05,
+      "loss": 0.2384,
+      "step": 26714
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1028159790438771e-05,
+      "loss": 0.2569,
+      "step": 26715
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1027823955132404e-05,
+      "loss": 0.4023,
+      "step": 26716
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1027488119826037e-05,
+      "loss": 0.4207,
+      "step": 26717
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1027152284519674e-05,
+      "loss": 0.2877,
+      "step": 26718
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1026816449213306e-05,
+      "loss": 0.4509,
+      "step": 26719
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.102648061390694e-05,
+      "loss": 0.1934,
+      "step": 26720
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1026144778600576e-05,
+      "loss": 0.1909,
+      "step": 26721
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1025808943294209e-05,
+      "loss": 0.2847,
+      "step": 26722
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1025473107987843e-05,
+      "loss": 0.3931,
+      "step": 26723
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1025137272681478e-05,
+      "loss": 0.4054,
+      "step": 26724
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1024801437375113e-05,
+      "loss": 0.3321,
+      "step": 26725
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1024465602068746e-05,
+      "loss": 0.2387,
+      "step": 26726
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1024129766762382e-05,
+      "loss": 0.4649,
+      "step": 26727
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1023793931456015e-05,
+      "loss": 0.1911,
+      "step": 26728
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1023458096149648e-05,
+      "loss": 0.338,
+      "step": 26729
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1023122260843284e-05,
+      "loss": 0.2633,
+      "step": 26730
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1022786425536917e-05,
+      "loss": 0.2919,
+      "step": 26731
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1022450590230552e-05,
+      "loss": 0.2641,
+      "step": 26732
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1022114754924186e-05,
+      "loss": 0.1893,
+      "step": 26733
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1021778919617821e-05,
+      "loss": 0.2636,
+      "step": 26734
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1021443084311454e-05,
+      "loss": 0.4787,
+      "step": 26735
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.102110724900509e-05,
+      "loss": 0.2668,
+      "step": 26736
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1020771413698723e-05,
+      "loss": 0.3822,
+      "step": 26737
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1020435578392356e-05,
+      "loss": 0.3075,
+      "step": 26738
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1020099743085993e-05,
+      "loss": 0.3943,
+      "step": 26739
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1019763907779626e-05,
+      "loss": 0.2998,
+      "step": 26740
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1019428072473259e-05,
+      "loss": 0.3066,
+      "step": 26741
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1019092237166895e-05,
+      "loss": 0.3062,
+      "step": 26742
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1018756401860528e-05,
+      "loss": 0.5036,
+      "step": 26743
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1018420566554163e-05,
+      "loss": 0.3093,
+      "step": 26744
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1018084731247797e-05,
+      "loss": 0.257,
+      "step": 26745
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1017748895941432e-05,
+      "loss": 0.1395,
+      "step": 26746
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1017413060635065e-05,
+      "loss": 0.3778,
+      "step": 26747
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1017077225328701e-05,
+      "loss": 0.2975,
+      "step": 26748
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1016741390022334e-05,
+      "loss": 0.3876,
+      "step": 26749
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1016405554715967e-05,
+      "loss": 0.2173,
+      "step": 26750
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1016069719409603e-05,
+      "loss": 0.3504,
+      "step": 26751
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1015733884103236e-05,
+      "loss": 0.3258,
+      "step": 26752
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1015398048796871e-05,
+      "loss": 0.3126,
+      "step": 26753
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1015062213490506e-05,
+      "loss": 0.2868,
+      "step": 26754
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.101472637818414e-05,
+      "loss": 0.3198,
+      "step": 26755
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1014390542877773e-05,
+      "loss": 0.3032,
+      "step": 26756
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.101405470757141e-05,
+      "loss": 0.2005,
+      "step": 26757
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1013718872265042e-05,
+      "loss": 0.3099,
+      "step": 26758
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1013383036958675e-05,
+      "loss": 0.2815,
+      "step": 26759
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1013047201652312e-05,
+      "loss": 0.2623,
+      "step": 26760
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1012711366345945e-05,
+      "loss": 0.3124,
+      "step": 26761
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.101237553103958e-05,
+      "loss": 0.2622,
+      "step": 26762
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1012039695733214e-05,
+      "loss": 0.3836,
+      "step": 26763
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1011703860426847e-05,
+      "loss": 0.5402,
+      "step": 26764
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1011368025120482e-05,
+      "loss": 0.3322,
+      "step": 26765
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1011032189814116e-05,
+      "loss": 0.2598,
+      "step": 26766
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1010696354507751e-05,
+      "loss": 0.3774,
+      "step": 26767
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1010360519201384e-05,
+      "loss": 0.2661,
+      "step": 26768
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.101002468389502e-05,
+      "loss": 0.4351,
+      "step": 26769
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1009688848588653e-05,
+      "loss": 0.1215,
+      "step": 26770
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1009353013282286e-05,
+      "loss": 0.311,
+      "step": 26771
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1009017177975922e-05,
+      "loss": 0.3291,
+      "step": 26772
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1008681342669555e-05,
+      "loss": 0.2014,
+      "step": 26773
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.100834550736319e-05,
+      "loss": 0.3378,
+      "step": 26774
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1008009672056825e-05,
+      "loss": 0.3125,
+      "step": 26775
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.100767383675046e-05,
+      "loss": 0.343,
+      "step": 26776
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1007338001444092e-05,
+      "loss": 0.3351,
+      "step": 26777
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1007002166137729e-05,
+      "loss": 0.359,
+      "step": 26778
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1006666330831362e-05,
+      "loss": 0.3639,
+      "step": 26779
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1006330495524995e-05,
+      "loss": 0.2611,
+      "step": 26780
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1005994660218631e-05,
+      "loss": 0.4522,
+      "step": 26781
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1005658824912264e-05,
+      "loss": 0.2224,
+      "step": 26782
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1005322989605899e-05,
+      "loss": 0.4231,
+      "step": 26783
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1004987154299533e-05,
+      "loss": 0.2392,
+      "step": 26784
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1004651318993168e-05,
+      "loss": 0.3269,
+      "step": 26785
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.10043154836868e-05,
+      "loss": 0.2926,
+      "step": 26786
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1003979648380435e-05,
+      "loss": 0.4235,
+      "step": 26787
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.100364381307407e-05,
+      "loss": 0.3266,
+      "step": 26788
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1003307977767703e-05,
+      "loss": 0.3661,
+      "step": 26789
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.100297214246134e-05,
+      "loss": 0.319,
+      "step": 26790
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1002636307154972e-05,
+      "loss": 0.2499,
+      "step": 26791
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1002300471848605e-05,
+      "loss": 0.3419,
+      "step": 26792
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1001964636542242e-05,
+      "loss": 0.2769,
+      "step": 26793
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1001628801235875e-05,
+      "loss": 0.3399,
+      "step": 26794
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.100129296592951e-05,
+      "loss": 0.2348,
+      "step": 26795
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1000957130623144e-05,
+      "loss": 0.1754,
+      "step": 26796
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1000621295316778e-05,
+      "loss": 0.1694,
+      "step": 26797
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.1000285460010411e-05,
+      "loss": 0.2628,
+      "step": 26798
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0999949624704044e-05,
+      "loss": 0.1417,
+      "step": 26799
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.099961378939768e-05,
+      "loss": 0.2036,
+      "step": 26800
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0999277954091314e-05,
+      "loss": 0.2418,
+      "step": 26801
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0998942118784948e-05,
+      "loss": 0.2189,
+      "step": 26802
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0998606283478583e-05,
+      "loss": 0.2547,
+      "step": 26803
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0998270448172218e-05,
+      "loss": 0.345,
+      "step": 26804
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.099793461286585e-05,
+      "loss": 0.3329,
+      "step": 26805
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0997598777559487e-05,
+      "loss": 0.2351,
+      "step": 26806
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.099726294225312e-05,
+      "loss": 0.2349,
+      "step": 26807
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0996927106946753e-05,
+      "loss": 0.244,
+      "step": 26808
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.099659127164039e-05,
+      "loss": 0.2173,
+      "step": 26809
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0996255436334022e-05,
+      "loss": 0.3369,
+      "step": 26810
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0995919601027655e-05,
+      "loss": 0.2758,
+      "step": 26811
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0995583765721291e-05,
+      "loss": 0.335,
+      "step": 26812
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0995247930414924e-05,
+      "loss": 0.2374,
+      "step": 26813
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0994912095108559e-05,
+      "loss": 0.3203,
+      "step": 26814
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0994576259802194e-05,
+      "loss": 0.2206,
+      "step": 26815
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0994240424495828e-05,
+      "loss": 0.4318,
+      "step": 26816
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0993904589189461e-05,
+      "loss": 0.2608,
+      "step": 26817
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0993568753883098e-05,
+      "loss": 0.2294,
+      "step": 26818
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.099323291857673e-05,
+      "loss": 0.2513,
+      "step": 26819
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0992897083270364e-05,
+      "loss": 0.3006,
+      "step": 26820
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0992561247964e-05,
+      "loss": 0.4051,
+      "step": 26821
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0992225412657633e-05,
+      "loss": 0.3423,
+      "step": 26822
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0991889577351267e-05,
+      "loss": 0.2234,
+      "step": 26823
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0991553742044902e-05,
+      "loss": 0.3186,
+      "step": 26824
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0991217906738537e-05,
+      "loss": 0.3915,
+      "step": 26825
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.099088207143217e-05,
+      "loss": 0.2213,
+      "step": 26826
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0990546236125806e-05,
+      "loss": 0.3311,
+      "step": 26827
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0990210400819439e-05,
+      "loss": 0.378,
+      "step": 26828
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0989874565513072e-05,
+      "loss": 0.4498,
+      "step": 26829
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0989538730206708e-05,
+      "loss": 0.2324,
+      "step": 26830
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0989202894900341e-05,
+      "loss": 0.5127,
+      "step": 26831
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0988867059593974e-05,
+      "loss": 0.1415,
+      "step": 26832
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.098853122428761e-05,
+      "loss": 0.2688,
+      "step": 26833
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0988195388981244e-05,
+      "loss": 0.3657,
+      "step": 26834
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0987859553674878e-05,
+      "loss": 0.1959,
+      "step": 26835
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0987523718368513e-05,
+      "loss": 0.2733,
+      "step": 26836
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0987187883062147e-05,
+      "loss": 0.3475,
+      "step": 26837
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.098685204775578e-05,
+      "loss": 0.3267,
+      "step": 26838
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0986516212449417e-05,
+      "loss": 0.2402,
+      "step": 26839
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.098618037714305e-05,
+      "loss": 0.3179,
+      "step": 26840
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0985844541836683e-05,
+      "loss": 0.3687,
+      "step": 26841
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0985508706530319e-05,
+      "loss": 0.3523,
+      "step": 26842
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0985172871223952e-05,
+      "loss": 0.3293,
+      "step": 26843
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0984837035917587e-05,
+      "loss": 0.189,
+      "step": 26844
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0984501200611221e-05,
+      "loss": 0.3895,
+      "step": 26845
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0984165365304856e-05,
+      "loss": 0.2586,
+      "step": 26846
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0983829529998489e-05,
+      "loss": 0.2736,
+      "step": 26847
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0983493694692125e-05,
+      "loss": 0.2652,
+      "step": 26848
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0983157859385758e-05,
+      "loss": 0.1815,
+      "step": 26849
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0982822024079391e-05,
+      "loss": 0.2071,
+      "step": 26850
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0982486188773027e-05,
+      "loss": 0.3606,
+      "step": 26851
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.098215035346666e-05,
+      "loss": 0.266,
+      "step": 26852
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0981814518160295e-05,
+      "loss": 0.504,
+      "step": 26853
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.098147868285393e-05,
+      "loss": 0.2589,
+      "step": 26854
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0981142847547563e-05,
+      "loss": 0.3136,
+      "step": 26855
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0980807012241197e-05,
+      "loss": 0.2754,
+      "step": 26856
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0980471176934832e-05,
+      "loss": 0.2927,
+      "step": 26857
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0980135341628467e-05,
+      "loss": 0.3286,
+      "step": 26858
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.09797995063221e-05,
+      "loss": 0.3313,
+      "step": 26859
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0979463671015736e-05,
+      "loss": 0.2709,
+      "step": 26860
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0979127835709369e-05,
+      "loss": 0.3425,
+      "step": 26861
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0978792000403002e-05,
+      "loss": 0.2784,
+      "step": 26862
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0978456165096638e-05,
+      "loss": 0.1959,
+      "step": 26863
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0978120329790271e-05,
+      "loss": 0.3664,
+      "step": 26864
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0977784494483906e-05,
+      "loss": 0.1742,
+      "step": 26865
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.097744865917754e-05,
+      "loss": 0.4734,
+      "step": 26866
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0977112823871175e-05,
+      "loss": 0.3205,
+      "step": 26867
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0976776988564808e-05,
+      "loss": 0.6203,
+      "step": 26868
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0976441153258444e-05,
+      "loss": 0.1985,
+      "step": 26869
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0976105317952077e-05,
+      "loss": 0.3334,
+      "step": 26870
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.097576948264571e-05,
+      "loss": 0.1376,
+      "step": 26871
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0975433647339347e-05,
+      "loss": 0.2188,
+      "step": 26872
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.097509781203298e-05,
+      "loss": 0.2568,
+      "step": 26873
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0974761976726614e-05,
+      "loss": 0.2566,
+      "step": 26874
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0974426141420249e-05,
+      "loss": 0.3524,
+      "step": 26875
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0974090306113882e-05,
+      "loss": 0.278,
+      "step": 26876
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0973754470807516e-05,
+      "loss": 0.433,
+      "step": 26877
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0973418635501151e-05,
+      "loss": 0.2495,
+      "step": 26878
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0973082800194786e-05,
+      "loss": 0.261,
+      "step": 26879
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0972746964888419e-05,
+      "loss": 0.5112,
+      "step": 26880
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0972411129582055e-05,
+      "loss": 0.4382,
+      "step": 26881
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0972075294275688e-05,
+      "loss": 0.3091,
+      "step": 26882
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0971739458969321e-05,
+      "loss": 0.3883,
+      "step": 26883
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0971403623662957e-05,
+      "loss": 0.3363,
+      "step": 26884
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.097106778835659e-05,
+      "loss": 0.3354,
+      "step": 26885
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0970731953050225e-05,
+      "loss": 0.3391,
+      "step": 26886
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.097039611774386e-05,
+      "loss": 0.175,
+      "step": 26887
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0970060282437494e-05,
+      "loss": 0.4235,
+      "step": 26888
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0969724447131127e-05,
+      "loss": 0.339,
+      "step": 26889
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0969388611824763e-05,
+      "loss": 0.1507,
+      "step": 26890
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0969052776518396e-05,
+      "loss": 0.3647,
+      "step": 26891
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.096871694121203e-05,
+      "loss": 0.2033,
+      "step": 26892
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0968381105905666e-05,
+      "loss": 0.2851,
+      "step": 26893
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0968045270599299e-05,
+      "loss": 0.4979,
+      "step": 26894
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0967709435292933e-05,
+      "loss": 0.5416,
+      "step": 26895
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0967373599986568e-05,
+      "loss": 0.2803,
+      "step": 26896
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0967037764680203e-05,
+      "loss": 0.3459,
+      "step": 26897
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 1.0966701929373836e-05,
+      "loss": 0.3896,
+      "step": 26898
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.096636609406747e-05,
+      "loss": 0.304,
+      "step": 26899
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0966030258761105e-05,
+      "loss": 0.3074,
+      "step": 26900
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0965694423454738e-05,
+      "loss": 0.2742,
+      "step": 26901
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0965358588148374e-05,
+      "loss": 0.3548,
+      "step": 26902
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0965022752842007e-05,
+      "loss": 0.4927,
+      "step": 26903
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.096468691753564e-05,
+      "loss": 0.4266,
+      "step": 26904
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0964351082229276e-05,
+      "loss": 0.2697,
+      "step": 26905
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.096401524692291e-05,
+      "loss": 0.2473,
+      "step": 26906
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0963679411616544e-05,
+      "loss": 0.2979,
+      "step": 26907
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0963343576310179e-05,
+      "loss": 0.3046,
+      "step": 26908
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0963007741003813e-05,
+      "loss": 0.191,
+      "step": 26909
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0962671905697446e-05,
+      "loss": 0.2602,
+      "step": 26910
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0962336070391083e-05,
+      "loss": 0.4182,
+      "step": 26911
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0962000235084716e-05,
+      "loss": 0.5234,
+      "step": 26912
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0961664399778348e-05,
+      "loss": 0.1895,
+      "step": 26913
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0961328564471985e-05,
+      "loss": 0.2928,
+      "step": 26914
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0960992729165618e-05,
+      "loss": 0.3987,
+      "step": 26915
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0960656893859252e-05,
+      "loss": 0.47,
+      "step": 26916
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0960321058552887e-05,
+      "loss": 0.2336,
+      "step": 26917
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0959985223246522e-05,
+      "loss": 0.1648,
+      "step": 26918
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0959649387940155e-05,
+      "loss": 0.2581,
+      "step": 26919
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0959313552633791e-05,
+      "loss": 0.2867,
+      "step": 26920
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0958977717327424e-05,
+      "loss": 0.2327,
+      "step": 26921
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0958641882021057e-05,
+      "loss": 0.4504,
+      "step": 26922
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0958306046714693e-05,
+      "loss": 0.1695,
+      "step": 26923
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0957970211408326e-05,
+      "loss": 0.1352,
+      "step": 26924
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0957634376101959e-05,
+      "loss": 0.2243,
+      "step": 26925
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0957298540795595e-05,
+      "loss": 0.2193,
+      "step": 26926
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0956962705489228e-05,
+      "loss": 0.3386,
+      "step": 26927
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0956626870182863e-05,
+      "loss": 0.2931,
+      "step": 26928
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0956291034876498e-05,
+      "loss": 0.2753,
+      "step": 26929
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0955955199570132e-05,
+      "loss": 0.3852,
+      "step": 26930
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0955619364263765e-05,
+      "loss": 0.3202,
+      "step": 26931
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0955283528957402e-05,
+      "loss": 0.2083,
+      "step": 26932
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0954947693651035e-05,
+      "loss": 0.4093,
+      "step": 26933
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0954611858344668e-05,
+      "loss": 0.2697,
+      "step": 26934
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0954276023038304e-05,
+      "loss": 0.42,
+      "step": 26935
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0953940187731937e-05,
+      "loss": 0.4431,
+      "step": 26936
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0953604352425572e-05,
+      "loss": 0.2683,
+      "step": 26937
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0953268517119206e-05,
+      "loss": 0.4387,
+      "step": 26938
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.095293268181284e-05,
+      "loss": 0.2553,
+      "step": 26939
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0952596846506474e-05,
+      "loss": 0.3147,
+      "step": 26940
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.095226101120011e-05,
+      "loss": 0.1435,
+      "step": 26941
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0951925175893743e-05,
+      "loss": 0.286,
+      "step": 26942
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0951589340587376e-05,
+      "loss": 0.345,
+      "step": 26943
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0951253505281012e-05,
+      "loss": 0.4224,
+      "step": 26944
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0950917669974645e-05,
+      "loss": 0.3117,
+      "step": 26945
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0950581834668278e-05,
+      "loss": 0.2845,
+      "step": 26946
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0950245999361915e-05,
+      "loss": 0.2377,
+      "step": 26947
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0949910164055548e-05,
+      "loss": 0.3056,
+      "step": 26948
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0949574328749182e-05,
+      "loss": 0.1934,
+      "step": 26949
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0949238493442817e-05,
+      "loss": 0.266,
+      "step": 26950
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0948902658136452e-05,
+      "loss": 0.5282,
+      "step": 26951
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0948566822830084e-05,
+      "loss": 0.1947,
+      "step": 26952
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.094823098752372e-05,
+      "loss": 0.4154,
+      "step": 26953
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0947895152217354e-05,
+      "loss": 0.2904,
+      "step": 26954
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0947559316910987e-05,
+      "loss": 0.4017,
+      "step": 26955
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0947223481604623e-05,
+      "loss": 0.3676,
+      "step": 26956
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0946887646298256e-05,
+      "loss": 0.2838,
+      "step": 26957
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.094655181099189e-05,
+      "loss": 0.1742,
+      "step": 26958
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0946215975685525e-05,
+      "loss": 0.1973,
+      "step": 26959
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.094588014037916e-05,
+      "loss": 0.2853,
+      "step": 26960
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0945544305072793e-05,
+      "loss": 0.1496,
+      "step": 26961
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.094520846976643e-05,
+      "loss": 0.2493,
+      "step": 26962
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0944872634460062e-05,
+      "loss": 0.3535,
+      "step": 26963
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0944536799153695e-05,
+      "loss": 0.4116,
+      "step": 26964
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0944200963847331e-05,
+      "loss": 0.2872,
+      "step": 26965
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0943865128540964e-05,
+      "loss": 0.2586,
+      "step": 26966
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0943529293234597e-05,
+      "loss": 0.3618,
+      "step": 26967
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0943193457928234e-05,
+      "loss": 0.411,
+      "step": 26968
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0942857622621867e-05,
+      "loss": 0.3428,
+      "step": 26969
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0942521787315501e-05,
+      "loss": 0.2863,
+      "step": 26970
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0942185952009136e-05,
+      "loss": 0.2299,
+      "step": 26971
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.094185011670277e-05,
+      "loss": 0.1999,
+      "step": 26972
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0941514281396404e-05,
+      "loss": 0.2328,
+      "step": 26973
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.094117844609004e-05,
+      "loss": 0.3911,
+      "step": 26974
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0940842610783673e-05,
+      "loss": 0.3428,
+      "step": 26975
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0940506775477306e-05,
+      "loss": 0.3916,
+      "step": 26976
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0940170940170942e-05,
+      "loss": 0.2905,
+      "step": 26977
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0939835104864575e-05,
+      "loss": 0.3316,
+      "step": 26978
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.093949926955821e-05,
+      "loss": 0.2249,
+      "step": 26979
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0939163434251843e-05,
+      "loss": 0.3578,
+      "step": 26980
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0938827598945479e-05,
+      "loss": 0.445,
+      "step": 26981
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0938491763639112e-05,
+      "loss": 0.2405,
+      "step": 26982
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0938155928332745e-05,
+      "loss": 0.2057,
+      "step": 26983
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0937820093026381e-05,
+      "loss": 0.2481,
+      "step": 26984
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0937484257720014e-05,
+      "loss": 0.2854,
+      "step": 26985
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0937148422413649e-05,
+      "loss": 0.3279,
+      "step": 26986
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0936812587107284e-05,
+      "loss": 0.4296,
+      "step": 26987
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0936476751800918e-05,
+      "loss": 0.1896,
+      "step": 26988
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0936140916494551e-05,
+      "loss": 0.3174,
+      "step": 26989
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0935805081188186e-05,
+      "loss": 0.3293,
+      "step": 26990
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.093546924588182e-05,
+      "loss": 0.4011,
+      "step": 26991
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0935133410575453e-05,
+      "loss": 0.1698,
+      "step": 26992
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.093479757526909e-05,
+      "loss": 0.299,
+      "step": 26993
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0934461739962723e-05,
+      "loss": 0.1629,
+      "step": 26994
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0934125904656356e-05,
+      "loss": 0.398,
+      "step": 26995
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0933790069349992e-05,
+      "loss": 0.128,
+      "step": 26996
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0933454234043625e-05,
+      "loss": 0.1762,
+      "step": 26997
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.093311839873726e-05,
+      "loss": 0.3575,
+      "step": 26998
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0932782563430894e-05,
+      "loss": 0.497,
+      "step": 26999
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0932446728124529e-05,
+      "loss": 0.3116,
+      "step": 27000
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0932110892818162e-05,
+      "loss": 0.2056,
+      "step": 27001
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0931775057511798e-05,
+      "loss": 0.346,
+      "step": 27002
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0931439222205431e-05,
+      "loss": 0.2432,
+      "step": 27003
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0931103386899064e-05,
+      "loss": 0.2876,
+      "step": 27004
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.09307675515927e-05,
+      "loss": 0.3917,
+      "step": 27005
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0930431716286333e-05,
+      "loss": 0.4437,
+      "step": 27006
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0930095880979968e-05,
+      "loss": 0.1609,
+      "step": 27007
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0929760045673603e-05,
+      "loss": 0.3034,
+      "step": 27008
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0929424210367237e-05,
+      "loss": 0.3523,
+      "step": 27009
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.092908837506087e-05,
+      "loss": 0.2584,
+      "step": 27010
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0928752539754505e-05,
+      "loss": 0.3331,
+      "step": 27011
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.092841670444814e-05,
+      "loss": 0.1766,
+      "step": 27012
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0928080869141773e-05,
+      "loss": 0.2666,
+      "step": 27013
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0927745033835409e-05,
+      "loss": 0.1216,
+      "step": 27014
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0927409198529042e-05,
+      "loss": 0.2614,
+      "step": 27015
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0927073363222675e-05,
+      "loss": 0.1842,
+      "step": 27016
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0926737527916311e-05,
+      "loss": 0.2726,
+      "step": 27017
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0926401692609944e-05,
+      "loss": 0.235,
+      "step": 27018
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0926065857303579e-05,
+      "loss": 0.4589,
+      "step": 27019
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0925730021997213e-05,
+      "loss": 0.2204,
+      "step": 27020
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0925394186690848e-05,
+      "loss": 0.2637,
+      "step": 27021
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0925058351384481e-05,
+      "loss": 0.3461,
+      "step": 27022
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0924722516078117e-05,
+      "loss": 0.192,
+      "step": 27023
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.092438668077175e-05,
+      "loss": 0.1567,
+      "step": 27024
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0924050845465383e-05,
+      "loss": 0.3271,
+      "step": 27025
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.092371501015902e-05,
+      "loss": 0.2472,
+      "step": 27026
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0923379174852653e-05,
+      "loss": 0.3567,
+      "step": 27027
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0923043339546287e-05,
+      "loss": 0.2942,
+      "step": 27028
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0922707504239922e-05,
+      "loss": 0.2871,
+      "step": 27029
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0922371668933556e-05,
+      "loss": 0.262,
+      "step": 27030
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.092203583362719e-05,
+      "loss": 0.237,
+      "step": 27031
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0921699998320826e-05,
+      "loss": 0.1571,
+      "step": 27032
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0921364163014459e-05,
+      "loss": 0.2908,
+      "step": 27033
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0921028327708092e-05,
+      "loss": 0.3021,
+      "step": 27034
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0920692492401728e-05,
+      "loss": 0.1929,
+      "step": 27035
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0920356657095361e-05,
+      "loss": 0.3793,
+      "step": 27036
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0920020821788994e-05,
+      "loss": 0.2218,
+      "step": 27037
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.091968498648263e-05,
+      "loss": 0.4274,
+      "step": 27038
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0919349151176263e-05,
+      "loss": 0.3946,
+      "step": 27039
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0919013315869898e-05,
+      "loss": 0.2753,
+      "step": 27040
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0918677480563533e-05,
+      "loss": 0.2208,
+      "step": 27041
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0918341645257167e-05,
+      "loss": 0.2016,
+      "step": 27042
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.09180058099508e-05,
+      "loss": 0.2317,
+      "step": 27043
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0917669974644436e-05,
+      "loss": 0.218,
+      "step": 27044
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.091733413933807e-05,
+      "loss": 0.2866,
+      "step": 27045
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0916998304031702e-05,
+      "loss": 0.3197,
+      "step": 27046
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0916662468725339e-05,
+      "loss": 0.2972,
+      "step": 27047
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0916326633418972e-05,
+      "loss": 0.6288,
+      "step": 27048
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0915990798112606e-05,
+      "loss": 0.3054,
+      "step": 27049
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0915654962806241e-05,
+      "loss": 0.2313,
+      "step": 27050
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0915319127499876e-05,
+      "loss": 0.3146,
+      "step": 27051
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0914983292193509e-05,
+      "loss": 0.2525,
+      "step": 27052
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0914647456887145e-05,
+      "loss": 0.3545,
+      "step": 27053
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0914311621580778e-05,
+      "loss": 0.4516,
+      "step": 27054
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.091397578627441e-05,
+      "loss": 0.377,
+      "step": 27055
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0913639950968047e-05,
+      "loss": 0.3984,
+      "step": 27056
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.091330411566168e-05,
+      "loss": 0.3018,
+      "step": 27057
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0912968280355313e-05,
+      "loss": 0.3668,
+      "step": 27058
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.091263244504895e-05,
+      "loss": 0.2515,
+      "step": 27059
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0912296609742582e-05,
+      "loss": 0.2035,
+      "step": 27060
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0911960774436217e-05,
+      "loss": 0.3268,
+      "step": 27061
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0911624939129852e-05,
+      "loss": 0.2323,
+      "step": 27062
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0911289103823486e-05,
+      "loss": 0.2116,
+      "step": 27063
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.091095326851712e-05,
+      "loss": 0.3399,
+      "step": 27064
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0910617433210756e-05,
+      "loss": 0.3766,
+      "step": 27065
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0910281597904389e-05,
+      "loss": 0.1618,
+      "step": 27066
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0909945762598021e-05,
+      "loss": 0.2911,
+      "step": 27067
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0909609927291658e-05,
+      "loss": 0.3098,
+      "step": 27068
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.090927409198529e-05,
+      "loss": 0.2933,
+      "step": 27069
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0908938256678925e-05,
+      "loss": 0.2995,
+      "step": 27070
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.090860242137256e-05,
+      "loss": 0.3157,
+      "step": 27071
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0908266586066195e-05,
+      "loss": 0.2181,
+      "step": 27072
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0907930750759828e-05,
+      "loss": 0.2192,
+      "step": 27073
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0907594915453464e-05,
+      "loss": 0.218,
+      "step": 27074
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0907259080147097e-05,
+      "loss": 0.36,
+      "step": 27075
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.090692324484073e-05,
+      "loss": 0.3737,
+      "step": 27076
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0906587409534366e-05,
+      "loss": 0.3578,
+      "step": 27077
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0906251574228e-05,
+      "loss": 0.3249,
+      "step": 27078
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0905915738921634e-05,
+      "loss": 0.2582,
+      "step": 27079
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0905579903615269e-05,
+      "loss": 0.367,
+      "step": 27080
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0905244068308901e-05,
+      "loss": 0.4736,
+      "step": 27081
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0904908233002536e-05,
+      "loss": 0.2713,
+      "step": 27082
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.090457239769617e-05,
+      "loss": 0.276,
+      "step": 27083
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0904236562389805e-05,
+      "loss": 0.4558,
+      "step": 27084
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0903900727083438e-05,
+      "loss": 0.43,
+      "step": 27085
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0903564891777075e-05,
+      "loss": 0.1503,
+      "step": 27086
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0903229056470708e-05,
+      "loss": 0.3803,
+      "step": 27087
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.090289322116434e-05,
+      "loss": 0.2762,
+      "step": 27088
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0902557385857977e-05,
+      "loss": 0.2781,
+      "step": 27089
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.090222155055161e-05,
+      "loss": 0.3391,
+      "step": 27090
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0901885715245245e-05,
+      "loss": 0.2295,
+      "step": 27091
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.090154987993888e-05,
+      "loss": 0.3562,
+      "step": 27092
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0901214044632514e-05,
+      "loss": 0.2509,
+      "step": 27093
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0900878209326147e-05,
+      "loss": 0.256,
+      "step": 27094
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0900542374019783e-05,
+      "loss": 0.291,
+      "step": 27095
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.0900206538713416e-05,
+      "loss": 0.4497,
+      "step": 27096
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0899870703407049e-05,
+      "loss": 0.2322,
+      "step": 27097
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0899534868100685e-05,
+      "loss": 0.4165,
+      "step": 27098
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0899199032794318e-05,
+      "loss": 0.3858,
+      "step": 27099
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0898863197487953e-05,
+      "loss": 0.4155,
+      "step": 27100
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0898527362181588e-05,
+      "loss": 0.2535,
+      "step": 27101
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.089819152687522e-05,
+      "loss": 0.3455,
+      "step": 27102
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0897855691568855e-05,
+      "loss": 0.2799,
+      "step": 27103
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.089751985626249e-05,
+      "loss": 0.4605,
+      "step": 27104
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0897184020956125e-05,
+      "loss": 0.2774,
+      "step": 27105
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0896848185649757e-05,
+      "loss": 0.3862,
+      "step": 27106
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0896512350343394e-05,
+      "loss": 0.4626,
+      "step": 27107
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0896176515037027e-05,
+      "loss": 0.3878,
+      "step": 27108
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.089584067973066e-05,
+      "loss": 0.1992,
+      "step": 27109
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0895504844424296e-05,
+      "loss": 0.342,
+      "step": 27110
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0895169009117929e-05,
+      "loss": 0.2491,
+      "step": 27111
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0894833173811564e-05,
+      "loss": 0.297,
+      "step": 27112
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0894497338505198e-05,
+      "loss": 0.271,
+      "step": 27113
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0894161503198833e-05,
+      "loss": 0.3475,
+      "step": 27114
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0893825667892466e-05,
+      "loss": 0.1479,
+      "step": 27115
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0893489832586102e-05,
+      "loss": 0.2386,
+      "step": 27116
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0893153997279735e-05,
+      "loss": 0.4376,
+      "step": 27117
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0892818161973368e-05,
+      "loss": 0.218,
+      "step": 27118
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0892482326667005e-05,
+      "loss": 0.0708,
+      "step": 27119
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0892146491360637e-05,
+      "loss": 0.3831,
+      "step": 27120
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0891810656054272e-05,
+      "loss": 0.3764,
+      "step": 27121
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0891474820747907e-05,
+      "loss": 0.2544,
+      "step": 27122
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0891138985441541e-05,
+      "loss": 0.2507,
+      "step": 27123
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0890803150135174e-05,
+      "loss": 0.3122,
+      "step": 27124
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0890467314828809e-05,
+      "loss": 0.1316,
+      "step": 27125
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0890131479522444e-05,
+      "loss": 0.1821,
+      "step": 27126
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0889795644216077e-05,
+      "loss": 0.1282,
+      "step": 27127
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0889459808909713e-05,
+      "loss": 0.1631,
+      "step": 27128
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0889123973603346e-05,
+      "loss": 0.332,
+      "step": 27129
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0888788138296979e-05,
+      "loss": 0.3678,
+      "step": 27130
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0888452302990615e-05,
+      "loss": 0.2055,
+      "step": 27131
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0888116467684248e-05,
+      "loss": 0.276,
+      "step": 27132
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0887780632377883e-05,
+      "loss": 0.3648,
+      "step": 27133
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0887444797071517e-05,
+      "loss": 0.2097,
+      "step": 27134
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0887108961765152e-05,
+      "loss": 0.3354,
+      "step": 27135
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0886773126458785e-05,
+      "loss": 0.3137,
+      "step": 27136
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0886437291152421e-05,
+      "loss": 0.264,
+      "step": 27137
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0886101455846054e-05,
+      "loss": 0.2099,
+      "step": 27138
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0885765620539687e-05,
+      "loss": 0.2978,
+      "step": 27139
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0885429785233324e-05,
+      "loss": 0.3137,
+      "step": 27140
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0885093949926957e-05,
+      "loss": 0.2948,
+      "step": 27141
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0884758114620591e-05,
+      "loss": 0.362,
+      "step": 27142
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0884422279314226e-05,
+      "loss": 0.357,
+      "step": 27143
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.088408644400786e-05,
+      "loss": 0.4769,
+      "step": 27144
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0883750608701493e-05,
+      "loss": 0.4798,
+      "step": 27145
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0883414773395128e-05,
+      "loss": 0.3725,
+      "step": 27146
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0883078938088763e-05,
+      "loss": 0.437,
+      "step": 27147
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0882743102782396e-05,
+      "loss": 0.2572,
+      "step": 27148
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0882407267476032e-05,
+      "loss": 0.2457,
+      "step": 27149
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0882071432169665e-05,
+      "loss": 0.329,
+      "step": 27150
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0881735596863298e-05,
+      "loss": 0.2982,
+      "step": 27151
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0881399761556934e-05,
+      "loss": 0.3148,
+      "step": 27152
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0881063926250567e-05,
+      "loss": 0.2223,
+      "step": 27153
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0880728090944202e-05,
+      "loss": 0.202,
+      "step": 27154
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0880392255637837e-05,
+      "loss": 0.3056,
+      "step": 27155
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0880056420331471e-05,
+      "loss": 0.1763,
+      "step": 27156
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0879720585025104e-05,
+      "loss": 0.4692,
+      "step": 27157
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.087938474971874e-05,
+      "loss": 0.3618,
+      "step": 27158
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0879048914412373e-05,
+      "loss": 0.3557,
+      "step": 27159
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0878713079106006e-05,
+      "loss": 0.2288,
+      "step": 27160
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0878377243799641e-05,
+      "loss": 0.2826,
+      "step": 27161
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0878041408493276e-05,
+      "loss": 0.3348,
+      "step": 27162
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.087770557318691e-05,
+      "loss": 0.1303,
+      "step": 27163
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0877369737880543e-05,
+      "loss": 0.5229,
+      "step": 27164
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.087703390257418e-05,
+      "loss": 0.3305,
+      "step": 27165
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0876698067267813e-05,
+      "loss": 0.3956,
+      "step": 27166
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0876362231961446e-05,
+      "loss": 0.2395,
+      "step": 27167
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0876026396655082e-05,
+      "loss": 0.2989,
+      "step": 27168
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0875690561348715e-05,
+      "loss": 0.2692,
+      "step": 27169
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0875354726042348e-05,
+      "loss": 0.1446,
+      "step": 27170
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0875018890735984e-05,
+      "loss": 0.199,
+      "step": 27171
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0874683055429617e-05,
+      "loss": 0.2916,
+      "step": 27172
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0874347220123252e-05,
+      "loss": 0.404,
+      "step": 27173
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0874011384816886e-05,
+      "loss": 0.1899,
+      "step": 27174
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0873675549510521e-05,
+      "loss": 0.2545,
+      "step": 27175
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0873339714204154e-05,
+      "loss": 0.4912,
+      "step": 27176
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.087300387889779e-05,
+      "loss": 0.2527,
+      "step": 27177
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0872668043591423e-05,
+      "loss": 0.4075,
+      "step": 27178
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0872332208285056e-05,
+      "loss": 0.1849,
+      "step": 27179
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0871996372978693e-05,
+      "loss": 0.3328,
+      "step": 27180
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0871660537672326e-05,
+      "loss": 0.236,
+      "step": 27181
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.087132470236596e-05,
+      "loss": 0.4145,
+      "step": 27182
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0870988867059595e-05,
+      "loss": 0.3673,
+      "step": 27183
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.087065303175323e-05,
+      "loss": 0.3072,
+      "step": 27184
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0870317196446862e-05,
+      "loss": 0.1985,
+      "step": 27185
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0869981361140499e-05,
+      "loss": 0.3007,
+      "step": 27186
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0869645525834132e-05,
+      "loss": 0.2999,
+      "step": 27187
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0869309690527765e-05,
+      "loss": 0.2375,
+      "step": 27188
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0868973855221401e-05,
+      "loss": 0.3409,
+      "step": 27189
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0868638019915034e-05,
+      "loss": 0.2583,
+      "step": 27190
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0868302184608669e-05,
+      "loss": 0.2581,
+      "step": 27191
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0867966349302303e-05,
+      "loss": 0.2803,
+      "step": 27192
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0867630513995936e-05,
+      "loss": 0.1614,
+      "step": 27193
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0867294678689571e-05,
+      "loss": 0.3734,
+      "step": 27194
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0866958843383206e-05,
+      "loss": 0.3754,
+      "step": 27195
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.086662300807684e-05,
+      "loss": 0.1822,
+      "step": 27196
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0866287172770473e-05,
+      "loss": 0.2872,
+      "step": 27197
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.086595133746411e-05,
+      "loss": 0.179,
+      "step": 27198
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0865615502157742e-05,
+      "loss": 0.1919,
+      "step": 27199
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0865279666851375e-05,
+      "loss": 0.2682,
+      "step": 27200
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0864943831545012e-05,
+      "loss": 0.2622,
+      "step": 27201
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0864607996238645e-05,
+      "loss": 0.3695,
+      "step": 27202
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.086427216093228e-05,
+      "loss": 0.2162,
+      "step": 27203
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0863936325625914e-05,
+      "loss": 0.4074,
+      "step": 27204
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0863600490319549e-05,
+      "loss": 0.3196,
+      "step": 27205
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0863264655013182e-05,
+      "loss": 0.3398,
+      "step": 27206
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0862928819706818e-05,
+      "loss": 0.1695,
+      "step": 27207
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0862592984400451e-05,
+      "loss": 0.4088,
+      "step": 27208
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0862257149094084e-05,
+      "loss": 0.5761,
+      "step": 27209
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.086192131378772e-05,
+      "loss": 0.3332,
+      "step": 27210
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0861585478481353e-05,
+      "loss": 0.2558,
+      "step": 27211
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0861249643174988e-05,
+      "loss": 0.1985,
+      "step": 27212
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0860913807868622e-05,
+      "loss": 0.2491,
+      "step": 27213
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0860577972562257e-05,
+      "loss": 0.2773,
+      "step": 27214
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.086024213725589e-05,
+      "loss": 0.1494,
+      "step": 27215
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0859906301949525e-05,
+      "loss": 0.2372,
+      "step": 27216
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.085957046664316e-05,
+      "loss": 0.2156,
+      "step": 27217
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0859234631336792e-05,
+      "loss": 0.2434,
+      "step": 27218
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0858898796030429e-05,
+      "loss": 0.3521,
+      "step": 27219
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0858562960724062e-05,
+      "loss": 0.3694,
+      "step": 27220
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0858227125417694e-05,
+      "loss": 0.2957,
+      "step": 27221
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.085789129011133e-05,
+      "loss": 0.1798,
+      "step": 27222
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0857555454804964e-05,
+      "loss": 0.2344,
+      "step": 27223
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0857219619498598e-05,
+      "loss": 0.3387,
+      "step": 27224
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0856883784192233e-05,
+      "loss": 0.2043,
+      "step": 27225
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0856547948885868e-05,
+      "loss": 0.1558,
+      "step": 27226
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.08562121135795e-05,
+      "loss": 0.3747,
+      "step": 27227
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0855876278273137e-05,
+      "loss": 0.2412,
+      "step": 27228
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.085554044296677e-05,
+      "loss": 0.3958,
+      "step": 27229
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0855204607660403e-05,
+      "loss": 0.39,
+      "step": 27230
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.085486877235404e-05,
+      "loss": 0.191,
+      "step": 27231
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0854532937047672e-05,
+      "loss": 0.1824,
+      "step": 27232
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0854197101741307e-05,
+      "loss": 0.5873,
+      "step": 27233
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0853861266434942e-05,
+      "loss": 0.2613,
+      "step": 27234
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0853525431128576e-05,
+      "loss": 0.3024,
+      "step": 27235
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0853189595822209e-05,
+      "loss": 0.2916,
+      "step": 27236
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0852853760515844e-05,
+      "loss": 0.2492,
+      "step": 27237
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0852517925209478e-05,
+      "loss": 0.2388,
+      "step": 27238
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0852182089903111e-05,
+      "loss": 0.4255,
+      "step": 27239
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0851846254596748e-05,
+      "loss": 0.1905,
+      "step": 27240
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.085151041929038e-05,
+      "loss": 0.2121,
+      "step": 27241
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0851174583984014e-05,
+      "loss": 0.2633,
+      "step": 27242
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.085083874867765e-05,
+      "loss": 0.2548,
+      "step": 27243
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0850502913371283e-05,
+      "loss": 0.3875,
+      "step": 27244
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0850167078064918e-05,
+      "loss": 0.3632,
+      "step": 27245
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0849831242758552e-05,
+      "loss": 0.2457,
+      "step": 27246
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0849495407452187e-05,
+      "loss": 0.2895,
+      "step": 27247
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.084915957214582e-05,
+      "loss": 0.2083,
+      "step": 27248
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0848823736839456e-05,
+      "loss": 0.2153,
+      "step": 27249
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0848487901533089e-05,
+      "loss": 0.3469,
+      "step": 27250
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0848152066226722e-05,
+      "loss": 0.3052,
+      "step": 27251
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0847816230920358e-05,
+      "loss": 0.2749,
+      "step": 27252
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0847480395613991e-05,
+      "loss": 0.3361,
+      "step": 27253
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0847144560307626e-05,
+      "loss": 0.2455,
+      "step": 27254
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.084680872500126e-05,
+      "loss": 0.3813,
+      "step": 27255
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0846472889694895e-05,
+      "loss": 0.132,
+      "step": 27256
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0846137054388528e-05,
+      "loss": 0.3232,
+      "step": 27257
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0845801219082165e-05,
+      "loss": 0.3996,
+      "step": 27258
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0845465383775798e-05,
+      "loss": 0.2497,
+      "step": 27259
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.084512954846943e-05,
+      "loss": 0.3472,
+      "step": 27260
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0844793713163067e-05,
+      "loss": 0.3462,
+      "step": 27261
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.08444578778567e-05,
+      "loss": 0.3564,
+      "step": 27262
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0844122042550333e-05,
+      "loss": 0.4415,
+      "step": 27263
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0843786207243969e-05,
+      "loss": 0.2079,
+      "step": 27264
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0843450371937602e-05,
+      "loss": 0.1925,
+      "step": 27265
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0843114536631237e-05,
+      "loss": 0.2255,
+      "step": 27266
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0842778701324871e-05,
+      "loss": 0.2137,
+      "step": 27267
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0842442866018506e-05,
+      "loss": 0.2333,
+      "step": 27268
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0842107030712139e-05,
+      "loss": 0.4054,
+      "step": 27269
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0841771195405775e-05,
+      "loss": 0.3033,
+      "step": 27270
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0841435360099408e-05,
+      "loss": 0.2443,
+      "step": 27271
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0841099524793041e-05,
+      "loss": 0.2268,
+      "step": 27272
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0840763689486678e-05,
+      "loss": 0.0806,
+      "step": 27273
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.084042785418031e-05,
+      "loss": 0.3106,
+      "step": 27274
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0840092018873945e-05,
+      "loss": 0.1591,
+      "step": 27275
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.083975618356758e-05,
+      "loss": 0.3491,
+      "step": 27276
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0839420348261214e-05,
+      "loss": 0.287,
+      "step": 27277
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0839084512954847e-05,
+      "loss": 0.4017,
+      "step": 27278
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0838748677648484e-05,
+      "loss": 0.2896,
+      "step": 27279
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0838412842342117e-05,
+      "loss": 0.1985,
+      "step": 27280
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.083807700703575e-05,
+      "loss": 0.2803,
+      "step": 27281
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0837741171729386e-05,
+      "loss": 0.1297,
+      "step": 27282
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0837405336423019e-05,
+      "loss": 0.2312,
+      "step": 27283
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0837069501116652e-05,
+      "loss": 0.4079,
+      "step": 27284
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0836733665810288e-05,
+      "loss": 0.2213,
+      "step": 27285
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0836397830503921e-05,
+      "loss": 0.4233,
+      "step": 27286
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0836061995197556e-05,
+      "loss": 0.2971,
+      "step": 27287
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.083572615989119e-05,
+      "loss": 0.2552,
+      "step": 27288
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0835390324584825e-05,
+      "loss": 0.342,
+      "step": 27289
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0835054489278458e-05,
+      "loss": 0.2327,
+      "step": 27290
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0834718653972094e-05,
+      "loss": 0.3906,
+      "step": 27291
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0834382818665727e-05,
+      "loss": 0.1286,
+      "step": 27292
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.083404698335936e-05,
+      "loss": 0.1927,
+      "step": 27293
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.0833711148052997e-05,
+      "loss": 0.3209,
+      "step": 27294
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 1.083337531274663e-05,
+      "loss": 0.1543,
+      "step": 27295
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0833039477440264e-05,
+      "loss": 0.3474,
+      "step": 27296
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0832703642133899e-05,
+      "loss": 0.2758,
+      "step": 27297
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0832367806827534e-05,
+      "loss": 0.4181,
+      "step": 27298
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0832031971521167e-05,
+      "loss": 0.1701,
+      "step": 27299
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0831696136214803e-05,
+      "loss": 0.2892,
+      "step": 27300
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0831360300908436e-05,
+      "loss": 0.2131,
+      "step": 27301
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0831024465602069e-05,
+      "loss": 0.2316,
+      "step": 27302
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0830688630295705e-05,
+      "loss": 0.3662,
+      "step": 27303
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0830352794989338e-05,
+      "loss": 0.3149,
+      "step": 27304
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0830016959682971e-05,
+      "loss": 0.2662,
+      "step": 27305
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0829681124376607e-05,
+      "loss": 0.34,
+      "step": 27306
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.082934528907024e-05,
+      "loss": 0.4376,
+      "step": 27307
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0829009453763875e-05,
+      "loss": 0.3572,
+      "step": 27308
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.082867361845751e-05,
+      "loss": 0.4852,
+      "step": 27309
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0828337783151144e-05,
+      "loss": 0.3359,
+      "step": 27310
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0828001947844777e-05,
+      "loss": 0.2895,
+      "step": 27311
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0827666112538414e-05,
+      "loss": 0.3366,
+      "step": 27312
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0827330277232046e-05,
+      "loss": 0.2963,
+      "step": 27313
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.082699444192568e-05,
+      "loss": 0.1891,
+      "step": 27314
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0826658606619316e-05,
+      "loss": 0.2358,
+      "step": 27315
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0826322771312949e-05,
+      "loss": 0.4394,
+      "step": 27316
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0825986936006583e-05,
+      "loss": 0.307,
+      "step": 27317
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0825651100700218e-05,
+      "loss": 0.3197,
+      "step": 27318
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0825315265393853e-05,
+      "loss": 0.4345,
+      "step": 27319
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0824979430087486e-05,
+      "loss": 0.2901,
+      "step": 27320
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0824643594781122e-05,
+      "loss": 0.372,
+      "step": 27321
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0824307759474755e-05,
+      "loss": 0.3341,
+      "step": 27322
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0823971924168388e-05,
+      "loss": 0.4277,
+      "step": 27323
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0823636088862024e-05,
+      "loss": 0.5357,
+      "step": 27324
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0823300253555657e-05,
+      "loss": 0.3229,
+      "step": 27325
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0822964418249292e-05,
+      "loss": 0.4274,
+      "step": 27326
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0822628582942926e-05,
+      "loss": 0.4967,
+      "step": 27327
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.082229274763656e-05,
+      "loss": 0.2098,
+      "step": 27328
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0821956912330194e-05,
+      "loss": 0.1781,
+      "step": 27329
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0821621077023829e-05,
+      "loss": 0.405,
+      "step": 27330
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0821285241717463e-05,
+      "loss": 0.3708,
+      "step": 27331
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0820949406411096e-05,
+      "loss": 0.2337,
+      "step": 27332
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0820613571104733e-05,
+      "loss": 0.3606,
+      "step": 27333
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0820277735798366e-05,
+      "loss": 0.2595,
+      "step": 27334
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0819941900491999e-05,
+      "loss": 0.3266,
+      "step": 27335
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0819606065185635e-05,
+      "loss": 0.4397,
+      "step": 27336
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0819270229879268e-05,
+      "loss": 0.3194,
+      "step": 27337
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0818934394572903e-05,
+      "loss": 0.3283,
+      "step": 27338
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0818598559266537e-05,
+      "loss": 0.395,
+      "step": 27339
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0818262723960172e-05,
+      "loss": 0.2252,
+      "step": 27340
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0817926888653805e-05,
+      "loss": 0.2981,
+      "step": 27341
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0817591053347438e-05,
+      "loss": 0.2612,
+      "step": 27342
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0817255218041074e-05,
+      "loss": 0.2291,
+      "step": 27343
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0816919382734707e-05,
+      "loss": 0.3233,
+      "step": 27344
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0816583547428342e-05,
+      "loss": 0.1243,
+      "step": 27345
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0816247712121976e-05,
+      "loss": 0.1789,
+      "step": 27346
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0815911876815611e-05,
+      "loss": 0.4109,
+      "step": 27347
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0815576041509244e-05,
+      "loss": 0.2435,
+      "step": 27348
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.081524020620288e-05,
+      "loss": 0.1942,
+      "step": 27349
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0814904370896513e-05,
+      "loss": 0.4803,
+      "step": 27350
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0814568535590146e-05,
+      "loss": 0.206,
+      "step": 27351
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0814232700283782e-05,
+      "loss": 0.3735,
+      "step": 27352
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0813896864977415e-05,
+      "loss": 0.4648,
+      "step": 27353
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0813561029671048e-05,
+      "loss": 0.3234,
+      "step": 27354
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0813225194364685e-05,
+      "loss": 0.3078,
+      "step": 27355
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0812889359058318e-05,
+      "loss": 0.2653,
+      "step": 27356
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0812553523751952e-05,
+      "loss": 0.4914,
+      "step": 27357
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0812217688445587e-05,
+      "loss": 0.3654,
+      "step": 27358
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0811881853139222e-05,
+      "loss": 0.2858,
+      "step": 27359
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0811546017832855e-05,
+      "loss": 0.2988,
+      "step": 27360
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0811210182526491e-05,
+      "loss": 0.4197,
+      "step": 27361
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0810874347220124e-05,
+      "loss": 0.3051,
+      "step": 27362
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0810538511913757e-05,
+      "loss": 0.3721,
+      "step": 27363
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0810202676607393e-05,
+      "loss": 0.3208,
+      "step": 27364
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0809866841301026e-05,
+      "loss": 0.1973,
+      "step": 27365
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.080953100599466e-05,
+      "loss": 0.3562,
+      "step": 27366
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0809195170688295e-05,
+      "loss": 0.4102,
+      "step": 27367
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.080885933538193e-05,
+      "loss": 0.3039,
+      "step": 27368
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0808523500075563e-05,
+      "loss": 0.3215,
+      "step": 27369
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.08081876647692e-05,
+      "loss": 0.195,
+      "step": 27370
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0807851829462832e-05,
+      "loss": 0.2521,
+      "step": 27371
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0807515994156465e-05,
+      "loss": 0.3449,
+      "step": 27372
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0807180158850102e-05,
+      "loss": 0.2928,
+      "step": 27373
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0806844323543735e-05,
+      "loss": 0.4235,
+      "step": 27374
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0806508488237368e-05,
+      "loss": 0.2213,
+      "step": 27375
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0806172652931004e-05,
+      "loss": 0.2478,
+      "step": 27376
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0805836817624637e-05,
+      "loss": 0.2024,
+      "step": 27377
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0805500982318271e-05,
+      "loss": 0.209,
+      "step": 27378
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0805165147011906e-05,
+      "loss": 0.4368,
+      "step": 27379
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.080482931170554e-05,
+      "loss": 0.3401,
+      "step": 27380
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0804493476399174e-05,
+      "loss": 0.2752,
+      "step": 27381
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.080415764109281e-05,
+      "loss": 0.1838,
+      "step": 27382
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0803821805786443e-05,
+      "loss": 0.3583,
+      "step": 27383
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0803485970480076e-05,
+      "loss": 0.411,
+      "step": 27384
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0803150135173712e-05,
+      "loss": 0.1886,
+      "step": 27385
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0802814299867345e-05,
+      "loss": 0.3262,
+      "step": 27386
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.080247846456098e-05,
+      "loss": 0.1282,
+      "step": 27387
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0802142629254615e-05,
+      "loss": 0.3799,
+      "step": 27388
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.080180679394825e-05,
+      "loss": 0.3298,
+      "step": 27389
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0801470958641882e-05,
+      "loss": 0.296,
+      "step": 27390
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0801135123335518e-05,
+      "loss": 0.3057,
+      "step": 27391
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0800799288029151e-05,
+      "loss": 0.3536,
+      "step": 27392
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0800463452722784e-05,
+      "loss": 0.2272,
+      "step": 27393
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.080012761741642e-05,
+      "loss": 0.3325,
+      "step": 27394
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0799791782110054e-05,
+      "loss": 0.2549,
+      "step": 27395
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0799455946803687e-05,
+      "loss": 0.2084,
+      "step": 27396
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0799120111497323e-05,
+      "loss": 0.3021,
+      "step": 27397
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0798784276190956e-05,
+      "loss": 0.2451,
+      "step": 27398
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.079844844088459e-05,
+      "loss": 0.2167,
+      "step": 27399
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0798112605578225e-05,
+      "loss": 0.3735,
+      "step": 27400
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.079777677027186e-05,
+      "loss": 0.314,
+      "step": 27401
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0797440934965493e-05,
+      "loss": 0.2975,
+      "step": 27402
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.079710509965913e-05,
+      "loss": 0.4146,
+      "step": 27403
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0796769264352762e-05,
+      "loss": 0.2911,
+      "step": 27404
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0796433429046395e-05,
+      "loss": 0.2279,
+      "step": 27405
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0796097593740031e-05,
+      "loss": 0.3277,
+      "step": 27406
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0795761758433664e-05,
+      "loss": 0.3264,
+      "step": 27407
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0795425923127299e-05,
+      "loss": 0.2104,
+      "step": 27408
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0795090087820934e-05,
+      "loss": 0.3205,
+      "step": 27409
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0794754252514568e-05,
+      "loss": 0.244,
+      "step": 27410
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0794418417208201e-05,
+      "loss": 0.3573,
+      "step": 27411
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0794082581901838e-05,
+      "loss": 0.351,
+      "step": 27412
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.079374674659547e-05,
+      "loss": 0.3919,
+      "step": 27413
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0793410911289104e-05,
+      "loss": 0.2812,
+      "step": 27414
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.079307507598274e-05,
+      "loss": 0.3091,
+      "step": 27415
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0792739240676373e-05,
+      "loss": 0.2857,
+      "step": 27416
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0792403405370007e-05,
+      "loss": 0.3144,
+      "step": 27417
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0792067570063642e-05,
+      "loss": 0.284,
+      "step": 27418
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0791731734757275e-05,
+      "loss": 0.4289,
+      "step": 27419
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.079139589945091e-05,
+      "loss": 0.3351,
+      "step": 27420
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0791060064144544e-05,
+      "loss": 0.3102,
+      "step": 27421
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0790724228838179e-05,
+      "loss": 0.2027,
+      "step": 27422
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0790388393531812e-05,
+      "loss": 0.1134,
+      "step": 27423
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0790052558225448e-05,
+      "loss": 0.1631,
+      "step": 27424
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0789716722919081e-05,
+      "loss": 0.3597,
+      "step": 27425
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0789380887612714e-05,
+      "loss": 0.3953,
+      "step": 27426
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.078904505230635e-05,
+      "loss": 0.2535,
+      "step": 27427
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0788709216999983e-05,
+      "loss": 0.1626,
+      "step": 27428
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0788373381693618e-05,
+      "loss": 0.2021,
+      "step": 27429
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0788037546387253e-05,
+      "loss": 0.3134,
+      "step": 27430
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0787701711080887e-05,
+      "loss": 0.2978,
+      "step": 27431
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.078736587577452e-05,
+      "loss": 0.3504,
+      "step": 27432
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0787030040468157e-05,
+      "loss": 0.3606,
+      "step": 27433
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.078669420516179e-05,
+      "loss": 0.1683,
+      "step": 27434
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0786358369855423e-05,
+      "loss": 0.3998,
+      "step": 27435
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0786022534549059e-05,
+      "loss": 0.2134,
+      "step": 27436
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0785686699242692e-05,
+      "loss": 0.4447,
+      "step": 27437
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0785350863936327e-05,
+      "loss": 0.2546,
+      "step": 27438
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0785015028629961e-05,
+      "loss": 0.3102,
+      "step": 27439
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0784679193323594e-05,
+      "loss": 0.3541,
+      "step": 27440
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0784343358017229e-05,
+      "loss": 0.2392,
+      "step": 27441
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0784007522710863e-05,
+      "loss": 0.3055,
+      "step": 27442
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0783671687404498e-05,
+      "loss": 0.3105,
+      "step": 27443
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0783335852098131e-05,
+      "loss": 0.2567,
+      "step": 27444
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0783000016791767e-05,
+      "loss": 0.3113,
+      "step": 27445
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.07826641814854e-05,
+      "loss": 0.2944,
+      "step": 27446
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0782328346179033e-05,
+      "loss": 0.2642,
+      "step": 27447
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.078199251087267e-05,
+      "loss": 0.1496,
+      "step": 27448
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0781656675566303e-05,
+      "loss": 0.1903,
+      "step": 27449
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0781320840259937e-05,
+      "loss": 0.2493,
+      "step": 27450
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0780985004953572e-05,
+      "loss": 0.3496,
+      "step": 27451
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0780649169647207e-05,
+      "loss": 0.2806,
+      "step": 27452
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.078031333434084e-05,
+      "loss": 0.1878,
+      "step": 27453
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0779977499034476e-05,
+      "loss": 0.2821,
+      "step": 27454
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0779641663728109e-05,
+      "loss": 0.4255,
+      "step": 27455
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0779305828421742e-05,
+      "loss": 0.3156,
+      "step": 27456
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0778969993115378e-05,
+      "loss": 0.3255,
+      "step": 27457
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0778634157809011e-05,
+      "loss": 0.3532,
+      "step": 27458
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0778298322502646e-05,
+      "loss": 0.1877,
+      "step": 27459
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.077796248719628e-05,
+      "loss": 0.2318,
+      "step": 27460
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0777626651889915e-05,
+      "loss": 0.4617,
+      "step": 27461
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0777290816583548e-05,
+      "loss": 0.3576,
+      "step": 27462
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0776954981277183e-05,
+      "loss": 0.347,
+      "step": 27463
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0776619145970817e-05,
+      "loss": 0.373,
+      "step": 27464
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.077628331066445e-05,
+      "loss": 0.2347,
+      "step": 27465
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0775947475358087e-05,
+      "loss": 0.1153,
+      "step": 27466
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.077561164005172e-05,
+      "loss": 0.2688,
+      "step": 27467
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0775275804745352e-05,
+      "loss": 0.273,
+      "step": 27468
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0774939969438989e-05,
+      "loss": 0.5039,
+      "step": 27469
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0774604134132622e-05,
+      "loss": 0.3491,
+      "step": 27470
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0774268298826256e-05,
+      "loss": 0.2732,
+      "step": 27471
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0773932463519891e-05,
+      "loss": 0.3987,
+      "step": 27472
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0773596628213526e-05,
+      "loss": 0.3281,
+      "step": 27473
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0773260792907159e-05,
+      "loss": 0.2501,
+      "step": 27474
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0772924957600795e-05,
+      "loss": 0.4088,
+      "step": 27475
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0772589122294428e-05,
+      "loss": 0.3298,
+      "step": 27476
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0772253286988061e-05,
+      "loss": 0.2233,
+      "step": 27477
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0771917451681697e-05,
+      "loss": 0.3189,
+      "step": 27478
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.077158161637533e-05,
+      "loss": 0.3349,
+      "step": 27479
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0771245781068965e-05,
+      "loss": 0.2758,
+      "step": 27480
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.07709099457626e-05,
+      "loss": 0.3001,
+      "step": 27481
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0770574110456234e-05,
+      "loss": 0.4524,
+      "step": 27482
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0770238275149867e-05,
+      "loss": 0.3581,
+      "step": 27483
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0769902439843503e-05,
+      "loss": 0.4288,
+      "step": 27484
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0769566604537136e-05,
+      "loss": 0.2204,
+      "step": 27485
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.076923076923077e-05,
+      "loss": 0.3658,
+      "step": 27486
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0768894933924406e-05,
+      "loss": 0.3831,
+      "step": 27487
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0768559098618039e-05,
+      "loss": 0.2077,
+      "step": 27488
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0768223263311672e-05,
+      "loss": 0.3302,
+      "step": 27489
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0767887428005308e-05,
+      "loss": 0.3148,
+      "step": 27490
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0767551592698941e-05,
+      "loss": 0.3379,
+      "step": 27491
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.0767215757392576e-05,
+      "loss": 0.361,
+      "step": 27492
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 1.076687992208621e-05,
+      "loss": 0.3858,
+      "step": 27493
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0766544086779845e-05,
+      "loss": 0.3471,
+      "step": 27494
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0766208251473478e-05,
+      "loss": 0.3173,
+      "step": 27495
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0765872416167114e-05,
+      "loss": 0.3197,
+      "step": 27496
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0765536580860747e-05,
+      "loss": 0.2732,
+      "step": 27497
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.076520074555438e-05,
+      "loss": 0.3649,
+      "step": 27498
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0764864910248016e-05,
+      "loss": 0.2271,
+      "step": 27499
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.076452907494165e-05,
+      "loss": 0.2177,
+      "step": 27500
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0764193239635284e-05,
+      "loss": 0.3881,
+      "step": 27501
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0763857404328919e-05,
+      "loss": 0.2211,
+      "step": 27502
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0763521569022553e-05,
+      "loss": 0.2064,
+      "step": 27503
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0763185733716186e-05,
+      "loss": 0.2779,
+      "step": 27504
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0762849898409823e-05,
+      "loss": 0.2458,
+      "step": 27505
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0762514063103456e-05,
+      "loss": 0.4506,
+      "step": 27506
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0762178227797088e-05,
+      "loss": 0.2315,
+      "step": 27507
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0761842392490725e-05,
+      "loss": 0.2754,
+      "step": 27508
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0761506557184358e-05,
+      "loss": 0.2182,
+      "step": 27509
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.076117072187799e-05,
+      "loss": 0.2918,
+      "step": 27510
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0760834886571627e-05,
+      "loss": 0.3141,
+      "step": 27511
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.076049905126526e-05,
+      "loss": 0.2972,
+      "step": 27512
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0760163215958895e-05,
+      "loss": 0.394,
+      "step": 27513
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.075982738065253e-05,
+      "loss": 0.2758,
+      "step": 27514
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0759491545346164e-05,
+      "loss": 0.2514,
+      "step": 27515
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0759155710039797e-05,
+      "loss": 0.209,
+      "step": 27516
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0758819874733433e-05,
+      "loss": 0.2424,
+      "step": 27517
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0758484039427066e-05,
+      "loss": 0.1612,
+      "step": 27518
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0758148204120699e-05,
+      "loss": 0.1797,
+      "step": 27519
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0757812368814335e-05,
+      "loss": 0.2499,
+      "step": 27520
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0757476533507968e-05,
+      "loss": 0.3912,
+      "step": 27521
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0757140698201603e-05,
+      "loss": 0.3297,
+      "step": 27522
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0756804862895236e-05,
+      "loss": 0.2139,
+      "step": 27523
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0756469027588872e-05,
+      "loss": 0.1965,
+      "step": 27524
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0756133192282505e-05,
+      "loss": 0.3318,
+      "step": 27525
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0755797356976138e-05,
+      "loss": 0.3275,
+      "step": 27526
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0755461521669775e-05,
+      "loss": 0.4216,
+      "step": 27527
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0755125686363408e-05,
+      "loss": 0.3689,
+      "step": 27528
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0754789851057042e-05,
+      "loss": 0.1164,
+      "step": 27529
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0754454015750677e-05,
+      "loss": 0.222,
+      "step": 27530
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.075411818044431e-05,
+      "loss": 0.2246,
+      "step": 27531
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0753782345137944e-05,
+      "loss": 0.3521,
+      "step": 27532
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0753446509831579e-05,
+      "loss": 0.2633,
+      "step": 27533
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0753110674525214e-05,
+      "loss": 0.4146,
+      "step": 27534
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0752774839218847e-05,
+      "loss": 0.3643,
+      "step": 27535
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0752439003912483e-05,
+      "loss": 0.2712,
+      "step": 27536
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0752103168606116e-05,
+      "loss": 0.2825,
+      "step": 27537
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0751767333299749e-05,
+      "loss": 0.3006,
+      "step": 27538
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0751431497993385e-05,
+      "loss": 0.1464,
+      "step": 27539
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0751095662687018e-05,
+      "loss": 0.1566,
+      "step": 27540
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0750759827380653e-05,
+      "loss": 0.2395,
+      "step": 27541
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0750423992074288e-05,
+      "loss": 0.4726,
+      "step": 27542
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0750088156767922e-05,
+      "loss": 0.2272,
+      "step": 27543
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0749752321461555e-05,
+      "loss": 0.4126,
+      "step": 27544
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0749416486155192e-05,
+      "loss": 0.2116,
+      "step": 27545
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0749080650848824e-05,
+      "loss": 0.3491,
+      "step": 27546
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0748744815542457e-05,
+      "loss": 0.3229,
+      "step": 27547
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0748408980236094e-05,
+      "loss": 0.4738,
+      "step": 27548
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0748073144929727e-05,
+      "loss": 0.407,
+      "step": 27549
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0747737309623361e-05,
+      "loss": 0.4221,
+      "step": 27550
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0747401474316996e-05,
+      "loss": 0.2893,
+      "step": 27551
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.074706563901063e-05,
+      "loss": 0.2546,
+      "step": 27552
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0746729803704264e-05,
+      "loss": 0.2629,
+      "step": 27553
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0746393968397898e-05,
+      "loss": 0.306,
+      "step": 27554
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0746058133091533e-05,
+      "loss": 0.2553,
+      "step": 27555
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0745722297785166e-05,
+      "loss": 0.3271,
+      "step": 27556
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0745386462478802e-05,
+      "loss": 0.4512,
+      "step": 27557
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0745050627172435e-05,
+      "loss": 0.1971,
+      "step": 27558
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0744714791866068e-05,
+      "loss": 0.3236,
+      "step": 27559
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0744378956559704e-05,
+      "loss": 0.2691,
+      "step": 27560
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0744043121253337e-05,
+      "loss": 0.3836,
+      "step": 27561
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0743707285946972e-05,
+      "loss": 0.1447,
+      "step": 27562
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0743371450640607e-05,
+      "loss": 0.2705,
+      "step": 27563
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0743035615334241e-05,
+      "loss": 0.2382,
+      "step": 27564
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0742699780027874e-05,
+      "loss": 0.2152,
+      "step": 27565
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.074236394472151e-05,
+      "loss": 0.3324,
+      "step": 27566
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0742028109415144e-05,
+      "loss": 0.2603,
+      "step": 27567
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0741692274108777e-05,
+      "loss": 0.408,
+      "step": 27568
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0741356438802413e-05,
+      "loss": 0.363,
+      "step": 27569
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0741020603496046e-05,
+      "loss": 0.2363,
+      "step": 27570
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.074068476818968e-05,
+      "loss": 0.4188,
+      "step": 27571
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0740348932883315e-05,
+      "loss": 0.1187,
+      "step": 27572
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.074001309757695e-05,
+      "loss": 0.0864,
+      "step": 27573
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0739677262270583e-05,
+      "loss": 0.3722,
+      "step": 27574
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0739341426964219e-05,
+      "loss": 0.2406,
+      "step": 27575
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0739005591657852e-05,
+      "loss": 0.3817,
+      "step": 27576
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0738669756351485e-05,
+      "loss": 0.2974,
+      "step": 27577
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0738333921045121e-05,
+      "loss": 0.2521,
+      "step": 27578
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0737998085738754e-05,
+      "loss": 0.355,
+      "step": 27579
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0737662250432387e-05,
+      "loss": 0.3377,
+      "step": 27580
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0737326415126024e-05,
+      "loss": 0.3178,
+      "step": 27581
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0736990579819657e-05,
+      "loss": 0.241,
+      "step": 27582
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0736654744513291e-05,
+      "loss": 0.207,
+      "step": 27583
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0736318909206926e-05,
+      "loss": 0.3067,
+      "step": 27584
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.073598307390056e-05,
+      "loss": 0.3471,
+      "step": 27585
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0735647238594193e-05,
+      "loss": 0.3825,
+      "step": 27586
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.073531140328783e-05,
+      "loss": 0.4452,
+      "step": 27587
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0734975567981463e-05,
+      "loss": 0.332,
+      "step": 27588
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0734639732675096e-05,
+      "loss": 0.3056,
+      "step": 27589
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0734303897368732e-05,
+      "loss": 0.2416,
+      "step": 27590
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0733968062062365e-05,
+      "loss": 0.2938,
+      "step": 27591
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0733632226756e-05,
+      "loss": 0.2948,
+      "step": 27592
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0733296391449634e-05,
+      "loss": 0.2811,
+      "step": 27593
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0732960556143269e-05,
+      "loss": 0.4007,
+      "step": 27594
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0732624720836902e-05,
+      "loss": 0.3895,
+      "step": 27595
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0732288885530538e-05,
+      "loss": 0.1808,
+      "step": 27596
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0731953050224171e-05,
+      "loss": 0.4082,
+      "step": 27597
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0731617214917804e-05,
+      "loss": 0.285,
+      "step": 27598
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.073128137961144e-05,
+      "loss": 0.314,
+      "step": 27599
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0730945544305073e-05,
+      "loss": 0.3513,
+      "step": 27600
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0730609708998706e-05,
+      "loss": 0.4507,
+      "step": 27601
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0730273873692343e-05,
+      "loss": 0.2219,
+      "step": 27602
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0729938038385976e-05,
+      "loss": 0.2704,
+      "step": 27603
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.072960220307961e-05,
+      "loss": 0.3105,
+      "step": 27604
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0729266367773245e-05,
+      "loss": 0.3425,
+      "step": 27605
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.072893053246688e-05,
+      "loss": 0.3277,
+      "step": 27606
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0728594697160513e-05,
+      "loss": 0.2737,
+      "step": 27607
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0728258861854149e-05,
+      "loss": 0.3795,
+      "step": 27608
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0727923026547782e-05,
+      "loss": 0.2071,
+      "step": 27609
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0727587191241415e-05,
+      "loss": 0.2828,
+      "step": 27610
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0727251355935051e-05,
+      "loss": 0.3097,
+      "step": 27611
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0726915520628684e-05,
+      "loss": 0.2272,
+      "step": 27612
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0726579685322319e-05,
+      "loss": 0.2793,
+      "step": 27613
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0726243850015953e-05,
+      "loss": 0.3701,
+      "step": 27614
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0725908014709588e-05,
+      "loss": 0.221,
+      "step": 27615
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0725572179403221e-05,
+      "loss": 0.4501,
+      "step": 27616
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0725236344096857e-05,
+      "loss": 0.4036,
+      "step": 27617
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.072490050879049e-05,
+      "loss": 0.376,
+      "step": 27618
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0724564673484123e-05,
+      "loss": 0.3682,
+      "step": 27619
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.072422883817776e-05,
+      "loss": 0.2628,
+      "step": 27620
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0723893002871393e-05,
+      "loss": 0.3835,
+      "step": 27621
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0723557167565025e-05,
+      "loss": 0.3254,
+      "step": 27622
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0723221332258662e-05,
+      "loss": 0.193,
+      "step": 27623
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0722885496952295e-05,
+      "loss": 0.2727,
+      "step": 27624
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.072254966164593e-05,
+      "loss": 0.2069,
+      "step": 27625
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0722213826339564e-05,
+      "loss": 0.3567,
+      "step": 27626
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0721877991033199e-05,
+      "loss": 0.3327,
+      "step": 27627
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0721542155726832e-05,
+      "loss": 0.288,
+      "step": 27628
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0721206320420468e-05,
+      "loss": 0.4258,
+      "step": 27629
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0720870485114101e-05,
+      "loss": 0.2042,
+      "step": 27630
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0720534649807734e-05,
+      "loss": 0.3449,
+      "step": 27631
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.072019881450137e-05,
+      "loss": 0.2525,
+      "step": 27632
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0719862979195003e-05,
+      "loss": 0.2797,
+      "step": 27633
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0719527143888638e-05,
+      "loss": 0.3094,
+      "step": 27634
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0719191308582272e-05,
+      "loss": 0.3251,
+      "step": 27635
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0718855473275907e-05,
+      "loss": 0.3124,
+      "step": 27636
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.071851963796954e-05,
+      "loss": 0.2905,
+      "step": 27637
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0718183802663176e-05,
+      "loss": 0.3266,
+      "step": 27638
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.071784796735681e-05,
+      "loss": 0.3159,
+      "step": 27639
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0717512132050442e-05,
+      "loss": 0.5101,
+      "step": 27640
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0717176296744079e-05,
+      "loss": 0.338,
+      "step": 27641
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0716840461437712e-05,
+      "loss": 0.3792,
+      "step": 27642
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0716504626131346e-05,
+      "loss": 0.3571,
+      "step": 27643
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0716168790824981e-05,
+      "loss": 0.4486,
+      "step": 27644
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0715832955518614e-05,
+      "loss": 0.1911,
+      "step": 27645
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0715497120212249e-05,
+      "loss": 0.5512,
+      "step": 27646
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0715161284905883e-05,
+      "loss": 0.4504,
+      "step": 27647
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0714825449599518e-05,
+      "loss": 0.2032,
+      "step": 27648
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.071448961429315e-05,
+      "loss": 0.2729,
+      "step": 27649
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0714153778986787e-05,
+      "loss": 0.5098,
+      "step": 27650
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.071381794368042e-05,
+      "loss": 0.2912,
+      "step": 27651
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0713482108374053e-05,
+      "loss": 0.2798,
+      "step": 27652
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.071314627306769e-05,
+      "loss": 0.2148,
+      "step": 27653
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0712810437761322e-05,
+      "loss": 0.3562,
+      "step": 27654
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0712474602454957e-05,
+      "loss": 0.2681,
+      "step": 27655
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0712138767148592e-05,
+      "loss": 0.2137,
+      "step": 27656
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0711802931842226e-05,
+      "loss": 0.269,
+      "step": 27657
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.071146709653586e-05,
+      "loss": 0.4188,
+      "step": 27658
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0711131261229496e-05,
+      "loss": 0.1496,
+      "step": 27659
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0710795425923129e-05,
+      "loss": 0.2706,
+      "step": 27660
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0710459590616761e-05,
+      "loss": 0.1778,
+      "step": 27661
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0710123755310398e-05,
+      "loss": 0.2461,
+      "step": 27662
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.070978792000403e-05,
+      "loss": 0.2279,
+      "step": 27663
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0709452084697665e-05,
+      "loss": 0.1856,
+      "step": 27664
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.07091162493913e-05,
+      "loss": 0.1942,
+      "step": 27665
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0708780414084933e-05,
+      "loss": 0.2869,
+      "step": 27666
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0708444578778568e-05,
+      "loss": 0.2203,
+      "step": 27667
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0708108743472202e-05,
+      "loss": 0.2973,
+      "step": 27668
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0707772908165837e-05,
+      "loss": 0.2278,
+      "step": 27669
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.070743707285947e-05,
+      "loss": 0.2994,
+      "step": 27670
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0707101237553106e-05,
+      "loss": 0.2382,
+      "step": 27671
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.070676540224674e-05,
+      "loss": 0.3609,
+      "step": 27672
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0706429566940372e-05,
+      "loss": 0.1901,
+      "step": 27673
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0706093731634008e-05,
+      "loss": 0.34,
+      "step": 27674
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0705757896327641e-05,
+      "loss": 0.4603,
+      "step": 27675
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0705422061021276e-05,
+      "loss": 0.2655,
+      "step": 27676
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.070508622571491e-05,
+      "loss": 0.4053,
+      "step": 27677
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0704750390408545e-05,
+      "loss": 0.4997,
+      "step": 27678
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0704414555102178e-05,
+      "loss": 0.3282,
+      "step": 27679
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0704078719795815e-05,
+      "loss": 0.3855,
+      "step": 27680
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0703742884489448e-05,
+      "loss": 0.2432,
+      "step": 27681
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.070340704918308e-05,
+      "loss": 0.3302,
+      "step": 27682
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0703071213876717e-05,
+      "loss": 0.2676,
+      "step": 27683
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.070273537857035e-05,
+      "loss": 0.3186,
+      "step": 27684
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0702399543263985e-05,
+      "loss": 0.3316,
+      "step": 27685
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.070206370795762e-05,
+      "loss": 0.48,
+      "step": 27686
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0701727872651254e-05,
+      "loss": 0.2478,
+      "step": 27687
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0701392037344887e-05,
+      "loss": 0.344,
+      "step": 27688
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0701056202038521e-05,
+      "loss": 0.2094,
+      "step": 27689
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0700720366732156e-05,
+      "loss": 0.3817,
+      "step": 27690
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0700384531425789e-05,
+      "loss": 0.3719,
+      "step": 27691
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.0700048696119425e-05,
+      "loss": 0.2522,
+      "step": 27692
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0699712860813058e-05,
+      "loss": 0.3137,
+      "step": 27693
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0699377025506691e-05,
+      "loss": 0.3191,
+      "step": 27694
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0699041190200328e-05,
+      "loss": 0.3868,
+      "step": 27695
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.069870535489396e-05,
+      "loss": 0.3803,
+      "step": 27696
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0698369519587595e-05,
+      "loss": 0.2878,
+      "step": 27697
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.069803368428123e-05,
+      "loss": 0.2268,
+      "step": 27698
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0697697848974865e-05,
+      "loss": 0.198,
+      "step": 27699
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0697362013668497e-05,
+      "loss": 0.2954,
+      "step": 27700
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0697026178362134e-05,
+      "loss": 0.2312,
+      "step": 27701
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0696690343055767e-05,
+      "loss": 0.2819,
+      "step": 27702
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.06963545077494e-05,
+      "loss": 0.2181,
+      "step": 27703
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0696018672443034e-05,
+      "loss": 0.2294,
+      "step": 27704
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0695682837136669e-05,
+      "loss": 0.3001,
+      "step": 27705
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0695347001830304e-05,
+      "loss": 0.1168,
+      "step": 27706
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0695011166523937e-05,
+      "loss": 0.2609,
+      "step": 27707
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0694675331217573e-05,
+      "loss": 0.4263,
+      "step": 27708
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0694339495911206e-05,
+      "loss": 0.2844,
+      "step": 27709
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0694003660604839e-05,
+      "loss": 0.183,
+      "step": 27710
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0693667825298475e-05,
+      "loss": 0.2281,
+      "step": 27711
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0693331989992108e-05,
+      "loss": 0.2601,
+      "step": 27712
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0692996154685741e-05,
+      "loss": 0.2836,
+      "step": 27713
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0692660319379377e-05,
+      "loss": 0.2276,
+      "step": 27714
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.069232448407301e-05,
+      "loss": 0.2862,
+      "step": 27715
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0691988648766645e-05,
+      "loss": 0.2856,
+      "step": 27716
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.069165281346028e-05,
+      "loss": 0.1676,
+      "step": 27717
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0691316978153914e-05,
+      "loss": 0.2724,
+      "step": 27718
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0690981142847547e-05,
+      "loss": 0.2383,
+      "step": 27719
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0690645307541184e-05,
+      "loss": 0.374,
+      "step": 27720
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0690309472234817e-05,
+      "loss": 0.2647,
+      "step": 27721
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.068997363692845e-05,
+      "loss": 0.233,
+      "step": 27722
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0689637801622086e-05,
+      "loss": 0.2976,
+      "step": 27723
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0689301966315719e-05,
+      "loss": 0.2851,
+      "step": 27724
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0688966131009353e-05,
+      "loss": 0.3192,
+      "step": 27725
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0688630295702988e-05,
+      "loss": 0.1626,
+      "step": 27726
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0688294460396623e-05,
+      "loss": 0.1917,
+      "step": 27727
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0687958625090256e-05,
+      "loss": 0.2091,
+      "step": 27728
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0687622789783892e-05,
+      "loss": 0.4723,
+      "step": 27729
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0687286954477525e-05,
+      "loss": 0.2808,
+      "step": 27730
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0686951119171158e-05,
+      "loss": 0.209,
+      "step": 27731
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0686615283864794e-05,
+      "loss": 0.2225,
+      "step": 27732
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0686279448558427e-05,
+      "loss": 0.3531,
+      "step": 27733
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0685943613252062e-05,
+      "loss": 0.2262,
+      "step": 27734
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0685607777945697e-05,
+      "loss": 0.1579,
+      "step": 27735
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.068527194263933e-05,
+      "loss": 0.1866,
+      "step": 27736
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0684936107332964e-05,
+      "loss": 0.1961,
+      "step": 27737
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0684600272026599e-05,
+      "loss": 0.2551,
+      "step": 27738
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0684264436720233e-05,
+      "loss": 0.2729,
+      "step": 27739
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0683928601413866e-05,
+      "loss": 0.3017,
+      "step": 27740
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0683592766107503e-05,
+      "loss": 0.2411,
+      "step": 27741
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0683256930801136e-05,
+      "loss": 0.2587,
+      "step": 27742
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0682921095494769e-05,
+      "loss": 0.1451,
+      "step": 27743
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0682585260188405e-05,
+      "loss": 0.1789,
+      "step": 27744
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0682249424882038e-05,
+      "loss": 0.2818,
+      "step": 27745
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0681913589575673e-05,
+      "loss": 0.4439,
+      "step": 27746
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0681577754269307e-05,
+      "loss": 0.1718,
+      "step": 27747
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0681241918962942e-05,
+      "loss": 0.2427,
+      "step": 27748
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0680906083656575e-05,
+      "loss": 0.3345,
+      "step": 27749
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0680570248350211e-05,
+      "loss": 0.2015,
+      "step": 27750
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0680234413043844e-05,
+      "loss": 0.206,
+      "step": 27751
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0679898577737477e-05,
+      "loss": 0.3478,
+      "step": 27752
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0679562742431113e-05,
+      "loss": 0.2522,
+      "step": 27753
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0679226907124746e-05,
+      "loss": 0.3095,
+      "step": 27754
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0678891071818381e-05,
+      "loss": 0.2805,
+      "step": 27755
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0678555236512016e-05,
+      "loss": 0.2995,
+      "step": 27756
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0678219401205649e-05,
+      "loss": 0.5297,
+      "step": 27757
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0677883565899283e-05,
+      "loss": 0.4471,
+      "step": 27758
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0677547730592918e-05,
+      "loss": 0.3415,
+      "step": 27759
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0677211895286553e-05,
+      "loss": 0.3866,
+      "step": 27760
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0676876059980186e-05,
+      "loss": 0.4171,
+      "step": 27761
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0676540224673822e-05,
+      "loss": 0.386,
+      "step": 27762
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0676204389367455e-05,
+      "loss": 0.2648,
+      "step": 27763
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0675868554061088e-05,
+      "loss": 0.2146,
+      "step": 27764
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0675532718754724e-05,
+      "loss": 0.2397,
+      "step": 27765
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0675196883448357e-05,
+      "loss": 0.2303,
+      "step": 27766
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0674861048141992e-05,
+      "loss": 0.4005,
+      "step": 27767
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0674525212835626e-05,
+      "loss": 0.2505,
+      "step": 27768
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0674189377529261e-05,
+      "loss": 0.2792,
+      "step": 27769
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0673853542222894e-05,
+      "loss": 0.3017,
+      "step": 27770
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.067351770691653e-05,
+      "loss": 0.2607,
+      "step": 27771
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0673181871610163e-05,
+      "loss": 0.3117,
+      "step": 27772
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0672846036303796e-05,
+      "loss": 0.3695,
+      "step": 27773
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0672510200997433e-05,
+      "loss": 0.2524,
+      "step": 27774
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0672174365691066e-05,
+      "loss": 0.3299,
+      "step": 27775
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.06718385303847e-05,
+      "loss": 0.2673,
+      "step": 27776
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0671502695078335e-05,
+      "loss": 0.2347,
+      "step": 27777
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.067116685977197e-05,
+      "loss": 0.3886,
+      "step": 27778
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0670831024465602e-05,
+      "loss": 0.1863,
+      "step": 27779
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0670495189159237e-05,
+      "loss": 0.2035,
+      "step": 27780
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0670159353852872e-05,
+      "loss": 0.3903,
+      "step": 27781
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0669823518546505e-05,
+      "loss": 0.2006,
+      "step": 27782
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0669487683240141e-05,
+      "loss": 0.2256,
+      "step": 27783
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0669151847933774e-05,
+      "loss": 0.3767,
+      "step": 27784
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0668816012627407e-05,
+      "loss": 0.3124,
+      "step": 27785
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0668480177321043e-05,
+      "loss": 0.237,
+      "step": 27786
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0668144342014676e-05,
+      "loss": 0.252,
+      "step": 27787
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0667808506708311e-05,
+      "loss": 0.4104,
+      "step": 27788
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0667472671401946e-05,
+      "loss": 0.2666,
+      "step": 27789
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.066713683609558e-05,
+      "loss": 0.3454,
+      "step": 27790
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0666801000789213e-05,
+      "loss": 0.2333,
+      "step": 27791
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.066646516548285e-05,
+      "loss": 0.2776,
+      "step": 27792
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0666129330176482e-05,
+      "loss": 0.3344,
+      "step": 27793
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0665793494870115e-05,
+      "loss": 0.2934,
+      "step": 27794
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0665457659563752e-05,
+      "loss": 0.2813,
+      "step": 27795
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0665121824257385e-05,
+      "loss": 0.2865,
+      "step": 27796
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.066478598895102e-05,
+      "loss": 0.3186,
+      "step": 27797
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0664450153644654e-05,
+      "loss": 0.2299,
+      "step": 27798
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0664114318338289e-05,
+      "loss": 0.2629,
+      "step": 27799
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0663778483031922e-05,
+      "loss": 0.2205,
+      "step": 27800
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0663442647725556e-05,
+      "loss": 0.2937,
+      "step": 27801
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0663106812419191e-05,
+      "loss": 0.1992,
+      "step": 27802
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0662770977112824e-05,
+      "loss": 0.1724,
+      "step": 27803
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.066243514180646e-05,
+      "loss": 0.286,
+      "step": 27804
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0662099306500093e-05,
+      "loss": 0.2863,
+      "step": 27805
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0661763471193726e-05,
+      "loss": 0.1649,
+      "step": 27806
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0661427635887362e-05,
+      "loss": 0.4347,
+      "step": 27807
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0661091800580995e-05,
+      "loss": 0.2999,
+      "step": 27808
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.066075596527463e-05,
+      "loss": 0.2627,
+      "step": 27809
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0660420129968265e-05,
+      "loss": 0.3026,
+      "step": 27810
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.06600842946619e-05,
+      "loss": 0.2689,
+      "step": 27811
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0659748459355532e-05,
+      "loss": 0.5727,
+      "step": 27812
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0659412624049169e-05,
+      "loss": 0.2516,
+      "step": 27813
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0659076788742802e-05,
+      "loss": 0.2856,
+      "step": 27814
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0658740953436434e-05,
+      "loss": 0.3244,
+      "step": 27815
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.065840511813007e-05,
+      "loss": 0.3264,
+      "step": 27816
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0658069282823704e-05,
+      "loss": 0.2483,
+      "step": 27817
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0657733447517338e-05,
+      "loss": 0.1925,
+      "step": 27818
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0657397612210973e-05,
+      "loss": 0.2441,
+      "step": 27819
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0657061776904608e-05,
+      "loss": 0.413,
+      "step": 27820
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.065672594159824e-05,
+      "loss": 0.1736,
+      "step": 27821
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0656390106291877e-05,
+      "loss": 0.2508,
+      "step": 27822
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.065605427098551e-05,
+      "loss": 0.1389,
+      "step": 27823
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0655718435679143e-05,
+      "loss": 0.224,
+      "step": 27824
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.065538260037278e-05,
+      "loss": 0.246,
+      "step": 27825
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0655046765066412e-05,
+      "loss": 0.2688,
+      "step": 27826
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0654710929760045e-05,
+      "loss": 0.2734,
+      "step": 27827
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0654375094453682e-05,
+      "loss": 0.4676,
+      "step": 27828
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0654039259147314e-05,
+      "loss": 0.282,
+      "step": 27829
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0653703423840949e-05,
+      "loss": 0.373,
+      "step": 27830
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0653367588534584e-05,
+      "loss": 0.2584,
+      "step": 27831
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0653031753228218e-05,
+      "loss": 0.5116,
+      "step": 27832
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0652695917921851e-05,
+      "loss": 0.2337,
+      "step": 27833
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0652360082615488e-05,
+      "loss": 0.3059,
+      "step": 27834
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.065202424730912e-05,
+      "loss": 0.3641,
+      "step": 27835
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0651688412002754e-05,
+      "loss": 0.1608,
+      "step": 27836
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.065135257669639e-05,
+      "loss": 0.3068,
+      "step": 27837
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0651016741390023e-05,
+      "loss": 0.1879,
+      "step": 27838
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0650680906083658e-05,
+      "loss": 0.2275,
+      "step": 27839
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0650345070777292e-05,
+      "loss": 0.3737,
+      "step": 27840
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0650009235470927e-05,
+      "loss": 0.2465,
+      "step": 27841
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.064967340016456e-05,
+      "loss": 0.276,
+      "step": 27842
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0649337564858196e-05,
+      "loss": 0.2642,
+      "step": 27843
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0649001729551829e-05,
+      "loss": 0.2564,
+      "step": 27844
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0648665894245462e-05,
+      "loss": 0.3427,
+      "step": 27845
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0648330058939098e-05,
+      "loss": 0.4338,
+      "step": 27846
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0647994223632731e-05,
+      "loss": 0.2514,
+      "step": 27847
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0647658388326364e-05,
+      "loss": 0.3016,
+      "step": 27848
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.064732255302e-05,
+      "loss": 0.274,
+      "step": 27849
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0646986717713634e-05,
+      "loss": 0.2032,
+      "step": 27850
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0646650882407268e-05,
+      "loss": 0.5339,
+      "step": 27851
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0646315047100903e-05,
+      "loss": 0.3778,
+      "step": 27852
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0645979211794538e-05,
+      "loss": 0.3468,
+      "step": 27853
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.064564337648817e-05,
+      "loss": 0.3592,
+      "step": 27854
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0645307541181807e-05,
+      "loss": 0.3679,
+      "step": 27855
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.064497170587544e-05,
+      "loss": 0.339,
+      "step": 27856
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0644635870569073e-05,
+      "loss": 0.3956,
+      "step": 27857
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0644300035262709e-05,
+      "loss": 0.4203,
+      "step": 27858
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0643964199956342e-05,
+      "loss": 0.1998,
+      "step": 27859
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0643628364649977e-05,
+      "loss": 0.4549,
+      "step": 27860
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0643292529343611e-05,
+      "loss": 0.1587,
+      "step": 27861
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0642956694037246e-05,
+      "loss": 0.4806,
+      "step": 27862
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0642620858730879e-05,
+      "loss": 0.1802,
+      "step": 27863
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0642285023424515e-05,
+      "loss": 0.3118,
+      "step": 27864
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0641949188118148e-05,
+      "loss": 0.2274,
+      "step": 27865
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0641613352811781e-05,
+      "loss": 0.2608,
+      "step": 27866
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0641277517505418e-05,
+      "loss": 0.3723,
+      "step": 27867
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.064094168219905e-05,
+      "loss": 0.1738,
+      "step": 27868
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0640605846892685e-05,
+      "loss": 0.3801,
+      "step": 27869
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.064027001158632e-05,
+      "loss": 0.356,
+      "step": 27870
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0639934176279953e-05,
+      "loss": 0.4383,
+      "step": 27871
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0639598340973587e-05,
+      "loss": 0.2405,
+      "step": 27872
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0639262505667222e-05,
+      "loss": 0.265,
+      "step": 27873
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0638926670360857e-05,
+      "loss": 0.3865,
+      "step": 27874
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.063859083505449e-05,
+      "loss": 0.4342,
+      "step": 27875
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0638254999748126e-05,
+      "loss": 0.2763,
+      "step": 27876
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0637919164441759e-05,
+      "loss": 0.3672,
+      "step": 27877
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0637583329135392e-05,
+      "loss": 0.1922,
+      "step": 27878
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0637247493829028e-05,
+      "loss": 0.3228,
+      "step": 27879
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0636911658522661e-05,
+      "loss": 0.1715,
+      "step": 27880
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0636575823216296e-05,
+      "loss": 0.3238,
+      "step": 27881
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0636239987909929e-05,
+      "loss": 0.3375,
+      "step": 27882
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0635904152603565e-05,
+      "loss": 0.3429,
+      "step": 27883
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0635568317297198e-05,
+      "loss": 0.4275,
+      "step": 27884
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0635232481990831e-05,
+      "loss": 0.2703,
+      "step": 27885
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0634896646684467e-05,
+      "loss": 0.3325,
+      "step": 27886
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.06345608113781e-05,
+      "loss": 0.3393,
+      "step": 27887
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0634224976071735e-05,
+      "loss": 0.2495,
+      "step": 27888
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.063388914076537e-05,
+      "loss": 0.1661,
+      "step": 27889
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 1.0633553305459004e-05,
+      "loss": 0.3958,
+      "step": 27890
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0633217470152637e-05,
+      "loss": 0.4265,
+      "step": 27891
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0632881634846272e-05,
+      "loss": 0.2287,
+      "step": 27892
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0632545799539906e-05,
+      "loss": 0.3895,
+      "step": 27893
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.063220996423354e-05,
+      "loss": 0.2669,
+      "step": 27894
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0631874128927176e-05,
+      "loss": 0.299,
+      "step": 27895
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0631538293620809e-05,
+      "loss": 0.2751,
+      "step": 27896
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0631202458314442e-05,
+      "loss": 0.2627,
+      "step": 27897
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0630866623008078e-05,
+      "loss": 0.336,
+      "step": 27898
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0630530787701711e-05,
+      "loss": 0.2299,
+      "step": 27899
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0630194952395346e-05,
+      "loss": 0.3035,
+      "step": 27900
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.062985911708898e-05,
+      "loss": 0.311,
+      "step": 27901
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0629523281782615e-05,
+      "loss": 0.3507,
+      "step": 27902
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0629187446476248e-05,
+      "loss": 0.1945,
+      "step": 27903
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0628851611169884e-05,
+      "loss": 0.171,
+      "step": 27904
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0628515775863517e-05,
+      "loss": 0.3232,
+      "step": 27905
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.062817994055715e-05,
+      "loss": 0.3066,
+      "step": 27906
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0627844105250786e-05,
+      "loss": 0.3808,
+      "step": 27907
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.062750826994442e-05,
+      "loss": 0.1844,
+      "step": 27908
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0627172434638054e-05,
+      "loss": 0.44,
+      "step": 27909
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0626836599331689e-05,
+      "loss": 0.3344,
+      "step": 27910
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0626500764025323e-05,
+      "loss": 0.2652,
+      "step": 27911
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0626164928718956e-05,
+      "loss": 0.1689,
+      "step": 27912
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0625829093412593e-05,
+      "loss": 0.2269,
+      "step": 27913
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0625493258106226e-05,
+      "loss": 0.3376,
+      "step": 27914
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0625157422799859e-05,
+      "loss": 0.2161,
+      "step": 27915
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0624821587493495e-05,
+      "loss": 0.3515,
+      "step": 27916
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0624485752187128e-05,
+      "loss": 0.238,
+      "step": 27917
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.062414991688076e-05,
+      "loss": 0.1539,
+      "step": 27918
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0623814081574397e-05,
+      "loss": 0.1965,
+      "step": 27919
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.062347824626803e-05,
+      "loss": 0.2798,
+      "step": 27920
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0623142410961665e-05,
+      "loss": 0.2566,
+      "step": 27921
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.06228065756553e-05,
+      "loss": 0.3384,
+      "step": 27922
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0622470740348934e-05,
+      "loss": 0.3867,
+      "step": 27923
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0622134905042567e-05,
+      "loss": 0.2617,
+      "step": 27924
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0621799069736203e-05,
+      "loss": 0.3309,
+      "step": 27925
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0621463234429836e-05,
+      "loss": 0.3541,
+      "step": 27926
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.062112739912347e-05,
+      "loss": 0.4311,
+      "step": 27927
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0620791563817106e-05,
+      "loss": 0.1557,
+      "step": 27928
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0620455728510739e-05,
+      "loss": 0.2935,
+      "step": 27929
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0620119893204373e-05,
+      "loss": 0.4618,
+      "step": 27930
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0619784057898008e-05,
+      "loss": 0.3048,
+      "step": 27931
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0619448222591642e-05,
+      "loss": 0.2585,
+      "step": 27932
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0619112387285275e-05,
+      "loss": 0.3723,
+      "step": 27933
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0618776551978912e-05,
+      "loss": 0.41,
+      "step": 27934
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0618440716672545e-05,
+      "loss": 0.3637,
+      "step": 27935
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0618104881366178e-05,
+      "loss": 0.2875,
+      "step": 27936
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0617769046059814e-05,
+      "loss": 0.2368,
+      "step": 27937
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0617433210753447e-05,
+      "loss": 0.3685,
+      "step": 27938
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.061709737544708e-05,
+      "loss": 0.271,
+      "step": 27939
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0616761540140716e-05,
+      "loss": 0.2462,
+      "step": 27940
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.061642570483435e-05,
+      "loss": 0.2669,
+      "step": 27941
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0616089869527984e-05,
+      "loss": 0.2786,
+      "step": 27942
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0615754034221619e-05,
+      "loss": 0.1822,
+      "step": 27943
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0615418198915253e-05,
+      "loss": 0.3125,
+      "step": 27944
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0615082363608886e-05,
+      "loss": 0.3506,
+      "step": 27945
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0614746528302522e-05,
+      "loss": 0.3349,
+      "step": 27946
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0614410692996155e-05,
+      "loss": 0.2069,
+      "step": 27947
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0614074857689788e-05,
+      "loss": 0.2773,
+      "step": 27948
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0613739022383425e-05,
+      "loss": 0.2755,
+      "step": 27949
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0613403187077058e-05,
+      "loss": 0.2628,
+      "step": 27950
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0613067351770692e-05,
+      "loss": 0.2018,
+      "step": 27951
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0612731516464327e-05,
+      "loss": 0.3268,
+      "step": 27952
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0612395681157962e-05,
+      "loss": 0.4625,
+      "step": 27953
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0612059845851595e-05,
+      "loss": 0.4677,
+      "step": 27954
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0611724010545231e-05,
+      "loss": 0.3163,
+      "step": 27955
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0611388175238864e-05,
+      "loss": 0.3339,
+      "step": 27956
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0611052339932497e-05,
+      "loss": 0.4277,
+      "step": 27957
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0610716504626133e-05,
+      "loss": 0.3141,
+      "step": 27958
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0610380669319766e-05,
+      "loss": 0.2748,
+      "step": 27959
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0610044834013399e-05,
+      "loss": 0.4691,
+      "step": 27960
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0609708998707035e-05,
+      "loss": 0.2429,
+      "step": 27961
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0609373163400668e-05,
+      "loss": 0.3251,
+      "step": 27962
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0609037328094303e-05,
+      "loss": 0.1951,
+      "step": 27963
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0608701492787938e-05,
+      "loss": 0.2322,
+      "step": 27964
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0608365657481572e-05,
+      "loss": 0.5784,
+      "step": 27965
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0608029822175205e-05,
+      "loss": 0.2498,
+      "step": 27966
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0607693986868842e-05,
+      "loss": 0.3117,
+      "step": 27967
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0607358151562475e-05,
+      "loss": 0.4314,
+      "step": 27968
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0607022316256108e-05,
+      "loss": 0.2028,
+      "step": 27969
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0606686480949744e-05,
+      "loss": 0.362,
+      "step": 27970
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0606350645643377e-05,
+      "loss": 0.3072,
+      "step": 27971
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0606014810337011e-05,
+      "loss": 0.382,
+      "step": 27972
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0605678975030646e-05,
+      "loss": 0.3297,
+      "step": 27973
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.060534313972428e-05,
+      "loss": 0.2198,
+      "step": 27974
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0605007304417914e-05,
+      "loss": 0.2545,
+      "step": 27975
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.060467146911155e-05,
+      "loss": 0.2604,
+      "step": 27976
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0604335633805183e-05,
+      "loss": 0.3703,
+      "step": 27977
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0603999798498816e-05,
+      "loss": 0.4247,
+      "step": 27978
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0603663963192452e-05,
+      "loss": 0.2779,
+      "step": 27979
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0603328127886085e-05,
+      "loss": 0.4064,
+      "step": 27980
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.060299229257972e-05,
+      "loss": 0.1644,
+      "step": 27981
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0602656457273355e-05,
+      "loss": 0.2792,
+      "step": 27982
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0602320621966987e-05,
+      "loss": 0.2669,
+      "step": 27983
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0601984786660622e-05,
+      "loss": 0.2273,
+      "step": 27984
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0601648951354257e-05,
+      "loss": 0.1808,
+      "step": 27985
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0601313116047891e-05,
+      "loss": 0.2471,
+      "step": 27986
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0600977280741524e-05,
+      "loss": 0.3964,
+      "step": 27987
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.060064144543516e-05,
+      "loss": 0.2809,
+      "step": 27988
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0600305610128794e-05,
+      "loss": 0.3663,
+      "step": 27989
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0599969774822427e-05,
+      "loss": 0.3079,
+      "step": 27990
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0599633939516063e-05,
+      "loss": 0.2888,
+      "step": 27991
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0599298104209696e-05,
+      "loss": 0.2202,
+      "step": 27992
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.059896226890333e-05,
+      "loss": 0.3113,
+      "step": 27993
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0598626433596965e-05,
+      "loss": 0.2502,
+      "step": 27994
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.05982905982906e-05,
+      "loss": 0.2974,
+      "step": 27995
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0597954762984233e-05,
+      "loss": 0.3235,
+      "step": 27996
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.059761892767787e-05,
+      "loss": 0.3842,
+      "step": 27997
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0597283092371502e-05,
+      "loss": 0.334,
+      "step": 27998
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0596947257065135e-05,
+      "loss": 0.3392,
+      "step": 27999
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0596611421758771e-05,
+      "loss": 0.3241,
+      "step": 28000
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0596275586452404e-05,
+      "loss": 0.3145,
+      "step": 28001
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0595939751146039e-05,
+      "loss": 0.2277,
+      "step": 28002
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0595603915839674e-05,
+      "loss": 0.2525,
+      "step": 28003
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0595268080533308e-05,
+      "loss": 0.1891,
+      "step": 28004
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0594932245226941e-05,
+      "loss": 0.2505,
+      "step": 28005
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0594596409920576e-05,
+      "loss": 0.2953,
+      "step": 28006
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.059426057461421e-05,
+      "loss": 0.2507,
+      "step": 28007
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0593924739307844e-05,
+      "loss": 0.2028,
+      "step": 28008
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.059358890400148e-05,
+      "loss": 0.243,
+      "step": 28009
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0593253068695113e-05,
+      "loss": 0.3592,
+      "step": 28010
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0592917233388746e-05,
+      "loss": 0.2632,
+      "step": 28011
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0592581398082382e-05,
+      "loss": 0.1322,
+      "step": 28012
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0592245562776015e-05,
+      "loss": 0.2516,
+      "step": 28013
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.059190972746965e-05,
+      "loss": 0.2963,
+      "step": 28014
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0591573892163284e-05,
+      "loss": 0.255,
+      "step": 28015
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0591238056856919e-05,
+      "loss": 0.1665,
+      "step": 28016
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0590902221550552e-05,
+      "loss": 0.3438,
+      "step": 28017
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0590566386244188e-05,
+      "loss": 0.4113,
+      "step": 28018
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0590230550937821e-05,
+      "loss": 0.2632,
+      "step": 28019
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0589894715631454e-05,
+      "loss": 0.4055,
+      "step": 28020
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.058955888032509e-05,
+      "loss": 0.296,
+      "step": 28021
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0589223045018723e-05,
+      "loss": 0.3516,
+      "step": 28022
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0588887209712358e-05,
+      "loss": 0.2333,
+      "step": 28023
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0588551374405993e-05,
+      "loss": 0.207,
+      "step": 28024
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0588215539099627e-05,
+      "loss": 0.4462,
+      "step": 28025
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.058787970379326e-05,
+      "loss": 0.3809,
+      "step": 28026
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0587543868486895e-05,
+      "loss": 0.3923,
+      "step": 28027
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.058720803318053e-05,
+      "loss": 0.3887,
+      "step": 28028
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0586872197874163e-05,
+      "loss": 0.1982,
+      "step": 28029
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0586536362567799e-05,
+      "loss": 0.2264,
+      "step": 28030
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0586200527261432e-05,
+      "loss": 0.1929,
+      "step": 28031
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0585864691955065e-05,
+      "loss": 0.431,
+      "step": 28032
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0585528856648701e-05,
+      "loss": 0.3336,
+      "step": 28033
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0585193021342334e-05,
+      "loss": 0.2473,
+      "step": 28034
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0584857186035969e-05,
+      "loss": 0.359,
+      "step": 28035
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0584521350729603e-05,
+      "loss": 0.2298,
+      "step": 28036
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0584185515423238e-05,
+      "loss": 0.3342,
+      "step": 28037
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0583849680116871e-05,
+      "loss": 0.3171,
+      "step": 28038
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0583513844810507e-05,
+      "loss": 0.3349,
+      "step": 28039
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.058317800950414e-05,
+      "loss": 0.3355,
+      "step": 28040
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0582842174197773e-05,
+      "loss": 0.3302,
+      "step": 28041
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.058250633889141e-05,
+      "loss": 0.3627,
+      "step": 28042
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0582170503585043e-05,
+      "loss": 0.3915,
+      "step": 28043
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0581834668278677e-05,
+      "loss": 0.4019,
+      "step": 28044
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0581498832972312e-05,
+      "loss": 0.1452,
+      "step": 28045
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0581162997665947e-05,
+      "loss": 0.2958,
+      "step": 28046
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.058082716235958e-05,
+      "loss": 0.4299,
+      "step": 28047
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0580491327053216e-05,
+      "loss": 0.3841,
+      "step": 28048
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0580155491746849e-05,
+      "loss": 0.257,
+      "step": 28049
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0579819656440482e-05,
+      "loss": 0.4641,
+      "step": 28050
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0579483821134118e-05,
+      "loss": 0.3148,
+      "step": 28051
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0579147985827751e-05,
+      "loss": 0.2461,
+      "step": 28052
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0578812150521384e-05,
+      "loss": 0.2289,
+      "step": 28053
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.057847631521502e-05,
+      "loss": 0.2521,
+      "step": 28054
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0578140479908653e-05,
+      "loss": 0.2674,
+      "step": 28055
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0577804644602288e-05,
+      "loss": 0.4053,
+      "step": 28056
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0577468809295923e-05,
+      "loss": 0.3349,
+      "step": 28057
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0577132973989557e-05,
+      "loss": 0.2082,
+      "step": 28058
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.057679713868319e-05,
+      "loss": 0.3018,
+      "step": 28059
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0576461303376827e-05,
+      "loss": 0.4625,
+      "step": 28060
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.057612546807046e-05,
+      "loss": 0.3713,
+      "step": 28061
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0575789632764092e-05,
+      "loss": 0.3648,
+      "step": 28062
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0575453797457727e-05,
+      "loss": 0.498,
+      "step": 28063
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0575117962151362e-05,
+      "loss": 0.259,
+      "step": 28064
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0574782126844996e-05,
+      "loss": 0.2458,
+      "step": 28065
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.057444629153863e-05,
+      "loss": 0.3138,
+      "step": 28066
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0574110456232266e-05,
+      "loss": 0.2579,
+      "step": 28067
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0573774620925899e-05,
+      "loss": 0.1932,
+      "step": 28068
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0573438785619532e-05,
+      "loss": 0.2773,
+      "step": 28069
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0573102950313168e-05,
+      "loss": 0.1737,
+      "step": 28070
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0572767115006801e-05,
+      "loss": 0.2875,
+      "step": 28071
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0572431279700436e-05,
+      "loss": 0.2649,
+      "step": 28072
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.057209544439407e-05,
+      "loss": 0.2677,
+      "step": 28073
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0571759609087703e-05,
+      "loss": 0.1661,
+      "step": 28074
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0571423773781338e-05,
+      "loss": 0.4556,
+      "step": 28075
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0571087938474972e-05,
+      "loss": 0.4455,
+      "step": 28076
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0570752103168607e-05,
+      "loss": 0.2681,
+      "step": 28077
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.057041626786224e-05,
+      "loss": 0.2462,
+      "step": 28078
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0570080432555876e-05,
+      "loss": 0.261,
+      "step": 28079
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.056974459724951e-05,
+      "loss": 0.4265,
+      "step": 28080
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0569408761943142e-05,
+      "loss": 0.3129,
+      "step": 28081
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0569072926636779e-05,
+      "loss": 0.2701,
+      "step": 28082
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0568737091330412e-05,
+      "loss": 0.3609,
+      "step": 28083
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0568401256024046e-05,
+      "loss": 0.3561,
+      "step": 28084
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0568065420717681e-05,
+      "loss": 0.3752,
+      "step": 28085
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0567729585411316e-05,
+      "loss": 0.3768,
+      "step": 28086
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0567393750104948e-05,
+      "loss": 0.2723,
+      "step": 28087
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0567057914798585e-05,
+      "loss": 0.2095,
+      "step": 28088
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.0566722079492218e-05,
+      "loss": 0.2478,
+      "step": 28089
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.056638624418585e-05,
+      "loss": 0.2562,
+      "step": 28090
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0566050408879487e-05,
+      "loss": 0.2239,
+      "step": 28091
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.056571457357312e-05,
+      "loss": 0.2908,
+      "step": 28092
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0565378738266755e-05,
+      "loss": 0.2884,
+      "step": 28093
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.056504290296039e-05,
+      "loss": 0.3422,
+      "step": 28094
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0564707067654022e-05,
+      "loss": 0.2204,
+      "step": 28095
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0564371232347657e-05,
+      "loss": 0.1891,
+      "step": 28096
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0564035397041292e-05,
+      "loss": 0.284,
+      "step": 28097
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0563699561734926e-05,
+      "loss": 0.3897,
+      "step": 28098
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0563363726428559e-05,
+      "loss": 0.4675,
+      "step": 28099
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0563027891122195e-05,
+      "loss": 0.3662,
+      "step": 28100
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0562692055815828e-05,
+      "loss": 0.3378,
+      "step": 28101
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0562356220509461e-05,
+      "loss": 0.3351,
+      "step": 28102
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0562020385203098e-05,
+      "loss": 0.2664,
+      "step": 28103
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.056168454989673e-05,
+      "loss": 0.2322,
+      "step": 28104
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0561348714590365e-05,
+      "loss": 0.2075,
+      "step": 28105
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0561012879284e-05,
+      "loss": 0.326,
+      "step": 28106
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0560677043977635e-05,
+      "loss": 0.2779,
+      "step": 28107
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0560341208671268e-05,
+      "loss": 0.264,
+      "step": 28108
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0560005373364904e-05,
+      "loss": 0.408,
+      "step": 28109
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0559669538058537e-05,
+      "loss": 0.3357,
+      "step": 28110
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.055933370275217e-05,
+      "loss": 0.2477,
+      "step": 28111
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0558997867445806e-05,
+      "loss": 0.3062,
+      "step": 28112
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0558662032139439e-05,
+      "loss": 0.3437,
+      "step": 28113
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0558326196833074e-05,
+      "loss": 0.2559,
+      "step": 28114
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0557990361526708e-05,
+      "loss": 0.2639,
+      "step": 28115
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0557654526220343e-05,
+      "loss": 0.3093,
+      "step": 28116
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0557318690913976e-05,
+      "loss": 0.285,
+      "step": 28117
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.055698285560761e-05,
+      "loss": 0.1232,
+      "step": 28118
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0556647020301245e-05,
+      "loss": 0.2703,
+      "step": 28119
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0556311184994878e-05,
+      "loss": 0.3037,
+      "step": 28120
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0555975349688515e-05,
+      "loss": 0.3166,
+      "step": 28121
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0555639514382148e-05,
+      "loss": 0.2792,
+      "step": 28122
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.055530367907578e-05,
+      "loss": 0.2587,
+      "step": 28123
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0554967843769417e-05,
+      "loss": 0.3114,
+      "step": 28124
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.055463200846305e-05,
+      "loss": 0.4138,
+      "step": 28125
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0554296173156684e-05,
+      "loss": 0.2876,
+      "step": 28126
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0553960337850319e-05,
+      "loss": 0.2248,
+      "step": 28127
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0553624502543954e-05,
+      "loss": 0.5006,
+      "step": 28128
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0553288667237587e-05,
+      "loss": 0.2398,
+      "step": 28129
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0552952831931223e-05,
+      "loss": 0.3336,
+      "step": 28130
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0552616996624856e-05,
+      "loss": 0.3178,
+      "step": 28131
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0552281161318489e-05,
+      "loss": 0.2646,
+      "step": 28132
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0551945326012125e-05,
+      "loss": 0.3168,
+      "step": 28133
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0551609490705758e-05,
+      "loss": 0.2365,
+      "step": 28134
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0551273655399393e-05,
+      "loss": 0.1805,
+      "step": 28135
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0550937820093028e-05,
+      "loss": 0.2561,
+      "step": 28136
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0550601984786662e-05,
+      "loss": 0.1873,
+      "step": 28137
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0550266149480295e-05,
+      "loss": 0.1889,
+      "step": 28138
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0549930314173931e-05,
+      "loss": 0.2523,
+      "step": 28139
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0549594478867564e-05,
+      "loss": 0.208,
+      "step": 28140
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0549258643561197e-05,
+      "loss": 0.466,
+      "step": 28141
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0548922808254834e-05,
+      "loss": 0.2564,
+      "step": 28142
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0548586972948467e-05,
+      "loss": 0.3452,
+      "step": 28143
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.05482511376421e-05,
+      "loss": 0.2567,
+      "step": 28144
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0547915302335736e-05,
+      "loss": 0.2859,
+      "step": 28145
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0547579467029369e-05,
+      "loss": 0.217,
+      "step": 28146
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0547243631723004e-05,
+      "loss": 0.2043,
+      "step": 28147
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0546907796416638e-05,
+      "loss": 0.2403,
+      "step": 28148
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0546571961110273e-05,
+      "loss": 0.3853,
+      "step": 28149
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0546236125803906e-05,
+      "loss": 0.4135,
+      "step": 28150
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0545900290497542e-05,
+      "loss": 0.1846,
+      "step": 28151
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0545564455191175e-05,
+      "loss": 0.3393,
+      "step": 28152
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0545228619884808e-05,
+      "loss": 0.3144,
+      "step": 28153
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0544892784578444e-05,
+      "loss": 0.4047,
+      "step": 28154
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0544556949272077e-05,
+      "loss": 0.3612,
+      "step": 28155
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0544221113965712e-05,
+      "loss": 0.2861,
+      "step": 28156
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0543885278659347e-05,
+      "loss": 0.2528,
+      "step": 28157
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0543549443352981e-05,
+      "loss": 0.2962,
+      "step": 28158
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0543213608046614e-05,
+      "loss": 0.067,
+      "step": 28159
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.054287777274025e-05,
+      "loss": 0.3126,
+      "step": 28160
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0542541937433884e-05,
+      "loss": 0.257,
+      "step": 28161
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0542206102127517e-05,
+      "loss": 0.581,
+      "step": 28162
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0541870266821153e-05,
+      "loss": 0.2401,
+      "step": 28163
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0541534431514786e-05,
+      "loss": 0.3857,
+      "step": 28164
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0541198596208419e-05,
+      "loss": 0.3867,
+      "step": 28165
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0540862760902055e-05,
+      "loss": 0.2844,
+      "step": 28166
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0540526925595688e-05,
+      "loss": 0.4682,
+      "step": 28167
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0540191090289323e-05,
+      "loss": 0.3375,
+      "step": 28168
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0539855254982957e-05,
+      "loss": 0.3003,
+      "step": 28169
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0539519419676592e-05,
+      "loss": 0.3045,
+      "step": 28170
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0539183584370225e-05,
+      "loss": 0.1957,
+      "step": 28171
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0538847749063861e-05,
+      "loss": 0.2596,
+      "step": 28172
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0538511913757494e-05,
+      "loss": 0.1419,
+      "step": 28173
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0538176078451127e-05,
+      "loss": 0.3376,
+      "step": 28174
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0537840243144764e-05,
+      "loss": 0.2765,
+      "step": 28175
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0537504407838397e-05,
+      "loss": 0.2235,
+      "step": 28176
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0537168572532031e-05,
+      "loss": 0.1635,
+      "step": 28177
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0536832737225666e-05,
+      "loss": 0.2291,
+      "step": 28178
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.05364969019193e-05,
+      "loss": 0.2785,
+      "step": 28179
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0536161066612933e-05,
+      "loss": 0.3508,
+      "step": 28180
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.053582523130657e-05,
+      "loss": 0.2972,
+      "step": 28181
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0535489396000203e-05,
+      "loss": 0.3239,
+      "step": 28182
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0535153560693836e-05,
+      "loss": 0.2946,
+      "step": 28183
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0534817725387472e-05,
+      "loss": 0.508,
+      "step": 28184
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0534481890081105e-05,
+      "loss": 0.3674,
+      "step": 28185
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0534146054774738e-05,
+      "loss": 0.2158,
+      "step": 28186
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0533810219468374e-05,
+      "loss": 0.2275,
+      "step": 28187
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0533474384162007e-05,
+      "loss": 0.2034,
+      "step": 28188
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0533138548855642e-05,
+      "loss": 0.5828,
+      "step": 28189
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0532802713549276e-05,
+      "loss": 0.1795,
+      "step": 28190
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0532466878242911e-05,
+      "loss": 0.3398,
+      "step": 28191
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0532131042936544e-05,
+      "loss": 0.3385,
+      "step": 28192
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.053179520763018e-05,
+      "loss": 0.3535,
+      "step": 28193
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0531459372323813e-05,
+      "loss": 0.2221,
+      "step": 28194
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0531123537017446e-05,
+      "loss": 0.299,
+      "step": 28195
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0530787701711083e-05,
+      "loss": 0.1437,
+      "step": 28196
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0530451866404716e-05,
+      "loss": 0.2771,
+      "step": 28197
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.053011603109835e-05,
+      "loss": 0.208,
+      "step": 28198
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0529780195791985e-05,
+      "loss": 0.4022,
+      "step": 28199
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.052944436048562e-05,
+      "loss": 0.335,
+      "step": 28200
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0529108525179253e-05,
+      "loss": 0.4238,
+      "step": 28201
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0528772689872889e-05,
+      "loss": 0.3429,
+      "step": 28202
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0528436854566522e-05,
+      "loss": 0.3437,
+      "step": 28203
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0528101019260155e-05,
+      "loss": 0.3279,
+      "step": 28204
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0527765183953791e-05,
+      "loss": 0.2092,
+      "step": 28205
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0527429348647424e-05,
+      "loss": 0.2529,
+      "step": 28206
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0527093513341059e-05,
+      "loss": 0.605,
+      "step": 28207
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0526757678034693e-05,
+      "loss": 0.4714,
+      "step": 28208
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0526421842728326e-05,
+      "loss": 0.2177,
+      "step": 28209
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0526086007421961e-05,
+      "loss": 0.2816,
+      "step": 28210
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0525750172115596e-05,
+      "loss": 0.3755,
+      "step": 28211
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.052541433680923e-05,
+      "loss": 0.2843,
+      "step": 28212
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0525078501502863e-05,
+      "loss": 0.3875,
+      "step": 28213
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.05247426661965e-05,
+      "loss": 0.2112,
+      "step": 28214
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0524406830890133e-05,
+      "loss": 0.4657,
+      "step": 28215
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0524070995583765e-05,
+      "loss": 0.4427,
+      "step": 28216
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0523735160277402e-05,
+      "loss": 0.3743,
+      "step": 28217
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0523399324971035e-05,
+      "loss": 0.2684,
+      "step": 28218
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.052306348966467e-05,
+      "loss": 0.2813,
+      "step": 28219
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0522727654358304e-05,
+      "loss": 0.4405,
+      "step": 28220
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0522391819051939e-05,
+      "loss": 0.2064,
+      "step": 28221
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0522055983745572e-05,
+      "loss": 0.3281,
+      "step": 28222
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0521720148439208e-05,
+      "loss": 0.1594,
+      "step": 28223
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0521384313132841e-05,
+      "loss": 0.2527,
+      "step": 28224
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0521048477826474e-05,
+      "loss": 0.3213,
+      "step": 28225
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.052071264252011e-05,
+      "loss": 0.2518,
+      "step": 28226
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0520376807213743e-05,
+      "loss": 0.4326,
+      "step": 28227
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0520040971907378e-05,
+      "loss": 0.2131,
+      "step": 28228
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0519705136601012e-05,
+      "loss": 0.2077,
+      "step": 28229
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0519369301294645e-05,
+      "loss": 0.2047,
+      "step": 28230
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.051903346598828e-05,
+      "loss": 0.2245,
+      "step": 28231
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0518697630681915e-05,
+      "loss": 0.3994,
+      "step": 28232
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.051836179537555e-05,
+      "loss": 0.353,
+      "step": 28233
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0518025960069182e-05,
+      "loss": 0.4247,
+      "step": 28234
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0517690124762819e-05,
+      "loss": 0.2751,
+      "step": 28235
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0517354289456452e-05,
+      "loss": 0.2225,
+      "step": 28236
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0517018454150085e-05,
+      "loss": 0.2272,
+      "step": 28237
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0516682618843721e-05,
+      "loss": 0.3614,
+      "step": 28238
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0516346783537354e-05,
+      "loss": 0.2355,
+      "step": 28239
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0516010948230989e-05,
+      "loss": 0.2224,
+      "step": 28240
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0515675112924623e-05,
+      "loss": 0.2058,
+      "step": 28241
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0515339277618258e-05,
+      "loss": 0.3462,
+      "step": 28242
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.051500344231189e-05,
+      "loss": 0.377,
+      "step": 28243
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0514667607005524e-05,
+      "loss": 0.1877,
+      "step": 28244
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.051433177169916e-05,
+      "loss": 0.3126,
+      "step": 28245
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0513995936392793e-05,
+      "loss": 0.3326,
+      "step": 28246
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0513660101086428e-05,
+      "loss": 0.2412,
+      "step": 28247
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0513324265780062e-05,
+      "loss": 0.2341,
+      "step": 28248
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0512988430473697e-05,
+      "loss": 0.1947,
+      "step": 28249
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.051265259516733e-05,
+      "loss": 0.2075,
+      "step": 28250
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0512316759860966e-05,
+      "loss": 0.2715,
+      "step": 28251
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.05119809245546e-05,
+      "loss": 0.3174,
+      "step": 28252
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0511645089248232e-05,
+      "loss": 0.2407,
+      "step": 28253
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0511309253941869e-05,
+      "loss": 0.3153,
+      "step": 28254
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0510973418635501e-05,
+      "loss": 0.3497,
+      "step": 28255
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0510637583329134e-05,
+      "loss": 0.2871,
+      "step": 28256
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.051030174802277e-05,
+      "loss": 0.1355,
+      "step": 28257
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0509965912716404e-05,
+      "loss": 0.4592,
+      "step": 28258
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0509630077410038e-05,
+      "loss": 0.2475,
+      "step": 28259
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0509294242103673e-05,
+      "loss": 0.2262,
+      "step": 28260
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0508958406797308e-05,
+      "loss": 0.1985,
+      "step": 28261
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.050862257149094e-05,
+      "loss": 0.3459,
+      "step": 28262
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0508286736184577e-05,
+      "loss": 0.4297,
+      "step": 28263
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.050795090087821e-05,
+      "loss": 0.4098,
+      "step": 28264
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0507615065571843e-05,
+      "loss": 0.3599,
+      "step": 28265
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.050727923026548e-05,
+      "loss": 0.2153,
+      "step": 28266
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0506943394959112e-05,
+      "loss": 0.2289,
+      "step": 28267
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0506607559652747e-05,
+      "loss": 0.2776,
+      "step": 28268
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0506271724346381e-05,
+      "loss": 0.2363,
+      "step": 28269
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0505935889040016e-05,
+      "loss": 0.4606,
+      "step": 28270
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0505600053733649e-05,
+      "loss": 0.291,
+      "step": 28271
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0505264218427285e-05,
+      "loss": 0.3587,
+      "step": 28272
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0504928383120918e-05,
+      "loss": 0.2419,
+      "step": 28273
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0504592547814551e-05,
+      "loss": 0.278,
+      "step": 28274
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0504256712508188e-05,
+      "loss": 0.5222,
+      "step": 28275
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.050392087720182e-05,
+      "loss": 0.1821,
+      "step": 28276
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0503585041895454e-05,
+      "loss": 0.2411,
+      "step": 28277
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.050324920658909e-05,
+      "loss": 0.2978,
+      "step": 28278
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0502913371282723e-05,
+      "loss": 0.2925,
+      "step": 28279
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0502577535976357e-05,
+      "loss": 0.2105,
+      "step": 28280
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0502241700669992e-05,
+      "loss": 0.3337,
+      "step": 28281
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0501905865363627e-05,
+      "loss": 0.2108,
+      "step": 28282
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.050157003005726e-05,
+      "loss": 0.2784,
+      "step": 28283
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0501234194750896e-05,
+      "loss": 0.1262,
+      "step": 28284
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0500898359444529e-05,
+      "loss": 0.2586,
+      "step": 28285
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0500562524138162e-05,
+      "loss": 0.3885,
+      "step": 28286
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 1.0500226688831798e-05,
+      "loss": 0.4105,
+      "step": 28287
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0499890853525431e-05,
+      "loss": 0.3915,
+      "step": 28288
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0499555018219066e-05,
+      "loss": 0.3645,
+      "step": 28289
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.04992191829127e-05,
+      "loss": 0.2466,
+      "step": 28290
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0498883347606335e-05,
+      "loss": 0.2417,
+      "step": 28291
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0498547512299968e-05,
+      "loss": 0.2931,
+      "step": 28292
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0498211676993605e-05,
+      "loss": 0.4047,
+      "step": 28293
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0497875841687237e-05,
+      "loss": 0.5746,
+      "step": 28294
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.049754000638087e-05,
+      "loss": 0.2937,
+      "step": 28295
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0497204171074507e-05,
+      "loss": 0.349,
+      "step": 28296
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.049686833576814e-05,
+      "loss": 0.1617,
+      "step": 28297
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0496532500461774e-05,
+      "loss": 0.2993,
+      "step": 28298
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0496196665155409e-05,
+      "loss": 0.4484,
+      "step": 28299
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0495860829849042e-05,
+      "loss": 0.1723,
+      "step": 28300
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0495524994542677e-05,
+      "loss": 0.3533,
+      "step": 28301
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0495189159236311e-05,
+      "loss": 0.2154,
+      "step": 28302
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0494853323929946e-05,
+      "loss": 0.3637,
+      "step": 28303
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0494517488623579e-05,
+      "loss": 0.2636,
+      "step": 28304
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0494181653317215e-05,
+      "loss": 0.1695,
+      "step": 28305
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0493845818010848e-05,
+      "loss": 0.256,
+      "step": 28306
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0493509982704481e-05,
+      "loss": 0.4176,
+      "step": 28307
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0493174147398117e-05,
+      "loss": 0.3236,
+      "step": 28308
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.049283831209175e-05,
+      "loss": 0.2768,
+      "step": 28309
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0492502476785385e-05,
+      "loss": 0.3022,
+      "step": 28310
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.049216664147902e-05,
+      "loss": 0.2054,
+      "step": 28311
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0491830806172654e-05,
+      "loss": 0.3125,
+      "step": 28312
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0491494970866287e-05,
+      "loss": 0.2988,
+      "step": 28313
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0491159135559924e-05,
+      "loss": 0.2601,
+      "step": 28314
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0490823300253557e-05,
+      "loss": 0.5375,
+      "step": 28315
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.049048746494719e-05,
+      "loss": 0.2968,
+      "step": 28316
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0490151629640826e-05,
+      "loss": 0.1962,
+      "step": 28317
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0489815794334459e-05,
+      "loss": 0.3668,
+      "step": 28318
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0489479959028093e-05,
+      "loss": 0.4091,
+      "step": 28319
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0489144123721728e-05,
+      "loss": 0.2874,
+      "step": 28320
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0488808288415361e-05,
+      "loss": 0.3266,
+      "step": 28321
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0488472453108996e-05,
+      "loss": 0.3021,
+      "step": 28322
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.048813661780263e-05,
+      "loss": 0.3315,
+      "step": 28323
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0487800782496265e-05,
+      "loss": 0.2859,
+      "step": 28324
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0487464947189898e-05,
+      "loss": 0.4522,
+      "step": 28325
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0487129111883534e-05,
+      "loss": 0.304,
+      "step": 28326
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0486793276577167e-05,
+      "loss": 0.2358,
+      "step": 28327
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.04864574412708e-05,
+      "loss": 0.3023,
+      "step": 28328
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0486121605964437e-05,
+      "loss": 0.1992,
+      "step": 28329
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.048578577065807e-05,
+      "loss": 0.2332,
+      "step": 28330
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0485449935351704e-05,
+      "loss": 0.1888,
+      "step": 28331
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0485114100045339e-05,
+      "loss": 0.3612,
+      "step": 28332
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0484778264738973e-05,
+      "loss": 0.3383,
+      "step": 28333
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0484442429432606e-05,
+      "loss": 0.224,
+      "step": 28334
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0484106594126243e-05,
+      "loss": 0.2264,
+      "step": 28335
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0483770758819876e-05,
+      "loss": 0.1998,
+      "step": 28336
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0483434923513509e-05,
+      "loss": 0.2902,
+      "step": 28337
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0483099088207145e-05,
+      "loss": 0.2549,
+      "step": 28338
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0482763252900778e-05,
+      "loss": 0.228,
+      "step": 28339
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0482427417594413e-05,
+      "loss": 0.3136,
+      "step": 28340
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0482091582288047e-05,
+      "loss": 0.2628,
+      "step": 28341
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0481755746981682e-05,
+      "loss": 0.3587,
+      "step": 28342
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0481419911675315e-05,
+      "loss": 0.2441,
+      "step": 28343
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.048108407636895e-05,
+      "loss": 0.4023,
+      "step": 28344
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0480748241062584e-05,
+      "loss": 0.2886,
+      "step": 28345
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0480412405756217e-05,
+      "loss": 0.4589,
+      "step": 28346
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0480076570449853e-05,
+      "loss": 0.428,
+      "step": 28347
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0479740735143486e-05,
+      "loss": 0.3918,
+      "step": 28348
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.047940489983712e-05,
+      "loss": 0.267,
+      "step": 28349
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0479069064530756e-05,
+      "loss": 0.4162,
+      "step": 28350
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0478733229224389e-05,
+      "loss": 0.4234,
+      "step": 28351
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0478397393918023e-05,
+      "loss": 0.2714,
+      "step": 28352
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0478061558611658e-05,
+      "loss": 0.2554,
+      "step": 28353
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0477725723305293e-05,
+      "loss": 0.209,
+      "step": 28354
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0477389887998926e-05,
+      "loss": 0.1669,
+      "step": 28355
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0477054052692562e-05,
+      "loss": 0.1876,
+      "step": 28356
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0476718217386195e-05,
+      "loss": 0.433,
+      "step": 28357
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0476382382079828e-05,
+      "loss": 0.3447,
+      "step": 28358
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0476046546773464e-05,
+      "loss": 0.373,
+      "step": 28359
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0475710711467097e-05,
+      "loss": 0.2697,
+      "step": 28360
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0475374876160732e-05,
+      "loss": 0.2283,
+      "step": 28361
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0475039040854366e-05,
+      "loss": 0.1867,
+      "step": 28362
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0474703205548001e-05,
+      "loss": 0.1668,
+      "step": 28363
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0474367370241634e-05,
+      "loss": 0.2812,
+      "step": 28364
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0474031534935269e-05,
+      "loss": 0.2311,
+      "step": 28365
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0473695699628903e-05,
+      "loss": 0.2753,
+      "step": 28366
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0473359864322536e-05,
+      "loss": 0.3012,
+      "step": 28367
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0473024029016173e-05,
+      "loss": 0.3185,
+      "step": 28368
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0472688193709806e-05,
+      "loss": 0.2418,
+      "step": 28369
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0472352358403438e-05,
+      "loss": 0.3297,
+      "step": 28370
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0472016523097075e-05,
+      "loss": 0.2569,
+      "step": 28371
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0471680687790708e-05,
+      "loss": 0.3313,
+      "step": 28372
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0471344852484342e-05,
+      "loss": 0.3575,
+      "step": 28373
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0471009017177977e-05,
+      "loss": 0.3473,
+      "step": 28374
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0470673181871612e-05,
+      "loss": 0.2556,
+      "step": 28375
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0470337346565245e-05,
+      "loss": 0.3818,
+      "step": 28376
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0470001511258881e-05,
+      "loss": 0.2225,
+      "step": 28377
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0469665675952514e-05,
+      "loss": 0.3669,
+      "step": 28378
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0469329840646147e-05,
+      "loss": 0.3817,
+      "step": 28379
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0468994005339783e-05,
+      "loss": 0.2091,
+      "step": 28380
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0468658170033416e-05,
+      "loss": 0.3047,
+      "step": 28381
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0468322334727051e-05,
+      "loss": 0.3982,
+      "step": 28382
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0467986499420686e-05,
+      "loss": 0.3388,
+      "step": 28383
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.046765066411432e-05,
+      "loss": 0.2846,
+      "step": 28384
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0467314828807953e-05,
+      "loss": 0.2017,
+      "step": 28385
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.046697899350159e-05,
+      "loss": 0.3393,
+      "step": 28386
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0466643158195222e-05,
+      "loss": 0.3234,
+      "step": 28387
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0466307322888855e-05,
+      "loss": 0.2798,
+      "step": 28388
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0465971487582492e-05,
+      "loss": 0.3429,
+      "step": 28389
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0465635652276125e-05,
+      "loss": 0.2766,
+      "step": 28390
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0465299816969758e-05,
+      "loss": 0.3597,
+      "step": 28391
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0464963981663394e-05,
+      "loss": 0.3232,
+      "step": 28392
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0464628146357027e-05,
+      "loss": 0.2005,
+      "step": 28393
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0464292311050662e-05,
+      "loss": 0.353,
+      "step": 28394
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0463956475744296e-05,
+      "loss": 0.3337,
+      "step": 28395
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.046362064043793e-05,
+      "loss": 0.3064,
+      "step": 28396
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0463284805131564e-05,
+      "loss": 0.3573,
+      "step": 28397
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.04629489698252e-05,
+      "loss": 0.259,
+      "step": 28398
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0462613134518833e-05,
+      "loss": 0.2657,
+      "step": 28399
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0462277299212466e-05,
+      "loss": 0.1883,
+      "step": 28400
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0461941463906102e-05,
+      "loss": 0.1892,
+      "step": 28401
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0461605628599735e-05,
+      "loss": 0.3169,
+      "step": 28402
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.046126979329337e-05,
+      "loss": 0.3276,
+      "step": 28403
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0460933957987005e-05,
+      "loss": 0.3455,
+      "step": 28404
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.046059812268064e-05,
+      "loss": 0.1861,
+      "step": 28405
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0460262287374272e-05,
+      "loss": 0.3991,
+      "step": 28406
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0459926452067909e-05,
+      "loss": 0.5376,
+      "step": 28407
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0459590616761542e-05,
+      "loss": 0.2899,
+      "step": 28408
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0459254781455174e-05,
+      "loss": 0.2684,
+      "step": 28409
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.045891894614881e-05,
+      "loss": 0.2944,
+      "step": 28410
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0458583110842444e-05,
+      "loss": 0.1903,
+      "step": 28411
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0458247275536077e-05,
+      "loss": 0.4335,
+      "step": 28412
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0457911440229713e-05,
+      "loss": 0.1674,
+      "step": 28413
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0457575604923346e-05,
+      "loss": 0.3591,
+      "step": 28414
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.045723976961698e-05,
+      "loss": 0.1854,
+      "step": 28415
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0456903934310615e-05,
+      "loss": 0.3182,
+      "step": 28416
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.045656809900425e-05,
+      "loss": 0.2206,
+      "step": 28417
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0456232263697883e-05,
+      "loss": 0.465,
+      "step": 28418
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.045589642839152e-05,
+      "loss": 0.3848,
+      "step": 28419
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0455560593085152e-05,
+      "loss": 0.4318,
+      "step": 28420
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0455224757778785e-05,
+      "loss": 0.2824,
+      "step": 28421
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0454888922472422e-05,
+      "loss": 0.2924,
+      "step": 28422
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0454553087166054e-05,
+      "loss": 0.3635,
+      "step": 28423
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0454217251859689e-05,
+      "loss": 0.4217,
+      "step": 28424
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0453881416553322e-05,
+      "loss": 0.2541,
+      "step": 28425
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0453545581246958e-05,
+      "loss": 0.2234,
+      "step": 28426
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0453209745940591e-05,
+      "loss": 0.1497,
+      "step": 28427
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0452873910634224e-05,
+      "loss": 0.1543,
+      "step": 28428
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.045253807532786e-05,
+      "loss": 0.239,
+      "step": 28429
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0452202240021494e-05,
+      "loss": 0.2272,
+      "step": 28430
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0451866404715128e-05,
+      "loss": 0.2706,
+      "step": 28431
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0451530569408763e-05,
+      "loss": 0.1823,
+      "step": 28432
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0451194734102398e-05,
+      "loss": 0.2259,
+      "step": 28433
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.045085889879603e-05,
+      "loss": 0.3559,
+      "step": 28434
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0450523063489665e-05,
+      "loss": 0.1557,
+      "step": 28435
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.04501872281833e-05,
+      "loss": 0.3652,
+      "step": 28436
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0449851392876933e-05,
+      "loss": 0.2465,
+      "step": 28437
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0449515557570569e-05,
+      "loss": 0.1199,
+      "step": 28438
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0449179722264202e-05,
+      "loss": 0.2309,
+      "step": 28439
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0448843886957835e-05,
+      "loss": 0.1876,
+      "step": 28440
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0448508051651471e-05,
+      "loss": 0.4795,
+      "step": 28441
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0448172216345104e-05,
+      "loss": 0.2882,
+      "step": 28442
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0447836381038739e-05,
+      "loss": 0.1721,
+      "step": 28443
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0447500545732374e-05,
+      "loss": 0.3033,
+      "step": 28444
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0447164710426008e-05,
+      "loss": 0.322,
+      "step": 28445
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0446828875119641e-05,
+      "loss": 0.2771,
+      "step": 28446
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0446493039813278e-05,
+      "loss": 0.3713,
+      "step": 28447
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.044615720450691e-05,
+      "loss": 0.2703,
+      "step": 28448
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0445821369200543e-05,
+      "loss": 0.4306,
+      "step": 28449
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.044548553389418e-05,
+      "loss": 0.2334,
+      "step": 28450
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0445149698587813e-05,
+      "loss": 0.2799,
+      "step": 28451
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0444813863281447e-05,
+      "loss": 0.4014,
+      "step": 28452
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0444478027975082e-05,
+      "loss": 0.3429,
+      "step": 28453
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0444142192668717e-05,
+      "loss": 0.3456,
+      "step": 28454
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.044380635736235e-05,
+      "loss": 0.1437,
+      "step": 28455
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0443470522055984e-05,
+      "loss": 0.2032,
+      "step": 28456
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0443134686749619e-05,
+      "loss": 0.2484,
+      "step": 28457
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0442798851443252e-05,
+      "loss": 0.3245,
+      "step": 28458
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0442463016136888e-05,
+      "loss": 0.2319,
+      "step": 28459
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0442127180830521e-05,
+      "loss": 0.2526,
+      "step": 28460
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0441791345524154e-05,
+      "loss": 0.2637,
+      "step": 28461
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.044145551021779e-05,
+      "loss": 0.2394,
+      "step": 28462
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0441119674911423e-05,
+      "loss": 0.2154,
+      "step": 28463
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0440783839605058e-05,
+      "loss": 0.2891,
+      "step": 28464
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0440448004298693e-05,
+      "loss": 0.291,
+      "step": 28465
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0440112168992327e-05,
+      "loss": 0.3292,
+      "step": 28466
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.043977633368596e-05,
+      "loss": 0.241,
+      "step": 28467
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0439440498379597e-05,
+      "loss": 0.196,
+      "step": 28468
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.043910466307323e-05,
+      "loss": 0.3501,
+      "step": 28469
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0438768827766863e-05,
+      "loss": 0.2429,
+      "step": 28470
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0438432992460499e-05,
+      "loss": 0.1278,
+      "step": 28471
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0438097157154132e-05,
+      "loss": 0.2989,
+      "step": 28472
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0437761321847766e-05,
+      "loss": 0.429,
+      "step": 28473
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0437425486541401e-05,
+      "loss": 0.1633,
+      "step": 28474
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0437089651235036e-05,
+      "loss": 0.4127,
+      "step": 28475
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0436753815928669e-05,
+      "loss": 0.2667,
+      "step": 28476
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0436417980622305e-05,
+      "loss": 0.1644,
+      "step": 28477
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0436082145315938e-05,
+      "loss": 0.3604,
+      "step": 28478
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0435746310009571e-05,
+      "loss": 0.3064,
+      "step": 28479
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0435410474703207e-05,
+      "loss": 0.2754,
+      "step": 28480
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.043507463939684e-05,
+      "loss": 0.2241,
+      "step": 28481
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0434738804090473e-05,
+      "loss": 0.3375,
+      "step": 28482
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.043440296878411e-05,
+      "loss": 0.26,
+      "step": 28483
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0434067133477743e-05,
+      "loss": 0.3689,
+      "step": 28484
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0433731298171377e-05,
+      "loss": 0.2135,
+      "step": 28485
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 1.0433395462865012e-05,
+      "loss": 0.3031,
+      "step": 28486
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0433059627558646e-05,
+      "loss": 0.2535,
+      "step": 28487
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.043272379225228e-05,
+      "loss": 0.5596,
+      "step": 28488
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0432387956945916e-05,
+      "loss": 0.2562,
+      "step": 28489
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0432052121639549e-05,
+      "loss": 0.3236,
+      "step": 28490
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0431716286333182e-05,
+      "loss": 0.5672,
+      "step": 28491
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0431380451026818e-05,
+      "loss": 0.3631,
+      "step": 28492
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0431044615720451e-05,
+      "loss": 0.2606,
+      "step": 28493
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0430708780414086e-05,
+      "loss": 0.3785,
+      "step": 28494
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.043037294510772e-05,
+      "loss": 0.3981,
+      "step": 28495
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0430037109801355e-05,
+      "loss": 0.2729,
+      "step": 28496
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0429701274494988e-05,
+      "loss": 0.3562,
+      "step": 28497
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0429365439188624e-05,
+      "loss": 0.3968,
+      "step": 28498
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0429029603882257e-05,
+      "loss": 0.2652,
+      "step": 28499
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.042869376857589e-05,
+      "loss": 0.3008,
+      "step": 28500
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0428357933269526e-05,
+      "loss": 0.403,
+      "step": 28501
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.042802209796316e-05,
+      "loss": 0.2311,
+      "step": 28502
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0427686262656792e-05,
+      "loss": 0.2669,
+      "step": 28503
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0427350427350429e-05,
+      "loss": 0.2689,
+      "step": 28504
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0427014592044062e-05,
+      "loss": 0.2687,
+      "step": 28505
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0426678756737696e-05,
+      "loss": 0.2281,
+      "step": 28506
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0426342921431331e-05,
+      "loss": 0.3775,
+      "step": 28507
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0426007086124966e-05,
+      "loss": 0.3941,
+      "step": 28508
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0425671250818599e-05,
+      "loss": 0.3359,
+      "step": 28509
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0425335415512235e-05,
+      "loss": 0.3098,
+      "step": 28510
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0424999580205868e-05,
+      "loss": 0.2828,
+      "step": 28511
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.04246637448995e-05,
+      "loss": 0.2982,
+      "step": 28512
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0424327909593137e-05,
+      "loss": 0.2284,
+      "step": 28513
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.042399207428677e-05,
+      "loss": 0.4519,
+      "step": 28514
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0423656238980405e-05,
+      "loss": 0.3451,
+      "step": 28515
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.042332040367404e-05,
+      "loss": 0.3887,
+      "step": 28516
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0422984568367674e-05,
+      "loss": 0.2206,
+      "step": 28517
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0422648733061307e-05,
+      "loss": 0.3592,
+      "step": 28518
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0422312897754943e-05,
+      "loss": 0.246,
+      "step": 28519
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0421977062448576e-05,
+      "loss": 0.3774,
+      "step": 28520
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.042164122714221e-05,
+      "loss": 0.2928,
+      "step": 28521
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0421305391835846e-05,
+      "loss": 0.3368,
+      "step": 28522
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0420969556529479e-05,
+      "loss": 0.148,
+      "step": 28523
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0420633721223113e-05,
+      "loss": 0.4829,
+      "step": 28524
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0420297885916748e-05,
+      "loss": 0.299,
+      "step": 28525
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.041996205061038e-05,
+      "loss": 0.378,
+      "step": 28526
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0419626215304015e-05,
+      "loss": 0.2635,
+      "step": 28527
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.041929037999765e-05,
+      "loss": 0.2997,
+      "step": 28528
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0418954544691285e-05,
+      "loss": 0.1777,
+      "step": 28529
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0418618709384918e-05,
+      "loss": 0.511,
+      "step": 28530
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0418282874078554e-05,
+      "loss": 0.1777,
+      "step": 28531
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0417947038772187e-05,
+      "loss": 0.3846,
+      "step": 28532
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.041761120346582e-05,
+      "loss": 0.3498,
+      "step": 28533
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0417275368159456e-05,
+      "loss": 0.1371,
+      "step": 28534
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.041693953285309e-05,
+      "loss": 0.3398,
+      "step": 28535
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0416603697546724e-05,
+      "loss": 0.3299,
+      "step": 28536
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0416267862240359e-05,
+      "loss": 0.2416,
+      "step": 28537
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0415932026933993e-05,
+      "loss": 0.3612,
+      "step": 28538
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0415596191627626e-05,
+      "loss": 0.2295,
+      "step": 28539
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0415260356321262e-05,
+      "loss": 0.319,
+      "step": 28540
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0414924521014895e-05,
+      "loss": 0.2842,
+      "step": 28541
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0414588685708528e-05,
+      "loss": 0.3031,
+      "step": 28542
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0414252850402165e-05,
+      "loss": 0.2661,
+      "step": 28543
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0413917015095798e-05,
+      "loss": 0.2185,
+      "step": 28544
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0413581179789432e-05,
+      "loss": 0.4919,
+      "step": 28545
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0413245344483067e-05,
+      "loss": 0.3576,
+      "step": 28546
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.04129095091767e-05,
+      "loss": 0.4028,
+      "step": 28547
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0412573673870335e-05,
+      "loss": 0.2683,
+      "step": 28548
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.041223783856397e-05,
+      "loss": 0.2679,
+      "step": 28549
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0411902003257604e-05,
+      "loss": 0.2976,
+      "step": 28550
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0411566167951237e-05,
+      "loss": 0.2871,
+      "step": 28551
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0411230332644873e-05,
+      "loss": 0.1826,
+      "step": 28552
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0410894497338506e-05,
+      "loss": 0.455,
+      "step": 28553
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0410558662032139e-05,
+      "loss": 0.2042,
+      "step": 28554
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0410222826725775e-05,
+      "loss": 0.3473,
+      "step": 28555
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0409886991419408e-05,
+      "loss": 0.3425,
+      "step": 28556
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0409551156113043e-05,
+      "loss": 0.3113,
+      "step": 28557
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0409215320806678e-05,
+      "loss": 0.3479,
+      "step": 28558
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0408879485500312e-05,
+      "loss": 0.3083,
+      "step": 28559
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0408543650193945e-05,
+      "loss": 0.2739,
+      "step": 28560
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0408207814887582e-05,
+      "loss": 0.3662,
+      "step": 28561
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0407871979581215e-05,
+      "loss": 0.2764,
+      "step": 28562
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0407536144274847e-05,
+      "loss": 0.1625,
+      "step": 28563
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0407200308968484e-05,
+      "loss": 0.2271,
+      "step": 28564
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0406864473662117e-05,
+      "loss": 0.2713,
+      "step": 28565
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0406528638355751e-05,
+      "loss": 0.1543,
+      "step": 28566
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0406192803049386e-05,
+      "loss": 0.2432,
+      "step": 28567
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.040585696774302e-05,
+      "loss": 0.2531,
+      "step": 28568
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0405521132436654e-05,
+      "loss": 0.3692,
+      "step": 28569
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0405185297130288e-05,
+      "loss": 0.1871,
+      "step": 28570
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0404849461823923e-05,
+      "loss": 0.2732,
+      "step": 28571
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0404513626517556e-05,
+      "loss": 0.3048,
+      "step": 28572
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0404177791211192e-05,
+      "loss": 0.2504,
+      "step": 28573
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0403841955904825e-05,
+      "loss": 0.149,
+      "step": 28574
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0403506120598458e-05,
+      "loss": 0.225,
+      "step": 28575
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0403170285292095e-05,
+      "loss": 0.2593,
+      "step": 28576
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0402834449985727e-05,
+      "loss": 0.2527,
+      "step": 28577
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0402498614679362e-05,
+      "loss": 0.2474,
+      "step": 28578
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0402162779372997e-05,
+      "loss": 0.4419,
+      "step": 28579
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0401826944066631e-05,
+      "loss": 0.1177,
+      "step": 28580
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0401491108760264e-05,
+      "loss": 0.2163,
+      "step": 28581
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.04011552734539e-05,
+      "loss": 0.3642,
+      "step": 28582
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0400819438147534e-05,
+      "loss": 0.4017,
+      "step": 28583
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0400483602841167e-05,
+      "loss": 0.3651,
+      "step": 28584
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0400147767534803e-05,
+      "loss": 0.3499,
+      "step": 28585
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0399811932228436e-05,
+      "loss": 0.3497,
+      "step": 28586
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.039947609692207e-05,
+      "loss": 0.2827,
+      "step": 28587
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0399140261615705e-05,
+      "loss": 0.5211,
+      "step": 28588
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.039880442630934e-05,
+      "loss": 0.2612,
+      "step": 28589
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0398468591002973e-05,
+      "loss": 0.3043,
+      "step": 28590
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0398132755696607e-05,
+      "loss": 0.2966,
+      "step": 28591
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0397796920390242e-05,
+      "loss": 0.3201,
+      "step": 28592
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0397461085083875e-05,
+      "loss": 0.1375,
+      "step": 28593
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0397125249777511e-05,
+      "loss": 0.2242,
+      "step": 28594
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0396789414471144e-05,
+      "loss": 0.2842,
+      "step": 28595
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0396453579164777e-05,
+      "loss": 0.257,
+      "step": 28596
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0396117743858414e-05,
+      "loss": 0.3167,
+      "step": 28597
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0395781908552047e-05,
+      "loss": 0.2464,
+      "step": 28598
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0395446073245681e-05,
+      "loss": 0.2876,
+      "step": 28599
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0395110237939316e-05,
+      "loss": 0.4554,
+      "step": 28600
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.039477440263295e-05,
+      "loss": 0.1367,
+      "step": 28601
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0394438567326583e-05,
+      "loss": 0.3407,
+      "step": 28602
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.039410273202022e-05,
+      "loss": 0.2658,
+      "step": 28603
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0393766896713853e-05,
+      "loss": 0.2875,
+      "step": 28604
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0393431061407486e-05,
+      "loss": 0.2722,
+      "step": 28605
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.039309522610112e-05,
+      "loss": 0.303,
+      "step": 28606
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0392759390794755e-05,
+      "loss": 0.3478,
+      "step": 28607
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.039242355548839e-05,
+      "loss": 0.2195,
+      "step": 28608
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0392087720182023e-05,
+      "loss": 0.2798,
+      "step": 28609
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0391751884875659e-05,
+      "loss": 0.3173,
+      "step": 28610
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0391416049569292e-05,
+      "loss": 0.2538,
+      "step": 28611
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0391080214262925e-05,
+      "loss": 0.3132,
+      "step": 28612
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0390744378956561e-05,
+      "loss": 0.2416,
+      "step": 28613
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0390408543650194e-05,
+      "loss": 0.234,
+      "step": 28614
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0390072708343827e-05,
+      "loss": 0.3651,
+      "step": 28615
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0389736873037463e-05,
+      "loss": 0.277,
+      "step": 28616
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0389401037731096e-05,
+      "loss": 0.4204,
+      "step": 28617
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0389065202424731e-05,
+      "loss": 0.2127,
+      "step": 28618
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0388729367118366e-05,
+      "loss": 0.4113,
+      "step": 28619
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0388393531812e-05,
+      "loss": 0.2664,
+      "step": 28620
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0388057696505633e-05,
+      "loss": 0.163,
+      "step": 28621
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.038772186119927e-05,
+      "loss": 0.1972,
+      "step": 28622
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0387386025892903e-05,
+      "loss": 0.1324,
+      "step": 28623
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0387050190586536e-05,
+      "loss": 0.3341,
+      "step": 28624
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0386714355280172e-05,
+      "loss": 0.6577,
+      "step": 28625
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0386378519973805e-05,
+      "loss": 0.3455,
+      "step": 28626
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.038604268466744e-05,
+      "loss": 0.2833,
+      "step": 28627
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0385706849361074e-05,
+      "loss": 0.4213,
+      "step": 28628
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0385371014054709e-05,
+      "loss": 0.2804,
+      "step": 28629
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0385035178748342e-05,
+      "loss": 0.3701,
+      "step": 28630
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0384699343441978e-05,
+      "loss": 0.3138,
+      "step": 28631
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0384363508135611e-05,
+      "loss": 0.3898,
+      "step": 28632
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0384027672829244e-05,
+      "loss": 0.3349,
+      "step": 28633
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.038369183752288e-05,
+      "loss": 0.2619,
+      "step": 28634
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0383356002216513e-05,
+      "loss": 0.3898,
+      "step": 28635
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0383020166910148e-05,
+      "loss": 0.3676,
+      "step": 28636
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0382684331603783e-05,
+      "loss": 0.1779,
+      "step": 28637
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0382348496297416e-05,
+      "loss": 0.3951,
+      "step": 28638
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.038201266099105e-05,
+      "loss": 0.1906,
+      "step": 28639
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0381676825684685e-05,
+      "loss": 0.3495,
+      "step": 28640
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.038134099037832e-05,
+      "loss": 0.25,
+      "step": 28641
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0381005155071952e-05,
+      "loss": 0.2563,
+      "step": 28642
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0380669319765589e-05,
+      "loss": 0.2812,
+      "step": 28643
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0380333484459222e-05,
+      "loss": 0.5112,
+      "step": 28644
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0379997649152855e-05,
+      "loss": 0.2114,
+      "step": 28645
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0379661813846491e-05,
+      "loss": 0.3942,
+      "step": 28646
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0379325978540124e-05,
+      "loss": 0.2559,
+      "step": 28647
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0378990143233759e-05,
+      "loss": 0.3029,
+      "step": 28648
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0378654307927393e-05,
+      "loss": 0.3264,
+      "step": 28649
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0378318472621028e-05,
+      "loss": 0.4032,
+      "step": 28650
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0377982637314661e-05,
+      "loss": 0.2522,
+      "step": 28651
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0377646802008297e-05,
+      "loss": 0.4318,
+      "step": 28652
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.037731096670193e-05,
+      "loss": 0.3621,
+      "step": 28653
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0376975131395563e-05,
+      "loss": 0.279,
+      "step": 28654
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.03766392960892e-05,
+      "loss": 0.231,
+      "step": 28655
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0376303460782832e-05,
+      "loss": 0.4079,
+      "step": 28656
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0375967625476467e-05,
+      "loss": 0.234,
+      "step": 28657
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0375631790170102e-05,
+      "loss": 0.2993,
+      "step": 28658
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0375295954863736e-05,
+      "loss": 0.1805,
+      "step": 28659
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.037496011955737e-05,
+      "loss": 0.3066,
+      "step": 28660
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0374624284251004e-05,
+      "loss": 0.3433,
+      "step": 28661
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0374288448944639e-05,
+      "loss": 0.2868,
+      "step": 28662
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0373952613638272e-05,
+      "loss": 0.2538,
+      "step": 28663
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0373616778331908e-05,
+      "loss": 0.2317,
+      "step": 28664
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0373280943025541e-05,
+      "loss": 0.2793,
+      "step": 28665
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0372945107719174e-05,
+      "loss": 0.2057,
+      "step": 28666
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.037260927241281e-05,
+      "loss": 0.2692,
+      "step": 28667
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0372273437106443e-05,
+      "loss": 0.5282,
+      "step": 28668
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0371937601800078e-05,
+      "loss": 0.4415,
+      "step": 28669
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0371601766493712e-05,
+      "loss": 0.4768,
+      "step": 28670
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0371265931187347e-05,
+      "loss": 0.2303,
+      "step": 28671
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.037093009588098e-05,
+      "loss": 0.2434,
+      "step": 28672
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0370594260574616e-05,
+      "loss": 0.3452,
+      "step": 28673
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.037025842526825e-05,
+      "loss": 0.2453,
+      "step": 28674
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0369922589961882e-05,
+      "loss": 0.2524,
+      "step": 28675
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0369586754655519e-05,
+      "loss": 0.184,
+      "step": 28676
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0369250919349152e-05,
+      "loss": 0.4099,
+      "step": 28677
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0368915084042786e-05,
+      "loss": 0.3197,
+      "step": 28678
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0368579248736421e-05,
+      "loss": 0.3419,
+      "step": 28679
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0368243413430055e-05,
+      "loss": 0.3081,
+      "step": 28680
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0367907578123688e-05,
+      "loss": 0.2168,
+      "step": 28681
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0367571742817323e-05,
+      "loss": 0.244,
+      "step": 28682
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.0367235907510958e-05,
+      "loss": 0.2845,
+      "step": 28683
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.036690007220459e-05,
+      "loss": 0.4369,
+      "step": 28684
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0366564236898227e-05,
+      "loss": 0.3519,
+      "step": 28685
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.036622840159186e-05,
+      "loss": 0.1153,
+      "step": 28686
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0365892566285493e-05,
+      "loss": 0.2465,
+      "step": 28687
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.036555673097913e-05,
+      "loss": 0.2777,
+      "step": 28688
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0365220895672762e-05,
+      "loss": 0.3787,
+      "step": 28689
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0364885060366397e-05,
+      "loss": 0.2441,
+      "step": 28690
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0364549225060032e-05,
+      "loss": 0.2013,
+      "step": 28691
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0364213389753666e-05,
+      "loss": 0.2462,
+      "step": 28692
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0363877554447299e-05,
+      "loss": 0.3112,
+      "step": 28693
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0363541719140935e-05,
+      "loss": 0.3274,
+      "step": 28694
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0363205883834568e-05,
+      "loss": 0.2848,
+      "step": 28695
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0362870048528201e-05,
+      "loss": 0.2521,
+      "step": 28696
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0362534213221838e-05,
+      "loss": 0.2882,
+      "step": 28697
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.036219837791547e-05,
+      "loss": 0.2727,
+      "step": 28698
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0361862542609105e-05,
+      "loss": 0.292,
+      "step": 28699
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.036152670730274e-05,
+      "loss": 0.2848,
+      "step": 28700
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0361190871996375e-05,
+      "loss": 0.5079,
+      "step": 28701
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0360855036690008e-05,
+      "loss": 0.2737,
+      "step": 28702
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0360519201383644e-05,
+      "loss": 0.2875,
+      "step": 28703
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0360183366077277e-05,
+      "loss": 0.3639,
+      "step": 28704
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.035984753077091e-05,
+      "loss": 0.4306,
+      "step": 28705
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0359511695464546e-05,
+      "loss": 0.1943,
+      "step": 28706
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0359175860158179e-05,
+      "loss": 0.3625,
+      "step": 28707
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0358840024851812e-05,
+      "loss": 0.2923,
+      "step": 28708
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0358504189545448e-05,
+      "loss": 0.1943,
+      "step": 28709
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0358168354239081e-05,
+      "loss": 0.2423,
+      "step": 28710
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0357832518932716e-05,
+      "loss": 0.2322,
+      "step": 28711
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.035749668362635e-05,
+      "loss": 0.2432,
+      "step": 28712
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0357160848319985e-05,
+      "loss": 0.375,
+      "step": 28713
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0356825013013618e-05,
+      "loss": 0.2494,
+      "step": 28714
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0356489177707255e-05,
+      "loss": 0.2193,
+      "step": 28715
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0356153342400888e-05,
+      "loss": 0.1954,
+      "step": 28716
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.035581750709452e-05,
+      "loss": 0.1924,
+      "step": 28717
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0355481671788157e-05,
+      "loss": 0.4253,
+      "step": 28718
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.035514583648179e-05,
+      "loss": 0.1459,
+      "step": 28719
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0354810001175424e-05,
+      "loss": 0.3477,
+      "step": 28720
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0354474165869059e-05,
+      "loss": 0.3523,
+      "step": 28721
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0354138330562694e-05,
+      "loss": 0.2711,
+      "step": 28722
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0353802495256327e-05,
+      "loss": 0.2554,
+      "step": 28723
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0353466659949963e-05,
+      "loss": 0.3797,
+      "step": 28724
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0353130824643596e-05,
+      "loss": 0.3644,
+      "step": 28725
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0352794989337229e-05,
+      "loss": 0.2801,
+      "step": 28726
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0352459154030865e-05,
+      "loss": 0.2641,
+      "step": 28727
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0352123318724498e-05,
+      "loss": 0.3258,
+      "step": 28728
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0351787483418131e-05,
+      "loss": 0.3399,
+      "step": 28729
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0351451648111768e-05,
+      "loss": 0.3173,
+      "step": 28730
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.03511158128054e-05,
+      "loss": 0.268,
+      "step": 28731
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0350779977499035e-05,
+      "loss": 0.5972,
+      "step": 28732
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.035044414219267e-05,
+      "loss": 0.2765,
+      "step": 28733
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0350108306886304e-05,
+      "loss": 0.2887,
+      "step": 28734
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0349772471579937e-05,
+      "loss": 0.3132,
+      "step": 28735
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0349436636273574e-05,
+      "loss": 0.4117,
+      "step": 28736
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0349100800967207e-05,
+      "loss": 0.3442,
+      "step": 28737
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.034876496566084e-05,
+      "loss": 0.3574,
+      "step": 28738
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0348429130354476e-05,
+      "loss": 0.3834,
+      "step": 28739
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0348093295048109e-05,
+      "loss": 0.3275,
+      "step": 28740
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0347757459741744e-05,
+      "loss": 0.2263,
+      "step": 28741
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0347421624435378e-05,
+      "loss": 0.2851,
+      "step": 28742
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0347085789129013e-05,
+      "loss": 0.1992,
+      "step": 28743
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0346749953822646e-05,
+      "loss": 0.2878,
+      "step": 28744
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0346414118516282e-05,
+      "loss": 0.2933,
+      "step": 28745
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0346078283209915e-05,
+      "loss": 0.2874,
+      "step": 28746
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0345742447903548e-05,
+      "loss": 0.4805,
+      "step": 28747
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0345406612597184e-05,
+      "loss": 0.2425,
+      "step": 28748
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0345070777290817e-05,
+      "loss": 0.406,
+      "step": 28749
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.034473494198445e-05,
+      "loss": 0.3152,
+      "step": 28750
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0344399106678087e-05,
+      "loss": 0.1895,
+      "step": 28751
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.034406327137172e-05,
+      "loss": 0.2723,
+      "step": 28752
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0343727436065354e-05,
+      "loss": 0.2498,
+      "step": 28753
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0343391600758989e-05,
+      "loss": 0.3595,
+      "step": 28754
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0343055765452624e-05,
+      "loss": 0.3279,
+      "step": 28755
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0342719930146257e-05,
+      "loss": 0.1935,
+      "step": 28756
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0342384094839893e-05,
+      "loss": 0.2924,
+      "step": 28757
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0342048259533526e-05,
+      "loss": 0.2872,
+      "step": 28758
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0341712424227159e-05,
+      "loss": 0.143,
+      "step": 28759
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0341376588920795e-05,
+      "loss": 0.3143,
+      "step": 28760
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0341040753614428e-05,
+      "loss": 0.3326,
+      "step": 28761
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0340704918308063e-05,
+      "loss": 0.2293,
+      "step": 28762
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0340369083001697e-05,
+      "loss": 0.2096,
+      "step": 28763
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0340033247695332e-05,
+      "loss": 0.1694,
+      "step": 28764
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0339697412388965e-05,
+      "loss": 0.2491,
+      "step": 28765
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0339361577082601e-05,
+      "loss": 0.3948,
+      "step": 28766
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0339025741776234e-05,
+      "loss": 0.3107,
+      "step": 28767
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0338689906469867e-05,
+      "loss": 0.3146,
+      "step": 28768
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0338354071163504e-05,
+      "loss": 0.4399,
+      "step": 28769
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0338018235857136e-05,
+      "loss": 0.2264,
+      "step": 28770
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0337682400550771e-05,
+      "loss": 0.1731,
+      "step": 28771
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0337346565244406e-05,
+      "loss": 0.1696,
+      "step": 28772
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0337010729938039e-05,
+      "loss": 0.3311,
+      "step": 28773
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0336674894631673e-05,
+      "loss": 0.3282,
+      "step": 28774
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0336339059325308e-05,
+      "loss": 0.2555,
+      "step": 28775
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0336003224018943e-05,
+      "loss": 0.3803,
+      "step": 28776
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0335667388712576e-05,
+      "loss": 0.3538,
+      "step": 28777
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0335331553406212e-05,
+      "loss": 0.443,
+      "step": 28778
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0334995718099845e-05,
+      "loss": 0.2397,
+      "step": 28779
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0334659882793478e-05,
+      "loss": 0.2565,
+      "step": 28780
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0334324047487114e-05,
+      "loss": 0.4255,
+      "step": 28781
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0333988212180747e-05,
+      "loss": 0.4321,
+      "step": 28782
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0333652376874382e-05,
+      "loss": 0.3111,
+      "step": 28783
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0333316541568015e-05,
+      "loss": 0.316,
+      "step": 28784
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0332980706261651e-05,
+      "loss": 0.388,
+      "step": 28785
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0332644870955284e-05,
+      "loss": 0.2531,
+      "step": 28786
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0332309035648917e-05,
+      "loss": 0.3028,
+      "step": 28787
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0331973200342553e-05,
+      "loss": 0.1875,
+      "step": 28788
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0331637365036186e-05,
+      "loss": 0.3356,
+      "step": 28789
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0331301529729821e-05,
+      "loss": 0.1687,
+      "step": 28790
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0330965694423456e-05,
+      "loss": 0.2542,
+      "step": 28791
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.033062985911709e-05,
+      "loss": 0.3945,
+      "step": 28792
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0330294023810723e-05,
+      "loss": 0.2944,
+      "step": 28793
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.032995818850436e-05,
+      "loss": 0.3409,
+      "step": 28794
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0329622353197993e-05,
+      "loss": 0.3309,
+      "step": 28795
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0329286517891625e-05,
+      "loss": 0.5421,
+      "step": 28796
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0328950682585262e-05,
+      "loss": 0.3396,
+      "step": 28797
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0328614847278895e-05,
+      "loss": 0.1612,
+      "step": 28798
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0328279011972528e-05,
+      "loss": 0.2097,
+      "step": 28799
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0327943176666164e-05,
+      "loss": 0.2604,
+      "step": 28800
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0327607341359797e-05,
+      "loss": 0.4634,
+      "step": 28801
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0327271506053432e-05,
+      "loss": 0.1428,
+      "step": 28802
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0326935670747066e-05,
+      "loss": 0.2251,
+      "step": 28803
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0326599835440701e-05,
+      "loss": 0.2173,
+      "step": 28804
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0326264000134334e-05,
+      "loss": 0.3312,
+      "step": 28805
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.032592816482797e-05,
+      "loss": 0.2845,
+      "step": 28806
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0325592329521603e-05,
+      "loss": 0.5003,
+      "step": 28807
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0325256494215236e-05,
+      "loss": 0.2893,
+      "step": 28808
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0324920658908872e-05,
+      "loss": 0.3354,
+      "step": 28809
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0324584823602505e-05,
+      "loss": 0.2354,
+      "step": 28810
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.032424898829614e-05,
+      "loss": 0.2991,
+      "step": 28811
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0323913152989775e-05,
+      "loss": 0.339,
+      "step": 28812
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.032357731768341e-05,
+      "loss": 0.3404,
+      "step": 28813
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0323241482377042e-05,
+      "loss": 0.2176,
+      "step": 28814
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0322905647070679e-05,
+      "loss": 0.2472,
+      "step": 28815
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0322569811764312e-05,
+      "loss": 0.3866,
+      "step": 28816
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0322233976457945e-05,
+      "loss": 0.2324,
+      "step": 28817
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0321898141151581e-05,
+      "loss": 0.2395,
+      "step": 28818
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0321562305845214e-05,
+      "loss": 0.3016,
+      "step": 28819
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0321226470538847e-05,
+      "loss": 0.3414,
+      "step": 28820
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0320890635232483e-05,
+      "loss": 0.3647,
+      "step": 28821
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0320554799926116e-05,
+      "loss": 0.316,
+      "step": 28822
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.032021896461975e-05,
+      "loss": 0.3056,
+      "step": 28823
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0319883129313385e-05,
+      "loss": 0.3273,
+      "step": 28824
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.031954729400702e-05,
+      "loss": 0.4379,
+      "step": 28825
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0319211458700653e-05,
+      "loss": 0.289,
+      "step": 28826
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.031887562339429e-05,
+      "loss": 0.3735,
+      "step": 28827
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0318539788087922e-05,
+      "loss": 0.278,
+      "step": 28828
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0318203952781555e-05,
+      "loss": 0.2496,
+      "step": 28829
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0317868117475192e-05,
+      "loss": 0.4728,
+      "step": 28830
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0317532282168825e-05,
+      "loss": 0.3615,
+      "step": 28831
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.031719644686246e-05,
+      "loss": 0.269,
+      "step": 28832
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0316860611556094e-05,
+      "loss": 0.4743,
+      "step": 28833
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0316524776249729e-05,
+      "loss": 0.1766,
+      "step": 28834
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0316188940943361e-05,
+      "loss": 0.3862,
+      "step": 28835
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0315853105636998e-05,
+      "loss": 0.3178,
+      "step": 28836
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.031551727033063e-05,
+      "loss": 0.3997,
+      "step": 28837
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0315181435024264e-05,
+      "loss": 0.2722,
+      "step": 28838
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.03148455997179e-05,
+      "loss": 0.2636,
+      "step": 28839
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0314509764411533e-05,
+      "loss": 0.3096,
+      "step": 28840
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0314173929105166e-05,
+      "loss": 0.2343,
+      "step": 28841
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0313838093798802e-05,
+      "loss": 0.1894,
+      "step": 28842
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0313502258492435e-05,
+      "loss": 0.2229,
+      "step": 28843
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.031316642318607e-05,
+      "loss": 0.1653,
+      "step": 28844
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0312830587879705e-05,
+      "loss": 0.1531,
+      "step": 28845
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.031249475257334e-05,
+      "loss": 0.2083,
+      "step": 28846
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0312158917266972e-05,
+      "loss": 0.2428,
+      "step": 28847
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0311823081960608e-05,
+      "loss": 0.2236,
+      "step": 28848
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0311487246654241e-05,
+      "loss": 0.1817,
+      "step": 28849
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0311151411347874e-05,
+      "loss": 0.263,
+      "step": 28850
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.031081557604151e-05,
+      "loss": 0.1456,
+      "step": 28851
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0310479740735144e-05,
+      "loss": 0.2859,
+      "step": 28852
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0310143905428778e-05,
+      "loss": 0.2619,
+      "step": 28853
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0309808070122413e-05,
+      "loss": 0.2785,
+      "step": 28854
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0309472234816048e-05,
+      "loss": 0.4723,
+      "step": 28855
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.030913639950968e-05,
+      "loss": 0.2295,
+      "step": 28856
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0308800564203317e-05,
+      "loss": 0.2813,
+      "step": 28857
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.030846472889695e-05,
+      "loss": 0.1981,
+      "step": 28858
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0308128893590583e-05,
+      "loss": 0.1884,
+      "step": 28859
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.030779305828422e-05,
+      "loss": 0.3314,
+      "step": 28860
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0307457222977852e-05,
+      "loss": 0.6013,
+      "step": 28861
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0307121387671487e-05,
+      "loss": 0.3016,
+      "step": 28862
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0306785552365121e-05,
+      "loss": 0.3689,
+      "step": 28863
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0306449717058754e-05,
+      "loss": 0.1721,
+      "step": 28864
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0306113881752389e-05,
+      "loss": 0.3145,
+      "step": 28865
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0305778046446024e-05,
+      "loss": 0.4538,
+      "step": 28866
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0305442211139658e-05,
+      "loss": 0.3043,
+      "step": 28867
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0305106375833291e-05,
+      "loss": 0.1897,
+      "step": 28868
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0304770540526928e-05,
+      "loss": 0.4952,
+      "step": 28869
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.030443470522056e-05,
+      "loss": 0.3732,
+      "step": 28870
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0304098869914194e-05,
+      "loss": 0.321,
+      "step": 28871
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.030376303460783e-05,
+      "loss": 0.3431,
+      "step": 28872
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0303427199301463e-05,
+      "loss": 0.215,
+      "step": 28873
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0303091363995097e-05,
+      "loss": 0.3854,
+      "step": 28874
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0302755528688732e-05,
+      "loss": 0.3021,
+      "step": 28875
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0302419693382367e-05,
+      "loss": 0.1797,
+      "step": 28876
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0302083858076e-05,
+      "loss": 0.2683,
+      "step": 28877
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0301748022769636e-05,
+      "loss": 0.5429,
+      "step": 28878
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0301412187463269e-05,
+      "loss": 0.278,
+      "step": 28879
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0301076352156902e-05,
+      "loss": 0.3825,
+      "step": 28880
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0300740516850538e-05,
+      "loss": 0.1567,
+      "step": 28881
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0300404681544171e-05,
+      "loss": 0.3675,
+      "step": 28882
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 1.0300068846237806e-05,
+      "loss": 0.2001,
+      "step": 28883
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.029973301093144e-05,
+      "loss": 0.2877,
+      "step": 28884
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0299397175625074e-05,
+      "loss": 0.194,
+      "step": 28885
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0299061340318708e-05,
+      "loss": 0.2677,
+      "step": 28886
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0298725505012343e-05,
+      "loss": 0.3276,
+      "step": 28887
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0298389669705977e-05,
+      "loss": 0.3034,
+      "step": 28888
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.029805383439961e-05,
+      "loss": 0.2513,
+      "step": 28889
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0297717999093247e-05,
+      "loss": 0.2995,
+      "step": 28890
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.029738216378688e-05,
+      "loss": 0.2312,
+      "step": 28891
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0297046328480513e-05,
+      "loss": 0.2941,
+      "step": 28892
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0296710493174149e-05,
+      "loss": 0.4144,
+      "step": 28893
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0296374657867782e-05,
+      "loss": 0.1552,
+      "step": 28894
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0296038822561417e-05,
+      "loss": 0.2402,
+      "step": 28895
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0295702987255051e-05,
+      "loss": 0.1965,
+      "step": 28896
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0295367151948686e-05,
+      "loss": 0.1747,
+      "step": 28897
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0295031316642319e-05,
+      "loss": 0.2672,
+      "step": 28898
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0294695481335955e-05,
+      "loss": 0.3687,
+      "step": 28899
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0294359646029588e-05,
+      "loss": 0.2348,
+      "step": 28900
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0294023810723221e-05,
+      "loss": 0.3344,
+      "step": 28901
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0293687975416857e-05,
+      "loss": 0.2236,
+      "step": 28902
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.029335214011049e-05,
+      "loss": 0.1997,
+      "step": 28903
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0293016304804125e-05,
+      "loss": 0.208,
+      "step": 28904
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.029268046949776e-05,
+      "loss": 0.2388,
+      "step": 28905
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0292344634191394e-05,
+      "loss": 0.1823,
+      "step": 28906
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0292008798885027e-05,
+      "loss": 0.1916,
+      "step": 28907
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0291672963578662e-05,
+      "loss": 0.3675,
+      "step": 28908
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0291337128272297e-05,
+      "loss": 0.155,
+      "step": 28909
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.029100129296593e-05,
+      "loss": 0.1815,
+      "step": 28910
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0290665457659566e-05,
+      "loss": 0.1096,
+      "step": 28911
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0290329622353199e-05,
+      "loss": 0.6377,
+      "step": 28912
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0289993787046832e-05,
+      "loss": 0.3114,
+      "step": 28913
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0289657951740468e-05,
+      "loss": 0.4092,
+      "step": 28914
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0289322116434101e-05,
+      "loss": 0.1829,
+      "step": 28915
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0288986281127736e-05,
+      "loss": 0.4308,
+      "step": 28916
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.028865044582137e-05,
+      "loss": 0.2667,
+      "step": 28917
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0288314610515005e-05,
+      "loss": 0.3218,
+      "step": 28918
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0287978775208638e-05,
+      "loss": 0.2125,
+      "step": 28919
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0287642939902274e-05,
+      "loss": 0.2948,
+      "step": 28920
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0287307104595907e-05,
+      "loss": 0.3878,
+      "step": 28921
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.028697126928954e-05,
+      "loss": 0.2263,
+      "step": 28922
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0286635433983177e-05,
+      "loss": 0.1887,
+      "step": 28923
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.028629959867681e-05,
+      "loss": 0.2942,
+      "step": 28924
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0285963763370444e-05,
+      "loss": 0.1823,
+      "step": 28925
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0285627928064079e-05,
+      "loss": 0.3241,
+      "step": 28926
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0285292092757713e-05,
+      "loss": 0.3192,
+      "step": 28927
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0284956257451346e-05,
+      "loss": 0.2349,
+      "step": 28928
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0284620422144983e-05,
+      "loss": 0.3015,
+      "step": 28929
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0284284586838616e-05,
+      "loss": 0.2328,
+      "step": 28930
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0283948751532249e-05,
+      "loss": 0.2477,
+      "step": 28931
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0283612916225885e-05,
+      "loss": 0.1796,
+      "step": 28932
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0283277080919518e-05,
+      "loss": 0.4307,
+      "step": 28933
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0282941245613151e-05,
+      "loss": 0.2707,
+      "step": 28934
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0282605410306787e-05,
+      "loss": 0.2921,
+      "step": 28935
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.028226957500042e-05,
+      "loss": 0.287,
+      "step": 28936
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0281933739694055e-05,
+      "loss": 0.279,
+      "step": 28937
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.028159790438769e-05,
+      "loss": 0.1474,
+      "step": 28938
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0281262069081324e-05,
+      "loss": 0.3757,
+      "step": 28939
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0280926233774957e-05,
+      "loss": 0.3076,
+      "step": 28940
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0280590398468593e-05,
+      "loss": 0.1469,
+      "step": 28941
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0280254563162226e-05,
+      "loss": 0.2751,
+      "step": 28942
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.027991872785586e-05,
+      "loss": 0.3789,
+      "step": 28943
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0279582892549496e-05,
+      "loss": 0.3023,
+      "step": 28944
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0279247057243129e-05,
+      "loss": 0.2362,
+      "step": 28945
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0278911221936763e-05,
+      "loss": 0.3464,
+      "step": 28946
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0278575386630398e-05,
+      "loss": 0.2669,
+      "step": 28947
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0278239551324033e-05,
+      "loss": 0.4198,
+      "step": 28948
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0277903716017666e-05,
+      "loss": 0.1839,
+      "step": 28949
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0277567880711302e-05,
+      "loss": 0.2186,
+      "step": 28950
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0277232045404935e-05,
+      "loss": 0.4269,
+      "step": 28951
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0276896210098568e-05,
+      "loss": 0.273,
+      "step": 28952
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0276560374792204e-05,
+      "loss": 0.2303,
+      "step": 28953
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0276224539485837e-05,
+      "loss": 0.2933,
+      "step": 28954
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.027588870417947e-05,
+      "loss": 0.2706,
+      "step": 28955
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0275552868873106e-05,
+      "loss": 0.3294,
+      "step": 28956
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.027521703356674e-05,
+      "loss": 0.3298,
+      "step": 28957
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0274881198260374e-05,
+      "loss": 0.3293,
+      "step": 28958
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0274545362954009e-05,
+      "loss": 0.1922,
+      "step": 28959
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0274209527647643e-05,
+      "loss": 0.2808,
+      "step": 28960
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0273873692341276e-05,
+      "loss": 0.3666,
+      "step": 28961
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0273537857034913e-05,
+      "loss": 0.3622,
+      "step": 28962
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0273202021728546e-05,
+      "loss": 0.4356,
+      "step": 28963
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0272866186422178e-05,
+      "loss": 0.2337,
+      "step": 28964
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0272530351115813e-05,
+      "loss": 0.2598,
+      "step": 28965
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0272194515809448e-05,
+      "loss": 0.1985,
+      "step": 28966
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0271858680503082e-05,
+      "loss": 0.1852,
+      "step": 28967
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0271522845196715e-05,
+      "loss": 0.3096,
+      "step": 28968
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0271187009890352e-05,
+      "loss": 0.2962,
+      "step": 28969
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0270851174583985e-05,
+      "loss": 0.3175,
+      "step": 28970
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0270515339277618e-05,
+      "loss": 0.1491,
+      "step": 28971
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0270179503971254e-05,
+      "loss": 0.3496,
+      "step": 28972
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0269843668664887e-05,
+      "loss": 0.4212,
+      "step": 28973
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0269507833358522e-05,
+      "loss": 0.2714,
+      "step": 28974
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0269171998052156e-05,
+      "loss": 0.2688,
+      "step": 28975
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0268836162745789e-05,
+      "loss": 0.1342,
+      "step": 28976
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0268500327439424e-05,
+      "loss": 0.2607,
+      "step": 28977
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0268164492133058e-05,
+      "loss": 0.5188,
+      "step": 28978
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0267828656826693e-05,
+      "loss": 0.3304,
+      "step": 28979
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0267492821520326e-05,
+      "loss": 0.2644,
+      "step": 28980
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0267156986213962e-05,
+      "loss": 0.2968,
+      "step": 28981
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0266821150907595e-05,
+      "loss": 0.2446,
+      "step": 28982
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0266485315601228e-05,
+      "loss": 0.1689,
+      "step": 28983
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0266149480294865e-05,
+      "loss": 0.3166,
+      "step": 28984
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0265813644988498e-05,
+      "loss": 0.2698,
+      "step": 28985
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0265477809682132e-05,
+      "loss": 0.146,
+      "step": 28986
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0265141974375767e-05,
+      "loss": 0.208,
+      "step": 28987
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0264806139069402e-05,
+      "loss": 0.2119,
+      "step": 28988
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0264470303763034e-05,
+      "loss": 0.3063,
+      "step": 28989
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.026413446845667e-05,
+      "loss": 0.386,
+      "step": 28990
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0263798633150304e-05,
+      "loss": 0.381,
+      "step": 28991
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0263462797843937e-05,
+      "loss": 0.3259,
+      "step": 28992
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0263126962537573e-05,
+      "loss": 0.2732,
+      "step": 28993
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0262791127231206e-05,
+      "loss": 0.2823,
+      "step": 28994
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.026245529192484e-05,
+      "loss": 0.1802,
+      "step": 28995
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0262119456618475e-05,
+      "loss": 0.1414,
+      "step": 28996
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.026178362131211e-05,
+      "loss": 0.1887,
+      "step": 28997
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0261447786005743e-05,
+      "loss": 0.396,
+      "step": 28998
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0261111950699378e-05,
+      "loss": 0.3165,
+      "step": 28999
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0260776115393012e-05,
+      "loss": 0.4133,
+      "step": 29000
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0260440280086645e-05,
+      "loss": 0.2316,
+      "step": 29001
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0260104444780282e-05,
+      "loss": 0.1685,
+      "step": 29002
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0259768609473914e-05,
+      "loss": 0.4598,
+      "step": 29003
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0259432774167547e-05,
+      "loss": 0.3027,
+      "step": 29004
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0259096938861184e-05,
+      "loss": 0.2387,
+      "step": 29005
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0258761103554817e-05,
+      "loss": 0.2904,
+      "step": 29006
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0258425268248451e-05,
+      "loss": 0.5022,
+      "step": 29007
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0258089432942086e-05,
+      "loss": 0.2558,
+      "step": 29008
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.025775359763572e-05,
+      "loss": 0.2219,
+      "step": 29009
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0257417762329354e-05,
+      "loss": 0.3253,
+      "step": 29010
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.025708192702299e-05,
+      "loss": 0.2036,
+      "step": 29011
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0256746091716623e-05,
+      "loss": 0.2547,
+      "step": 29012
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0256410256410256e-05,
+      "loss": 0.274,
+      "step": 29013
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0256074421103892e-05,
+      "loss": 0.289,
+      "step": 29014
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0255738585797525e-05,
+      "loss": 0.1435,
+      "step": 29015
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.025540275049116e-05,
+      "loss": 0.2084,
+      "step": 29016
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0255066915184794e-05,
+      "loss": 0.1528,
+      "step": 29017
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0254731079878429e-05,
+      "loss": 0.3054,
+      "step": 29018
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0254395244572062e-05,
+      "loss": 0.255,
+      "step": 29019
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0254059409265697e-05,
+      "loss": 0.2624,
+      "step": 29020
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0253723573959331e-05,
+      "loss": 0.1706,
+      "step": 29021
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0253387738652964e-05,
+      "loss": 0.2653,
+      "step": 29022
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.02530519033466e-05,
+      "loss": 0.2933,
+      "step": 29023
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0252716068040234e-05,
+      "loss": 0.1872,
+      "step": 29024
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0252380232733867e-05,
+      "loss": 0.2081,
+      "step": 29025
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0252044397427503e-05,
+      "loss": 0.2847,
+      "step": 29026
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0251708562121136e-05,
+      "loss": 0.2173,
+      "step": 29027
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.025137272681477e-05,
+      "loss": 0.2583,
+      "step": 29028
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0251036891508405e-05,
+      "loss": 0.3707,
+      "step": 29029
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.025070105620204e-05,
+      "loss": 0.2008,
+      "step": 29030
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0250365220895673e-05,
+      "loss": 0.1893,
+      "step": 29031
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0250029385589309e-05,
+      "loss": 0.2994,
+      "step": 29032
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0249693550282942e-05,
+      "loss": 0.1912,
+      "step": 29033
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0249357714976575e-05,
+      "loss": 0.2914,
+      "step": 29034
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0249021879670211e-05,
+      "loss": 0.4501,
+      "step": 29035
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0248686044363844e-05,
+      "loss": 0.2529,
+      "step": 29036
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0248350209057479e-05,
+      "loss": 0.2064,
+      "step": 29037
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0248014373751114e-05,
+      "loss": 0.413,
+      "step": 29038
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0247678538444748e-05,
+      "loss": 0.3406,
+      "step": 29039
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0247342703138381e-05,
+      "loss": 0.2838,
+      "step": 29040
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0247006867832018e-05,
+      "loss": 0.3434,
+      "step": 29041
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.024667103252565e-05,
+      "loss": 0.2333,
+      "step": 29042
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0246335197219283e-05,
+      "loss": 0.4333,
+      "step": 29043
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.024599936191292e-05,
+      "loss": 0.3114,
+      "step": 29044
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0245663526606553e-05,
+      "loss": 0.1728,
+      "step": 29045
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0245327691300186e-05,
+      "loss": 0.2988,
+      "step": 29046
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0244991855993822e-05,
+      "loss": 0.3265,
+      "step": 29047
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0244656020687455e-05,
+      "loss": 0.2676,
+      "step": 29048
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.024432018538109e-05,
+      "loss": 0.1842,
+      "step": 29049
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0243984350074724e-05,
+      "loss": 0.4738,
+      "step": 29050
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0243648514768359e-05,
+      "loss": 0.295,
+      "step": 29051
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0243312679461992e-05,
+      "loss": 0.2724,
+      "step": 29052
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0242976844155628e-05,
+      "loss": 0.3211,
+      "step": 29053
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0242641008849261e-05,
+      "loss": 0.4487,
+      "step": 29054
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0242305173542894e-05,
+      "loss": 0.2211,
+      "step": 29055
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.024196933823653e-05,
+      "loss": 0.3233,
+      "step": 29056
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0241633502930163e-05,
+      "loss": 0.2519,
+      "step": 29057
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0241297667623798e-05,
+      "loss": 0.3307,
+      "step": 29058
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0240961832317433e-05,
+      "loss": 0.1791,
+      "step": 29059
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0240625997011067e-05,
+      "loss": 0.2509,
+      "step": 29060
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.02402901617047e-05,
+      "loss": 0.3567,
+      "step": 29061
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0239954326398337e-05,
+      "loss": 0.249,
+      "step": 29062
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.023961849109197e-05,
+      "loss": 0.22,
+      "step": 29063
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0239282655785603e-05,
+      "loss": 0.1721,
+      "step": 29064
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0238946820479239e-05,
+      "loss": 0.2112,
+      "step": 29065
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0238610985172872e-05,
+      "loss": 0.2816,
+      "step": 29066
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0238275149866505e-05,
+      "loss": 0.2367,
+      "step": 29067
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0237939314560141e-05,
+      "loss": 0.3874,
+      "step": 29068
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0237603479253774e-05,
+      "loss": 0.3532,
+      "step": 29069
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0237267643947409e-05,
+      "loss": 0.3517,
+      "step": 29070
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0236931808641043e-05,
+      "loss": 0.1621,
+      "step": 29071
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0236595973334678e-05,
+      "loss": 0.1893,
+      "step": 29072
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0236260138028311e-05,
+      "loss": 0.2254,
+      "step": 29073
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0235924302721947e-05,
+      "loss": 0.2842,
+      "step": 29074
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.023558846741558e-05,
+      "loss": 0.1975,
+      "step": 29075
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0235252632109213e-05,
+      "loss": 0.3238,
+      "step": 29076
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.023491679680285e-05,
+      "loss": 0.3893,
+      "step": 29077
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0234580961496483e-05,
+      "loss": 0.2994,
+      "step": 29078
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0234245126190117e-05,
+      "loss": 0.3669,
+      "step": 29079
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0233909290883752e-05,
+      "loss": 0.2262,
+      "step": 29080
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 1.0233573455577386e-05,
+      "loss": 0.3032,
+      "step": 29081
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.023323762027102e-05,
+      "loss": 0.309,
+      "step": 29082
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0232901784964656e-05,
+      "loss": 0.1996,
+      "step": 29083
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0232565949658289e-05,
+      "loss": 0.2857,
+      "step": 29084
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0232230114351922e-05,
+      "loss": 0.2858,
+      "step": 29085
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0231894279045558e-05,
+      "loss": 0.255,
+      "step": 29086
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0231558443739191e-05,
+      "loss": 0.4174,
+      "step": 29087
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0231222608432826e-05,
+      "loss": 0.4321,
+      "step": 29088
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.023088677312646e-05,
+      "loss": 0.2433,
+      "step": 29089
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0230550937820093e-05,
+      "loss": 0.2915,
+      "step": 29090
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0230215102513728e-05,
+      "loss": 0.2976,
+      "step": 29091
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0229879267207363e-05,
+      "loss": 0.3116,
+      "step": 29092
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0229543431900997e-05,
+      "loss": 0.3959,
+      "step": 29093
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.022920759659463e-05,
+      "loss": 0.3201,
+      "step": 29094
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0228871761288266e-05,
+      "loss": 0.2285,
+      "step": 29095
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.02285359259819e-05,
+      "loss": 0.389,
+      "step": 29096
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0228200090675532e-05,
+      "loss": 0.3386,
+      "step": 29097
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0227864255369169e-05,
+      "loss": 0.3409,
+      "step": 29098
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0227528420062802e-05,
+      "loss": 0.313,
+      "step": 29099
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0227192584756436e-05,
+      "loss": 0.2894,
+      "step": 29100
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0226856749450071e-05,
+      "loss": 0.1969,
+      "step": 29101
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0226520914143706e-05,
+      "loss": 0.242,
+      "step": 29102
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0226185078837339e-05,
+      "loss": 0.3036,
+      "step": 29103
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0225849243530975e-05,
+      "loss": 0.4027,
+      "step": 29104
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0225513408224608e-05,
+      "loss": 0.1644,
+      "step": 29105
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.022517757291824e-05,
+      "loss": 0.287,
+      "step": 29106
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0224841737611877e-05,
+      "loss": 0.3612,
+      "step": 29107
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.022450590230551e-05,
+      "loss": 0.3381,
+      "step": 29108
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0224170066999145e-05,
+      "loss": 0.2109,
+      "step": 29109
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.022383423169278e-05,
+      "loss": 0.1936,
+      "step": 29110
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0223498396386412e-05,
+      "loss": 0.1895,
+      "step": 29111
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0223162561080047e-05,
+      "loss": 0.3619,
+      "step": 29112
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0222826725773682e-05,
+      "loss": 0.257,
+      "step": 29113
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0222490890467316e-05,
+      "loss": 0.3407,
+      "step": 29114
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.022215505516095e-05,
+      "loss": 0.1652,
+      "step": 29115
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0221819219854586e-05,
+      "loss": 0.1448,
+      "step": 29116
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0221483384548219e-05,
+      "loss": 0.1662,
+      "step": 29117
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0221147549241851e-05,
+      "loss": 0.3383,
+      "step": 29118
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0220811713935488e-05,
+      "loss": 0.2899,
+      "step": 29119
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.022047587862912e-05,
+      "loss": 0.3603,
+      "step": 29120
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0220140043322755e-05,
+      "loss": 0.3359,
+      "step": 29121
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.021980420801639e-05,
+      "loss": 0.2501,
+      "step": 29122
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0219468372710025e-05,
+      "loss": 0.4446,
+      "step": 29123
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0219132537403658e-05,
+      "loss": 0.1622,
+      "step": 29124
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0218796702097294e-05,
+      "loss": 0.49,
+      "step": 29125
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0218460866790927e-05,
+      "loss": 0.332,
+      "step": 29126
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.021812503148456e-05,
+      "loss": 0.2565,
+      "step": 29127
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0217789196178196e-05,
+      "loss": 0.5186,
+      "step": 29128
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.021745336087183e-05,
+      "loss": 0.2216,
+      "step": 29129
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0217117525565464e-05,
+      "loss": 0.4384,
+      "step": 29130
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0216781690259099e-05,
+      "loss": 0.3435,
+      "step": 29131
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0216445854952733e-05,
+      "loss": 0.2134,
+      "step": 29132
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0216110019646366e-05,
+      "loss": 0.3447,
+      "step": 29133
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.021577418434e-05,
+      "loss": 0.2802,
+      "step": 29134
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0215438349033635e-05,
+      "loss": 0.3326,
+      "step": 29135
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0215102513727268e-05,
+      "loss": 0.385,
+      "step": 29136
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0214766678420905e-05,
+      "loss": 0.2462,
+      "step": 29137
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0214430843114538e-05,
+      "loss": 0.2876,
+      "step": 29138
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.021409500780817e-05,
+      "loss": 0.3441,
+      "step": 29139
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0213759172501807e-05,
+      "loss": 0.2704,
+      "step": 29140
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.021342333719544e-05,
+      "loss": 0.1879,
+      "step": 29141
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0213087501889075e-05,
+      "loss": 0.2912,
+      "step": 29142
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.021275166658271e-05,
+      "loss": 0.2361,
+      "step": 29143
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0212415831276344e-05,
+      "loss": 0.3465,
+      "step": 29144
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0212079995969977e-05,
+      "loss": 0.4728,
+      "step": 29145
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.021174416066361e-05,
+      "loss": 0.2765,
+      "step": 29146
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0211408325357246e-05,
+      "loss": 0.1661,
+      "step": 29147
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0211072490050879e-05,
+      "loss": 0.3104,
+      "step": 29148
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0210736654744514e-05,
+      "loss": 0.2231,
+      "step": 29149
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0210400819438148e-05,
+      "loss": 0.377,
+      "step": 29150
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0210064984131783e-05,
+      "loss": 0.3298,
+      "step": 29151
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0209729148825416e-05,
+      "loss": 0.3958,
+      "step": 29152
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0209393313519052e-05,
+      "loss": 0.2319,
+      "step": 29153
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0209057478212685e-05,
+      "loss": 0.1806,
+      "step": 29154
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0208721642906318e-05,
+      "loss": 0.1549,
+      "step": 29155
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0208385807599955e-05,
+      "loss": 0.465,
+      "step": 29156
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0208049972293587e-05,
+      "loss": 0.3325,
+      "step": 29157
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.020771413698722e-05,
+      "loss": 0.3145,
+      "step": 29158
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0207378301680857e-05,
+      "loss": 0.3513,
+      "step": 29159
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.020704246637449e-05,
+      "loss": 0.1261,
+      "step": 29160
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0206706631068124e-05,
+      "loss": 0.3307,
+      "step": 29161
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0206370795761759e-05,
+      "loss": 0.1565,
+      "step": 29162
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0206034960455394e-05,
+      "loss": 0.3056,
+      "step": 29163
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0205699125149027e-05,
+      "loss": 0.2653,
+      "step": 29164
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0205363289842663e-05,
+      "loss": 0.1556,
+      "step": 29165
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0205027454536296e-05,
+      "loss": 0.2074,
+      "step": 29166
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0204691619229929e-05,
+      "loss": 0.3751,
+      "step": 29167
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0204355783923565e-05,
+      "loss": 0.4165,
+      "step": 29168
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0204019948617198e-05,
+      "loss": 0.3667,
+      "step": 29169
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0203684113310833e-05,
+      "loss": 0.3405,
+      "step": 29170
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0203348278004467e-05,
+      "loss": 0.2158,
+      "step": 29171
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0203012442698102e-05,
+      "loss": 0.1691,
+      "step": 29172
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0202676607391735e-05,
+      "loss": 0.3524,
+      "step": 29173
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0202340772085371e-05,
+      "loss": 0.3159,
+      "step": 29174
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0202004936779004e-05,
+      "loss": 0.4997,
+      "step": 29175
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0201669101472637e-05,
+      "loss": 0.4586,
+      "step": 29176
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0201333266166274e-05,
+      "loss": 0.4991,
+      "step": 29177
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0200997430859907e-05,
+      "loss": 0.3419,
+      "step": 29178
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.020066159555354e-05,
+      "loss": 0.3576,
+      "step": 29179
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0200325760247176e-05,
+      "loss": 0.2831,
+      "step": 29180
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0199989924940809e-05,
+      "loss": 0.2514,
+      "step": 29181
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0199654089634444e-05,
+      "loss": 0.3822,
+      "step": 29182
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0199318254328078e-05,
+      "loss": 0.224,
+      "step": 29183
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0198982419021713e-05,
+      "loss": 0.5823,
+      "step": 29184
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0198646583715346e-05,
+      "loss": 0.2442,
+      "step": 29185
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0198310748408982e-05,
+      "loss": 0.3915,
+      "step": 29186
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0197974913102615e-05,
+      "loss": 0.2472,
+      "step": 29187
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0197639077796248e-05,
+      "loss": 0.2037,
+      "step": 29188
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0197303242489884e-05,
+      "loss": 0.2608,
+      "step": 29189
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0196967407183517e-05,
+      "loss": 0.3569,
+      "step": 29190
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0196631571877152e-05,
+      "loss": 0.1985,
+      "step": 29191
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0196295736570787e-05,
+      "loss": 0.4143,
+      "step": 29192
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0195959901264421e-05,
+      "loss": 0.3407,
+      "step": 29193
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0195624065958054e-05,
+      "loss": 0.262,
+      "step": 29194
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.019528823065169e-05,
+      "loss": 0.3686,
+      "step": 29195
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0194952395345323e-05,
+      "loss": 0.323,
+      "step": 29196
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0194616560038956e-05,
+      "loss": 0.4497,
+      "step": 29197
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0194280724732593e-05,
+      "loss": 0.1958,
+      "step": 29198
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0193944889426226e-05,
+      "loss": 0.186,
+      "step": 29199
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.019360905411986e-05,
+      "loss": 0.3108,
+      "step": 29200
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0193273218813495e-05,
+      "loss": 0.3681,
+      "step": 29201
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0192937383507128e-05,
+      "loss": 0.4456,
+      "step": 29202
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0192601548200763e-05,
+      "loss": 0.3648,
+      "step": 29203
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0192265712894397e-05,
+      "loss": 0.2163,
+      "step": 29204
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0191929877588032e-05,
+      "loss": 0.2757,
+      "step": 29205
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0191594042281665e-05,
+      "loss": 0.2934,
+      "step": 29206
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0191258206975301e-05,
+      "loss": 0.4442,
+      "step": 29207
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0190922371668934e-05,
+      "loss": 0.2999,
+      "step": 29208
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0190586536362567e-05,
+      "loss": 0.3412,
+      "step": 29209
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0190250701056203e-05,
+      "loss": 0.1657,
+      "step": 29210
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0189914865749836e-05,
+      "loss": 0.2603,
+      "step": 29211
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0189579030443471e-05,
+      "loss": 0.3185,
+      "step": 29212
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0189243195137106e-05,
+      "loss": 0.2015,
+      "step": 29213
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.018890735983074e-05,
+      "loss": 0.2233,
+      "step": 29214
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0188571524524373e-05,
+      "loss": 0.2821,
+      "step": 29215
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.018823568921801e-05,
+      "loss": 0.417,
+      "step": 29216
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0187899853911643e-05,
+      "loss": 0.2025,
+      "step": 29217
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0187564018605276e-05,
+      "loss": 0.2934,
+      "step": 29218
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0187228183298912e-05,
+      "loss": 0.3567,
+      "step": 29219
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0186892347992545e-05,
+      "loss": 0.326,
+      "step": 29220
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.018655651268618e-05,
+      "loss": 0.2239,
+      "step": 29221
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0186220677379814e-05,
+      "loss": 0.4408,
+      "step": 29222
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0185884842073449e-05,
+      "loss": 0.1462,
+      "step": 29223
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0185549006767082e-05,
+      "loss": 0.2953,
+      "step": 29224
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0185213171460716e-05,
+      "loss": 0.1748,
+      "step": 29225
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0184877336154351e-05,
+      "loss": 0.2536,
+      "step": 29226
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0184541500847984e-05,
+      "loss": 0.2959,
+      "step": 29227
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.018420566554162e-05,
+      "loss": 0.2873,
+      "step": 29228
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0183869830235253e-05,
+      "loss": 0.2099,
+      "step": 29229
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0183533994928886e-05,
+      "loss": 0.2744,
+      "step": 29230
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0183198159622523e-05,
+      "loss": 0.3347,
+      "step": 29231
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0182862324316156e-05,
+      "loss": 0.1673,
+      "step": 29232
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.018252648900979e-05,
+      "loss": 0.13,
+      "step": 29233
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0182190653703425e-05,
+      "loss": 0.3045,
+      "step": 29234
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.018185481839706e-05,
+      "loss": 0.299,
+      "step": 29235
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0181518983090692e-05,
+      "loss": 0.3136,
+      "step": 29236
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0181183147784329e-05,
+      "loss": 0.5137,
+      "step": 29237
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0180847312477962e-05,
+      "loss": 0.4443,
+      "step": 29238
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0180511477171595e-05,
+      "loss": 0.1684,
+      "step": 29239
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0180175641865231e-05,
+      "loss": 0.3679,
+      "step": 29240
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0179839806558864e-05,
+      "loss": 0.2157,
+      "step": 29241
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0179503971252499e-05,
+      "loss": 0.3699,
+      "step": 29242
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0179168135946133e-05,
+      "loss": 0.2292,
+      "step": 29243
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0178832300639768e-05,
+      "loss": 0.1211,
+      "step": 29244
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0178496465333401e-05,
+      "loss": 0.5124,
+      "step": 29245
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0178160630027036e-05,
+      "loss": 0.1815,
+      "step": 29246
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.017782479472067e-05,
+      "loss": 0.2286,
+      "step": 29247
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0177488959414303e-05,
+      "loss": 0.1644,
+      "step": 29248
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.017715312410794e-05,
+      "loss": 0.224,
+      "step": 29249
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0176817288801572e-05,
+      "loss": 0.26,
+      "step": 29250
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0176481453495205e-05,
+      "loss": 0.2865,
+      "step": 29251
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0176145618188842e-05,
+      "loss": 0.3084,
+      "step": 29252
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0175809782882475e-05,
+      "loss": 0.3167,
+      "step": 29253
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.017547394757611e-05,
+      "loss": 0.29,
+      "step": 29254
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0175138112269744e-05,
+      "loss": 0.416,
+      "step": 29255
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0174802276963379e-05,
+      "loss": 0.3998,
+      "step": 29256
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0174466441657012e-05,
+      "loss": 0.1699,
+      "step": 29257
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0174130606350648e-05,
+      "loss": 0.1584,
+      "step": 29258
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0173794771044281e-05,
+      "loss": 0.3205,
+      "step": 29259
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0173458935737914e-05,
+      "loss": 0.3453,
+      "step": 29260
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.017312310043155e-05,
+      "loss": 0.3964,
+      "step": 29261
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0172787265125183e-05,
+      "loss": 0.2166,
+      "step": 29262
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0172451429818818e-05,
+      "loss": 0.3017,
+      "step": 29263
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0172115594512452e-05,
+      "loss": 0.276,
+      "step": 29264
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0171779759206087e-05,
+      "loss": 0.2397,
+      "step": 29265
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.017144392389972e-05,
+      "loss": 0.355,
+      "step": 29266
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0171108088593356e-05,
+      "loss": 0.2765,
+      "step": 29267
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.017077225328699e-05,
+      "loss": 0.2152,
+      "step": 29268
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0170436417980622e-05,
+      "loss": 0.2533,
+      "step": 29269
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0170100582674259e-05,
+      "loss": 0.1753,
+      "step": 29270
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0169764747367892e-05,
+      "loss": 0.2415,
+      "step": 29271
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0169428912061524e-05,
+      "loss": 0.3662,
+      "step": 29272
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.016909307675516e-05,
+      "loss": 0.2065,
+      "step": 29273
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0168757241448794e-05,
+      "loss": 0.3234,
+      "step": 29274
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0168421406142428e-05,
+      "loss": 0.4112,
+      "step": 29275
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0168085570836063e-05,
+      "loss": 0.2612,
+      "step": 29276
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0167749735529698e-05,
+      "loss": 0.532,
+      "step": 29277
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.016741390022333e-05,
+      "loss": 0.3737,
+      "step": 29278
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.0167078064916967e-05,
+      "loss": 0.3302,
+      "step": 29279
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.01667422296106e-05,
+      "loss": 0.6248,
+      "step": 29280
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0166406394304233e-05,
+      "loss": 0.2346,
+      "step": 29281
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.016607055899787e-05,
+      "loss": 0.4309,
+      "step": 29282
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0165734723691502e-05,
+      "loss": 0.3155,
+      "step": 29283
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0165398888385137e-05,
+      "loss": 0.2269,
+      "step": 29284
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0165063053078772e-05,
+      "loss": 0.3529,
+      "step": 29285
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0164727217772406e-05,
+      "loss": 0.3201,
+      "step": 29286
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0164391382466039e-05,
+      "loss": 0.3514,
+      "step": 29287
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0164055547159675e-05,
+      "loss": 0.1871,
+      "step": 29288
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0163719711853308e-05,
+      "loss": 0.3544,
+      "step": 29289
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0163383876546941e-05,
+      "loss": 0.3979,
+      "step": 29290
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0163048041240578e-05,
+      "loss": 0.2373,
+      "step": 29291
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.016271220593421e-05,
+      "loss": 0.2295,
+      "step": 29292
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0162376370627844e-05,
+      "loss": 0.1755,
+      "step": 29293
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.016204053532148e-05,
+      "loss": 0.2309,
+      "step": 29294
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0161704700015113e-05,
+      "loss": 0.2583,
+      "step": 29295
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0161368864708748e-05,
+      "loss": 0.2963,
+      "step": 29296
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0161033029402382e-05,
+      "loss": 0.208,
+      "step": 29297
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0160697194096017e-05,
+      "loss": 0.3048,
+      "step": 29298
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.016036135878965e-05,
+      "loss": 0.2091,
+      "step": 29299
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0160025523483286e-05,
+      "loss": 0.2098,
+      "step": 29300
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0159689688176919e-05,
+      "loss": 0.2536,
+      "step": 29301
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0159353852870552e-05,
+      "loss": 0.2116,
+      "step": 29302
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0159018017564188e-05,
+      "loss": 0.3143,
+      "step": 29303
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0158682182257821e-05,
+      "loss": 0.3144,
+      "step": 29304
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0158346346951456e-05,
+      "loss": 0.2578,
+      "step": 29305
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.015801051164509e-05,
+      "loss": 0.3589,
+      "step": 29306
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0157674676338725e-05,
+      "loss": 0.2476,
+      "step": 29307
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0157338841032358e-05,
+      "loss": 0.2096,
+      "step": 29308
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0157003005725995e-05,
+      "loss": 0.3211,
+      "step": 29309
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0156667170419628e-05,
+      "loss": 0.3256,
+      "step": 29310
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.015633133511326e-05,
+      "loss": 0.3848,
+      "step": 29311
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0155995499806897e-05,
+      "loss": 0.168,
+      "step": 29312
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.015565966450053e-05,
+      "loss": 0.269,
+      "step": 29313
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0155323829194163e-05,
+      "loss": 0.303,
+      "step": 29314
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0154987993887799e-05,
+      "loss": 0.2792,
+      "step": 29315
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0154652158581432e-05,
+      "loss": 0.2432,
+      "step": 29316
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0154316323275067e-05,
+      "loss": 0.3258,
+      "step": 29317
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0153980487968701e-05,
+      "loss": 0.407,
+      "step": 29318
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0153644652662336e-05,
+      "loss": 0.2192,
+      "step": 29319
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0153308817355969e-05,
+      "loss": 0.2971,
+      "step": 29320
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0152972982049605e-05,
+      "loss": 0.2757,
+      "step": 29321
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0152637146743238e-05,
+      "loss": 0.5185,
+      "step": 29322
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0152301311436871e-05,
+      "loss": 0.1795,
+      "step": 29323
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0151965476130508e-05,
+      "loss": 0.3296,
+      "step": 29324
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.015162964082414e-05,
+      "loss": 0.2757,
+      "step": 29325
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0151293805517775e-05,
+      "loss": 0.373,
+      "step": 29326
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0150957970211408e-05,
+      "loss": 0.3937,
+      "step": 29327
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0150622134905044e-05,
+      "loss": 0.2243,
+      "step": 29328
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0150286299598677e-05,
+      "loss": 0.347,
+      "step": 29329
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.014995046429231e-05,
+      "loss": 0.3019,
+      "step": 29330
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0149614628985947e-05,
+      "loss": 0.4023,
+      "step": 29331
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.014927879367958e-05,
+      "loss": 0.3196,
+      "step": 29332
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0148942958373214e-05,
+      "loss": 0.3942,
+      "step": 29333
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0148607123066849e-05,
+      "loss": 0.1929,
+      "step": 29334
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0148271287760484e-05,
+      "loss": 0.2391,
+      "step": 29335
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0147935452454117e-05,
+      "loss": 0.3253,
+      "step": 29336
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0147599617147751e-05,
+      "loss": 0.3922,
+      "step": 29337
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0147263781841386e-05,
+      "loss": 0.2488,
+      "step": 29338
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0146927946535019e-05,
+      "loss": 0.4802,
+      "step": 29339
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0146592111228655e-05,
+      "loss": 0.2922,
+      "step": 29340
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0146256275922288e-05,
+      "loss": 0.2996,
+      "step": 29341
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0145920440615921e-05,
+      "loss": 0.3877,
+      "step": 29342
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0145584605309557e-05,
+      "loss": 0.0998,
+      "step": 29343
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.014524877000319e-05,
+      "loss": 0.2834,
+      "step": 29344
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0144912934696825e-05,
+      "loss": 0.2364,
+      "step": 29345
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.014457709939046e-05,
+      "loss": 0.1577,
+      "step": 29346
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0144241264084094e-05,
+      "loss": 0.4621,
+      "step": 29347
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0143905428777727e-05,
+      "loss": 0.3761,
+      "step": 29348
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0143569593471364e-05,
+      "loss": 0.2892,
+      "step": 29349
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0143233758164997e-05,
+      "loss": 0.1888,
+      "step": 29350
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.014289792285863e-05,
+      "loss": 0.6107,
+      "step": 29351
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0142562087552266e-05,
+      "loss": 0.3424,
+      "step": 29352
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0142226252245899e-05,
+      "loss": 0.1826,
+      "step": 29353
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0141890416939533e-05,
+      "loss": 0.275,
+      "step": 29354
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0141554581633168e-05,
+      "loss": 0.3243,
+      "step": 29355
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0141218746326803e-05,
+      "loss": 0.2859,
+      "step": 29356
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0140882911020436e-05,
+      "loss": 0.3676,
+      "step": 29357
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0140547075714072e-05,
+      "loss": 0.1205,
+      "step": 29358
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0140211240407705e-05,
+      "loss": 0.2996,
+      "step": 29359
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0139875405101338e-05,
+      "loss": 0.2941,
+      "step": 29360
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0139539569794974e-05,
+      "loss": 0.3751,
+      "step": 29361
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0139203734488607e-05,
+      "loss": 0.2707,
+      "step": 29362
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.013886789918224e-05,
+      "loss": 0.2615,
+      "step": 29363
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0138532063875876e-05,
+      "loss": 0.3789,
+      "step": 29364
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.013819622856951e-05,
+      "loss": 0.2955,
+      "step": 29365
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0137860393263144e-05,
+      "loss": 0.3272,
+      "step": 29366
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0137524557956779e-05,
+      "loss": 0.1754,
+      "step": 29367
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0137188722650413e-05,
+      "loss": 0.3151,
+      "step": 29368
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0136852887344046e-05,
+      "loss": 0.1908,
+      "step": 29369
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0136517052037683e-05,
+      "loss": 0.3152,
+      "step": 29370
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0136181216731316e-05,
+      "loss": 0.3029,
+      "step": 29371
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0135845381424949e-05,
+      "loss": 0.2052,
+      "step": 29372
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0135509546118585e-05,
+      "loss": 0.3476,
+      "step": 29373
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0135173710812218e-05,
+      "loss": 0.3493,
+      "step": 29374
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0134837875505853e-05,
+      "loss": 0.195,
+      "step": 29375
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0134502040199487e-05,
+      "loss": 0.2915,
+      "step": 29376
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0134166204893122e-05,
+      "loss": 0.2926,
+      "step": 29377
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0133830369586755e-05,
+      "loss": 0.3518,
+      "step": 29378
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0133494534280391e-05,
+      "loss": 0.1485,
+      "step": 29379
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0133158698974024e-05,
+      "loss": 0.2643,
+      "step": 29380
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0132822863667657e-05,
+      "loss": 0.5378,
+      "step": 29381
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0132487028361293e-05,
+      "loss": 0.2483,
+      "step": 29382
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0132151193054926e-05,
+      "loss": 0.2415,
+      "step": 29383
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.013181535774856e-05,
+      "loss": 0.5612,
+      "step": 29384
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0131479522442196e-05,
+      "loss": 0.4902,
+      "step": 29385
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0131143687135829e-05,
+      "loss": 0.2546,
+      "step": 29386
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0130807851829463e-05,
+      "loss": 0.2764,
+      "step": 29387
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0130472016523098e-05,
+      "loss": 0.1857,
+      "step": 29388
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0130136181216733e-05,
+      "loss": 0.1904,
+      "step": 29389
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0129800345910365e-05,
+      "loss": 0.523,
+      "step": 29390
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0129464510604002e-05,
+      "loss": 0.3075,
+      "step": 29391
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0129128675297635e-05,
+      "loss": 0.1831,
+      "step": 29392
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0128792839991268e-05,
+      "loss": 0.3958,
+      "step": 29393
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0128457004684904e-05,
+      "loss": 0.4323,
+      "step": 29394
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0128121169378537e-05,
+      "loss": 0.2674,
+      "step": 29395
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0127785334072172e-05,
+      "loss": 0.3249,
+      "step": 29396
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0127449498765806e-05,
+      "loss": 0.3004,
+      "step": 29397
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0127113663459441e-05,
+      "loss": 0.2936,
+      "step": 29398
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0126777828153074e-05,
+      "loss": 0.3088,
+      "step": 29399
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.012644199284671e-05,
+      "loss": 0.4158,
+      "step": 29400
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0126106157540343e-05,
+      "loss": 0.222,
+      "step": 29401
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0125770322233976e-05,
+      "loss": 0.3941,
+      "step": 29402
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0125434486927612e-05,
+      "loss": 0.291,
+      "step": 29403
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0125098651621245e-05,
+      "loss": 0.2842,
+      "step": 29404
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0124762816314878e-05,
+      "loss": 0.3464,
+      "step": 29405
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0124426981008515e-05,
+      "loss": 0.1894,
+      "step": 29406
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0124091145702148e-05,
+      "loss": 0.4399,
+      "step": 29407
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0123755310395782e-05,
+      "loss": 0.3198,
+      "step": 29408
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0123419475089417e-05,
+      "loss": 0.1802,
+      "step": 29409
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0123083639783052e-05,
+      "loss": 0.2116,
+      "step": 29410
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0122747804476685e-05,
+      "loss": 0.2946,
+      "step": 29411
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0122411969170321e-05,
+      "loss": 0.3302,
+      "step": 29412
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0122076133863954e-05,
+      "loss": 0.2847,
+      "step": 29413
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0121740298557587e-05,
+      "loss": 0.191,
+      "step": 29414
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0121404463251223e-05,
+      "loss": 0.2471,
+      "step": 29415
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0121068627944856e-05,
+      "loss": 0.2681,
+      "step": 29416
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.012073279263849e-05,
+      "loss": 0.1994,
+      "step": 29417
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0120396957332125e-05,
+      "loss": 0.2036,
+      "step": 29418
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.012006112202576e-05,
+      "loss": 0.3401,
+      "step": 29419
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0119725286719393e-05,
+      "loss": 0.2843,
+      "step": 29420
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.011938945141303e-05,
+      "loss": 0.3057,
+      "step": 29421
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0119053616106662e-05,
+      "loss": 0.2044,
+      "step": 29422
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0118717780800295e-05,
+      "loss": 0.2042,
+      "step": 29423
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0118381945493932e-05,
+      "loss": 0.1926,
+      "step": 29424
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0118046110187565e-05,
+      "loss": 0.1819,
+      "step": 29425
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.01177102748812e-05,
+      "loss": 0.3506,
+      "step": 29426
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0117374439574834e-05,
+      "loss": 0.3947,
+      "step": 29427
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0117038604268467e-05,
+      "loss": 0.4688,
+      "step": 29428
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0116702768962101e-05,
+      "loss": 0.3953,
+      "step": 29429
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0116366933655736e-05,
+      "loss": 0.2384,
+      "step": 29430
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.011603109834937e-05,
+      "loss": 0.2659,
+      "step": 29431
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0115695263043004e-05,
+      "loss": 0.2426,
+      "step": 29432
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.011535942773664e-05,
+      "loss": 0.5413,
+      "step": 29433
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0115023592430273e-05,
+      "loss": 0.2316,
+      "step": 29434
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0114687757123906e-05,
+      "loss": 0.2012,
+      "step": 29435
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0114351921817542e-05,
+      "loss": 0.1661,
+      "step": 29436
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0114016086511175e-05,
+      "loss": 0.1317,
+      "step": 29437
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.011368025120481e-05,
+      "loss": 0.3307,
+      "step": 29438
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0113344415898445e-05,
+      "loss": 0.2553,
+      "step": 29439
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.011300858059208e-05,
+      "loss": 0.1646,
+      "step": 29440
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0112672745285712e-05,
+      "loss": 0.4903,
+      "step": 29441
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0112336909979348e-05,
+      "loss": 0.3863,
+      "step": 29442
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0112001074672981e-05,
+      "loss": 0.5938,
+      "step": 29443
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0111665239366614e-05,
+      "loss": 0.3727,
+      "step": 29444
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.011132940406025e-05,
+      "loss": 0.2288,
+      "step": 29445
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0110993568753884e-05,
+      "loss": 0.1879,
+      "step": 29446
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0110657733447518e-05,
+      "loss": 0.3957,
+      "step": 29447
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0110321898141153e-05,
+      "loss": 0.3514,
+      "step": 29448
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0109986062834788e-05,
+      "loss": 0.3319,
+      "step": 29449
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.010965022752842e-05,
+      "loss": 0.3519,
+      "step": 29450
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0109314392222055e-05,
+      "loss": 0.2348,
+      "step": 29451
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.010897855691569e-05,
+      "loss": 0.1743,
+      "step": 29452
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0108642721609323e-05,
+      "loss": 0.3256,
+      "step": 29453
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.010830688630296e-05,
+      "loss": 0.2382,
+      "step": 29454
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0107971050996592e-05,
+      "loss": 0.2487,
+      "step": 29455
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0107635215690225e-05,
+      "loss": 0.3604,
+      "step": 29456
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0107299380383861e-05,
+      "loss": 0.399,
+      "step": 29457
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0106963545077494e-05,
+      "loss": 0.2125,
+      "step": 29458
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0106627709771129e-05,
+      "loss": 0.4021,
+      "step": 29459
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0106291874464764e-05,
+      "loss": 0.195,
+      "step": 29460
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0105956039158398e-05,
+      "loss": 0.1851,
+      "step": 29461
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0105620203852031e-05,
+      "loss": 0.2803,
+      "step": 29462
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0105284368545668e-05,
+      "loss": 0.3595,
+      "step": 29463
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.01049485332393e-05,
+      "loss": 0.2382,
+      "step": 29464
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0104612697932934e-05,
+      "loss": 0.2503,
+      "step": 29465
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.010427686262657e-05,
+      "loss": 0.2638,
+      "step": 29466
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0103941027320203e-05,
+      "loss": 0.2197,
+      "step": 29467
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0103605192013837e-05,
+      "loss": 0.4369,
+      "step": 29468
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0103269356707472e-05,
+      "loss": 0.3811,
+      "step": 29469
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0102933521401107e-05,
+      "loss": 0.2033,
+      "step": 29470
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.010259768609474e-05,
+      "loss": 0.4756,
+      "step": 29471
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0102261850788374e-05,
+      "loss": 0.2282,
+      "step": 29472
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0101926015482009e-05,
+      "loss": 0.1691,
+      "step": 29473
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0101590180175642e-05,
+      "loss": 0.3537,
+      "step": 29474
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0101254344869278e-05,
+      "loss": 0.1723,
+      "step": 29475
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0100918509562911e-05,
+      "loss": 0.2503,
+      "step": 29476
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.0100582674256544e-05,
+      "loss": 0.3158,
+      "step": 29477
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 1.010024683895018e-05,
+      "loss": 0.4448,
+      "step": 29478
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0099911003643813e-05,
+      "loss": 0.3436,
+      "step": 29479
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0099575168337448e-05,
+      "loss": 0.3183,
+      "step": 29480
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0099239333031083e-05,
+      "loss": 0.2618,
+      "step": 29481
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0098903497724717e-05,
+      "loss": 0.188,
+      "step": 29482
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.009856766241835e-05,
+      "loss": 0.2245,
+      "step": 29483
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0098231827111987e-05,
+      "loss": 0.34,
+      "step": 29484
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.009789599180562e-05,
+      "loss": 0.2239,
+      "step": 29485
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0097560156499253e-05,
+      "loss": 0.2937,
+      "step": 29486
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0097224321192889e-05,
+      "loss": 0.2824,
+      "step": 29487
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0096888485886522e-05,
+      "loss": 0.4679,
+      "step": 29488
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0096552650580157e-05,
+      "loss": 0.4391,
+      "step": 29489
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0096216815273791e-05,
+      "loss": 0.1344,
+      "step": 29490
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0095880979967426e-05,
+      "loss": 0.4299,
+      "step": 29491
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0095545144661059e-05,
+      "loss": 0.2447,
+      "step": 29492
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0095209309354695e-05,
+      "loss": 0.296,
+      "step": 29493
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0094873474048328e-05,
+      "loss": 0.229,
+      "step": 29494
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0094537638741961e-05,
+      "loss": 0.4472,
+      "step": 29495
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0094201803435597e-05,
+      "loss": 0.3463,
+      "step": 29496
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.009386596812923e-05,
+      "loss": 0.2034,
+      "step": 29497
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0093530132822863e-05,
+      "loss": 0.4052,
+      "step": 29498
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.00931942975165e-05,
+      "loss": 0.2505,
+      "step": 29499
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0092858462210133e-05,
+      "loss": 0.3365,
+      "step": 29500
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0092522626903767e-05,
+      "loss": 0.371,
+      "step": 29501
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0092186791597402e-05,
+      "loss": 0.2004,
+      "step": 29502
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0091850956291037e-05,
+      "loss": 0.3213,
+      "step": 29503
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.009151512098467e-05,
+      "loss": 0.3041,
+      "step": 29504
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0091179285678306e-05,
+      "loss": 0.1746,
+      "step": 29505
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0090843450371939e-05,
+      "loss": 0.353,
+      "step": 29506
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0090507615065572e-05,
+      "loss": 0.3578,
+      "step": 29507
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0090171779759206e-05,
+      "loss": 0.3445,
+      "step": 29508
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0089835944452841e-05,
+      "loss": 0.3105,
+      "step": 29509
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0089500109146476e-05,
+      "loss": 0.4576,
+      "step": 29510
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0089164273840109e-05,
+      "loss": 0.2495,
+      "step": 29511
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0088828438533745e-05,
+      "loss": 0.4633,
+      "step": 29512
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0088492603227378e-05,
+      "loss": 0.1933,
+      "step": 29513
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0088156767921011e-05,
+      "loss": 0.4596,
+      "step": 29514
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0087820932614647e-05,
+      "loss": 0.2745,
+      "step": 29515
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.008748509730828e-05,
+      "loss": 0.4011,
+      "step": 29516
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0087149262001915e-05,
+      "loss": 0.5335,
+      "step": 29517
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.008681342669555e-05,
+      "loss": 0.3128,
+      "step": 29518
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0086477591389182e-05,
+      "loss": 0.2807,
+      "step": 29519
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0086141756082817e-05,
+      "loss": 0.4159,
+      "step": 29520
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0085805920776452e-05,
+      "loss": 0.2926,
+      "step": 29521
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0085470085470086e-05,
+      "loss": 0.3077,
+      "step": 29522
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.008513425016372e-05,
+      "loss": 0.2496,
+      "step": 29523
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0084798414857356e-05,
+      "loss": 0.2196,
+      "step": 29524
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0084462579550989e-05,
+      "loss": 0.324,
+      "step": 29525
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0084126744244622e-05,
+      "loss": 0.2803,
+      "step": 29526
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0083790908938258e-05,
+      "loss": 0.3629,
+      "step": 29527
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0083455073631891e-05,
+      "loss": 0.2349,
+      "step": 29528
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0083119238325526e-05,
+      "loss": 0.3428,
+      "step": 29529
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.008278340301916e-05,
+      "loss": 0.4121,
+      "step": 29530
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0082447567712795e-05,
+      "loss": 0.2983,
+      "step": 29531
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0082111732406428e-05,
+      "loss": 0.4023,
+      "step": 29532
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0081775897100064e-05,
+      "loss": 0.3509,
+      "step": 29533
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0081440061793697e-05,
+      "loss": 0.3091,
+      "step": 29534
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.008110422648733e-05,
+      "loss": 0.1984,
+      "step": 29535
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0080768391180966e-05,
+      "loss": 0.2622,
+      "step": 29536
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.00804325558746e-05,
+      "loss": 0.2714,
+      "step": 29537
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0080096720568234e-05,
+      "loss": 0.2688,
+      "step": 29538
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0079760885261869e-05,
+      "loss": 0.4279,
+      "step": 29539
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0079425049955502e-05,
+      "loss": 0.3116,
+      "step": 29540
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0079089214649136e-05,
+      "loss": 0.2862,
+      "step": 29541
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0078753379342771e-05,
+      "loss": 0.3257,
+      "step": 29542
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0078417544036406e-05,
+      "loss": 0.2993,
+      "step": 29543
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0078081708730038e-05,
+      "loss": 0.3009,
+      "step": 29544
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0077745873423675e-05,
+      "loss": 0.368,
+      "step": 29545
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0077410038117308e-05,
+      "loss": 0.2421,
+      "step": 29546
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.007707420281094e-05,
+      "loss": 0.3109,
+      "step": 29547
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0076738367504577e-05,
+      "loss": 0.3409,
+      "step": 29548
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.007640253219821e-05,
+      "loss": 0.1694,
+      "step": 29549
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0076066696891845e-05,
+      "loss": 0.2621,
+      "step": 29550
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.007573086158548e-05,
+      "loss": 0.1238,
+      "step": 29551
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0075395026279114e-05,
+      "loss": 0.2927,
+      "step": 29552
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0075059190972747e-05,
+      "loss": 0.2586,
+      "step": 29553
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0074723355666383e-05,
+      "loss": 0.3576,
+      "step": 29554
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0074387520360016e-05,
+      "loss": 0.2503,
+      "step": 29555
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.007405168505365e-05,
+      "loss": 0.3213,
+      "step": 29556
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0073715849747286e-05,
+      "loss": 0.3731,
+      "step": 29557
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0073380014440918e-05,
+      "loss": 0.2612,
+      "step": 29558
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0073044179134553e-05,
+      "loss": 0.2424,
+      "step": 29559
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0072708343828188e-05,
+      "loss": 0.3218,
+      "step": 29560
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0072372508521822e-05,
+      "loss": 0.2859,
+      "step": 29561
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0072036673215455e-05,
+      "loss": 0.3792,
+      "step": 29562
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.007170083790909e-05,
+      "loss": 0.2718,
+      "step": 29563
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0071365002602725e-05,
+      "loss": 0.3564,
+      "step": 29564
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0071029167296358e-05,
+      "loss": 0.1802,
+      "step": 29565
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0070693331989994e-05,
+      "loss": 0.247,
+      "step": 29566
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0070357496683627e-05,
+      "loss": 0.326,
+      "step": 29567
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.007002166137726e-05,
+      "loss": 0.2169,
+      "step": 29568
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0069685826070896e-05,
+      "loss": 0.2597,
+      "step": 29569
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0069349990764529e-05,
+      "loss": 0.4896,
+      "step": 29570
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0069014155458164e-05,
+      "loss": 0.326,
+      "step": 29571
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0068678320151798e-05,
+      "loss": 0.2854,
+      "step": 29572
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0068342484845433e-05,
+      "loss": 0.2884,
+      "step": 29573
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0068006649539066e-05,
+      "loss": 0.3481,
+      "step": 29574
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0067670814232702e-05,
+      "loss": 0.1408,
+      "step": 29575
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0067334978926335e-05,
+      "loss": 0.259,
+      "step": 29576
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0066999143619968e-05,
+      "loss": 0.2107,
+      "step": 29577
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0066663308313605e-05,
+      "loss": 0.3732,
+      "step": 29578
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0066327473007238e-05,
+      "loss": 0.2648,
+      "step": 29579
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0065991637700872e-05,
+      "loss": 0.3933,
+      "step": 29580
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0065655802394507e-05,
+      "loss": 0.2263,
+      "step": 29581
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0065319967088142e-05,
+      "loss": 0.2867,
+      "step": 29582
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0064984131781774e-05,
+      "loss": 0.3176,
+      "step": 29583
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.006464829647541e-05,
+      "loss": 0.2572,
+      "step": 29584
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0064312461169044e-05,
+      "loss": 0.2879,
+      "step": 29585
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0063976625862677e-05,
+      "loss": 0.2628,
+      "step": 29586
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0063640790556313e-05,
+      "loss": 0.3364,
+      "step": 29587
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0063304955249946e-05,
+      "loss": 0.2518,
+      "step": 29588
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0062969119943579e-05,
+      "loss": 0.2376,
+      "step": 29589
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0062633284637215e-05,
+      "loss": 0.4466,
+      "step": 29590
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0062297449330848e-05,
+      "loss": 0.3821,
+      "step": 29591
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0061961614024483e-05,
+      "loss": 0.3949,
+      "step": 29592
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0061625778718118e-05,
+      "loss": 0.3947,
+      "step": 29593
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0061289943411752e-05,
+      "loss": 0.3782,
+      "step": 29594
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0060954108105385e-05,
+      "loss": 0.277,
+      "step": 29595
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0060618272799022e-05,
+      "loss": 0.3661,
+      "step": 29596
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0060282437492654e-05,
+      "loss": 0.3004,
+      "step": 29597
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0059946602186287e-05,
+      "loss": 0.2705,
+      "step": 29598
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0059610766879924e-05,
+      "loss": 0.2795,
+      "step": 29599
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0059274931573557e-05,
+      "loss": 0.383,
+      "step": 29600
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0058939096267191e-05,
+      "loss": 0.2179,
+      "step": 29601
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0058603260960826e-05,
+      "loss": 0.1966,
+      "step": 29602
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.005826742565446e-05,
+      "loss": 0.2075,
+      "step": 29603
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0057931590348094e-05,
+      "loss": 0.4265,
+      "step": 29604
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.005759575504173e-05,
+      "loss": 0.4222,
+      "step": 29605
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0057259919735363e-05,
+      "loss": 0.4236,
+      "step": 29606
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0056924084428996e-05,
+      "loss": 0.2539,
+      "step": 29607
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0056588249122632e-05,
+      "loss": 0.2986,
+      "step": 29608
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0056252413816265e-05,
+      "loss": 0.1159,
+      "step": 29609
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0055916578509898e-05,
+      "loss": 0.1938,
+      "step": 29610
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0055580743203534e-05,
+      "loss": 0.1709,
+      "step": 29611
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0055244907897167e-05,
+      "loss": 0.3871,
+      "step": 29612
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0054909072590802e-05,
+      "loss": 0.349,
+      "step": 29613
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0054573237284437e-05,
+      "loss": 0.2457,
+      "step": 29614
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0054237401978071e-05,
+      "loss": 0.2354,
+      "step": 29615
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0053901566671704e-05,
+      "loss": 0.3693,
+      "step": 29616
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.005356573136534e-05,
+      "loss": 0.231,
+      "step": 29617
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0053229896058974e-05,
+      "loss": 0.3314,
+      "step": 29618
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0052894060752607e-05,
+      "loss": 0.3558,
+      "step": 29619
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0052558225446243e-05,
+      "loss": 0.3358,
+      "step": 29620
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0052222390139876e-05,
+      "loss": 0.2801,
+      "step": 29621
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.005188655483351e-05,
+      "loss": 0.3604,
+      "step": 29622
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0051550719527145e-05,
+      "loss": 0.1704,
+      "step": 29623
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.005121488422078e-05,
+      "loss": 0.3067,
+      "step": 29624
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0050879048914413e-05,
+      "loss": 0.3884,
+      "step": 29625
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0050543213608049e-05,
+      "loss": 0.2901,
+      "step": 29626
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0050207378301682e-05,
+      "loss": 0.1657,
+      "step": 29627
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0049871542995315e-05,
+      "loss": 0.338,
+      "step": 29628
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0049535707688951e-05,
+      "loss": 0.5435,
+      "step": 29629
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0049199872382584e-05,
+      "loss": 0.3822,
+      "step": 29630
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0048864037076217e-05,
+      "loss": 0.2239,
+      "step": 29631
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0048528201769854e-05,
+      "loss": 0.1738,
+      "step": 29632
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0048192366463487e-05,
+      "loss": 0.3671,
+      "step": 29633
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0047856531157121e-05,
+      "loss": 0.3192,
+      "step": 29634
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0047520695850756e-05,
+      "loss": 0.2079,
+      "step": 29635
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.004718486054439e-05,
+      "loss": 0.3324,
+      "step": 29636
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0046849025238023e-05,
+      "loss": 0.1431,
+      "step": 29637
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.004651318993166e-05,
+      "loss": 0.3845,
+      "step": 29638
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0046177354625293e-05,
+      "loss": 0.5063,
+      "step": 29639
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0045841519318926e-05,
+      "loss": 0.2362,
+      "step": 29640
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0045505684012562e-05,
+      "loss": 0.3448,
+      "step": 29641
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0045169848706195e-05,
+      "loss": 0.2432,
+      "step": 29642
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.004483401339983e-05,
+      "loss": 0.3752,
+      "step": 29643
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0044498178093464e-05,
+      "loss": 0.5245,
+      "step": 29644
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0044162342787099e-05,
+      "loss": 0.3233,
+      "step": 29645
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0043826507480732e-05,
+      "loss": 0.3714,
+      "step": 29646
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0043490672174368e-05,
+      "loss": 0.2871,
+      "step": 29647
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0043154836868001e-05,
+      "loss": 0.2537,
+      "step": 29648
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0042819001561634e-05,
+      "loss": 0.3468,
+      "step": 29649
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.004248316625527e-05,
+      "loss": 0.2884,
+      "step": 29650
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0042147330948903e-05,
+      "loss": 0.2821,
+      "step": 29651
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0041811495642538e-05,
+      "loss": 0.2374,
+      "step": 29652
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0041475660336173e-05,
+      "loss": 0.2171,
+      "step": 29653
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0041139825029806e-05,
+      "loss": 0.2254,
+      "step": 29654
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.004080398972344e-05,
+      "loss": 0.4116,
+      "step": 29655
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0040468154417075e-05,
+      "loss": 0.3622,
+      "step": 29656
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.004013231911071e-05,
+      "loss": 0.3931,
+      "step": 29657
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0039796483804343e-05,
+      "loss": 0.2753,
+      "step": 29658
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0039460648497979e-05,
+      "loss": 0.3685,
+      "step": 29659
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0039124813191612e-05,
+      "loss": 0.2488,
+      "step": 29660
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0038788977885245e-05,
+      "loss": 0.2806,
+      "step": 29661
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0038453142578881e-05,
+      "loss": 0.249,
+      "step": 29662
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0038117307272514e-05,
+      "loss": 0.1853,
+      "step": 29663
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0037781471966149e-05,
+      "loss": 0.2954,
+      "step": 29664
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0037445636659783e-05,
+      "loss": 0.2542,
+      "step": 29665
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0037109801353418e-05,
+      "loss": 0.2417,
+      "step": 29666
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0036773966047051e-05,
+      "loss": 0.2231,
+      "step": 29667
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0036438130740687e-05,
+      "loss": 0.3356,
+      "step": 29668
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.003610229543432e-05,
+      "loss": 0.3997,
+      "step": 29669
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0035766460127953e-05,
+      "loss": 0.3127,
+      "step": 29670
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.003543062482159e-05,
+      "loss": 0.4808,
+      "step": 29671
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0035094789515223e-05,
+      "loss": 0.3041,
+      "step": 29672
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0034758954208857e-05,
+      "loss": 0.2468,
+      "step": 29673
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0034423118902492e-05,
+      "loss": 0.2689,
+      "step": 29674
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0034087283596125e-05,
+      "loss": 0.2434,
+      "step": 29675
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.003375144828976e-05,
+      "loss": 0.2204,
+      "step": 29676
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 1.0033415612983394e-05,
+      "loss": 0.4925,
+      "step": 29677
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0033079777677029e-05,
+      "loss": 0.2507,
+      "step": 29678
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0032743942370662e-05,
+      "loss": 0.2419,
+      "step": 29679
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0032408107064298e-05,
+      "loss": 0.4326,
+      "step": 29680
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0032072271757931e-05,
+      "loss": 0.1612,
+      "step": 29681
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0031736436451564e-05,
+      "loss": 0.3176,
+      "step": 29682
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.00314006011452e-05,
+      "loss": 0.281,
+      "step": 29683
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0031064765838833e-05,
+      "loss": 0.3643,
+      "step": 29684
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0030728930532468e-05,
+      "loss": 0.2769,
+      "step": 29685
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0030393095226102e-05,
+      "loss": 0.2441,
+      "step": 29686
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0030057259919737e-05,
+      "loss": 0.2341,
+      "step": 29687
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.002972142461337e-05,
+      "loss": 0.1297,
+      "step": 29688
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0029385589307003e-05,
+      "loss": 0.2571,
+      "step": 29689
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.002904975400064e-05,
+      "loss": 0.3026,
+      "step": 29690
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0028713918694272e-05,
+      "loss": 0.2272,
+      "step": 29691
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0028378083387907e-05,
+      "loss": 0.2626,
+      "step": 29692
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0028042248081542e-05,
+      "loss": 0.2187,
+      "step": 29693
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0027706412775176e-05,
+      "loss": 0.3903,
+      "step": 29694
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.002737057746881e-05,
+      "loss": 0.208,
+      "step": 29695
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0027034742162446e-05,
+      "loss": 0.3022,
+      "step": 29696
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0026698906856079e-05,
+      "loss": 0.3459,
+      "step": 29697
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0026363071549711e-05,
+      "loss": 0.1765,
+      "step": 29698
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0026027236243348e-05,
+      "loss": 0.407,
+      "step": 29699
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.002569140093698e-05,
+      "loss": 0.4739,
+      "step": 29700
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0025355565630614e-05,
+      "loss": 0.3328,
+      "step": 29701
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.002501973032425e-05,
+      "loss": 0.2506,
+      "step": 29702
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0024683895017883e-05,
+      "loss": 0.3535,
+      "step": 29703
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0024348059711518e-05,
+      "loss": 0.4059,
+      "step": 29704
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0024012224405152e-05,
+      "loss": 0.2581,
+      "step": 29705
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0023676389098787e-05,
+      "loss": 0.2579,
+      "step": 29706
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.002334055379242e-05,
+      "loss": 0.2172,
+      "step": 29707
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0023004718486056e-05,
+      "loss": 0.3287,
+      "step": 29708
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.002266888317969e-05,
+      "loss": 0.2332,
+      "step": 29709
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0022333047873322e-05,
+      "loss": 0.2567,
+      "step": 29710
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0021997212566959e-05,
+      "loss": 0.2152,
+      "step": 29711
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0021661377260591e-05,
+      "loss": 0.1903,
+      "step": 29712
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0021325541954226e-05,
+      "loss": 0.2154,
+      "step": 29713
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.002098970664786e-05,
+      "loss": 0.2079,
+      "step": 29714
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0020653871341495e-05,
+      "loss": 0.37,
+      "step": 29715
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0020318036035128e-05,
+      "loss": 0.4047,
+      "step": 29716
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0019982200728765e-05,
+      "loss": 0.3413,
+      "step": 29717
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0019646365422398e-05,
+      "loss": 0.4635,
+      "step": 29718
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.001931053011603e-05,
+      "loss": 0.4202,
+      "step": 29719
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0018974694809667e-05,
+      "loss": 0.1605,
+      "step": 29720
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.00186388595033e-05,
+      "loss": 0.313,
+      "step": 29721
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0018303024196933e-05,
+      "loss": 0.1858,
+      "step": 29722
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.001796718889057e-05,
+      "loss": 0.3501,
+      "step": 29723
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0017631353584202e-05,
+      "loss": 0.3243,
+      "step": 29724
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0017295518277837e-05,
+      "loss": 0.2508,
+      "step": 29725
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0016959682971471e-05,
+      "loss": 0.2385,
+      "step": 29726
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0016623847665106e-05,
+      "loss": 0.2609,
+      "step": 29727
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0016288012358739e-05,
+      "loss": 0.3928,
+      "step": 29728
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0015952177052375e-05,
+      "loss": 0.3052,
+      "step": 29729
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0015616341746008e-05,
+      "loss": 0.2696,
+      "step": 29730
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0015280506439641e-05,
+      "loss": 0.3888,
+      "step": 29731
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0014944671133278e-05,
+      "loss": 0.1848,
+      "step": 29732
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.001460883582691e-05,
+      "loss": 0.4109,
+      "step": 29733
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0014273000520545e-05,
+      "loss": 0.326,
+      "step": 29734
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.001393716521418e-05,
+      "loss": 0.3221,
+      "step": 29735
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0013601329907815e-05,
+      "loss": 0.3666,
+      "step": 29736
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0013265494601447e-05,
+      "loss": 0.4026,
+      "step": 29737
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0012929659295084e-05,
+      "loss": 0.2678,
+      "step": 29738
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0012593823988717e-05,
+      "loss": 0.2316,
+      "step": 29739
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.001225798868235e-05,
+      "loss": 0.249,
+      "step": 29740
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0011922153375986e-05,
+      "loss": 0.2113,
+      "step": 29741
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0011586318069619e-05,
+      "loss": 0.3325,
+      "step": 29742
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0011250482763254e-05,
+      "loss": 0.2175,
+      "step": 29743
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0010914647456888e-05,
+      "loss": 0.2822,
+      "step": 29744
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0010578812150521e-05,
+      "loss": 0.6173,
+      "step": 29745
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0010242976844156e-05,
+      "loss": 0.402,
+      "step": 29746
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.000990714153779e-05,
+      "loss": 0.3327,
+      "step": 29747
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0009571306231425e-05,
+      "loss": 0.3124,
+      "step": 29748
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0009235470925058e-05,
+      "loss": 0.3298,
+      "step": 29749
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0008899635618695e-05,
+      "loss": 0.2138,
+      "step": 29750
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0008563800312327e-05,
+      "loss": 0.4096,
+      "step": 29751
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.000822796500596e-05,
+      "loss": 0.1231,
+      "step": 29752
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0007892129699597e-05,
+      "loss": 0.2046,
+      "step": 29753
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.000755629439323e-05,
+      "loss": 0.4135,
+      "step": 29754
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0007220459086864e-05,
+      "loss": 0.1693,
+      "step": 29755
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0006884623780499e-05,
+      "loss": 0.2543,
+      "step": 29756
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0006548788474134e-05,
+      "loss": 0.2667,
+      "step": 29757
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0006212953167767e-05,
+      "loss": 0.3144,
+      "step": 29758
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0005877117861403e-05,
+      "loss": 0.2613,
+      "step": 29759
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0005541282555036e-05,
+      "loss": 0.2009,
+      "step": 29760
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0005205447248669e-05,
+      "loss": 0.1865,
+      "step": 29761
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0004869611942305e-05,
+      "loss": 0.1971,
+      "step": 29762
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0004533776635938e-05,
+      "loss": 0.3309,
+      "step": 29763
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0004197941329573e-05,
+      "loss": 0.3312,
+      "step": 29764
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0003862106023207e-05,
+      "loss": 0.3842,
+      "step": 29765
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.000352627071684e-05,
+      "loss": 0.2222,
+      "step": 29766
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0003190435410475e-05,
+      "loss": 0.1907,
+      "step": 29767
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.000285460010411e-05,
+      "loss": 0.2141,
+      "step": 29768
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0002518764797744e-05,
+      "loss": 0.3073,
+      "step": 29769
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0002182929491377e-05,
+      "loss": 0.2943,
+      "step": 29770
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0001847094185014e-05,
+      "loss": 0.5908,
+      "step": 29771
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0001511258878647e-05,
+      "loss": 0.1622,
+      "step": 29772
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.000117542357228e-05,
+      "loss": 0.2723,
+      "step": 29773
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0000839588265916e-05,
+      "loss": 0.3618,
+      "step": 29774
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0000503752959549e-05,
+      "loss": 0.4524,
+      "step": 29775
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.0000167917653183e-05,
+      "loss": 0.2439,
+      "step": 29776
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.999832082346818e-06,
+      "loss": 0.2346,
+      "step": 29777
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.999496247040453e-06,
+      "loss": 0.1094,
+      "step": 29778
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.999160411734087e-06,
+      "loss": 0.2287,
+      "step": 29779
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.99882457642772e-06,
+      "loss": 0.1509,
+      "step": 29780
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.998488741121355e-06,
+      "loss": 0.3453,
+      "step": 29781
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.99815290581499e-06,
+      "loss": 0.435,
+      "step": 29782
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.997817070508623e-06,
+      "loss": 0.376,
+      "step": 29783
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.997481235202257e-06,
+      "loss": 0.442,
+      "step": 29784
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.997145399895892e-06,
+      "loss": 0.2551,
+      "step": 29785
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.996809564589527e-06,
+      "loss": 0.173,
+      "step": 29786
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.996473729283161e-06,
+      "loss": 0.156,
+      "step": 29787
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.996137893976794e-06,
+      "loss": 0.2462,
+      "step": 29788
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.995802058670429e-06,
+      "loss": 0.1896,
+      "step": 29789
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.995466223364063e-06,
+      "loss": 0.3779,
+      "step": 29790
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.995130388057698e-06,
+      "loss": 0.2579,
+      "step": 29791
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.994794552751331e-06,
+      "loss": 0.3308,
+      "step": 29792
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.994458717444966e-06,
+      "loss": 0.3371,
+      "step": 29793
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.9941228821386e-06,
+      "loss": 0.2226,
+      "step": 29794
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.993787046832233e-06,
+      "loss": 0.2728,
+      "step": 29795
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.993451211525868e-06,
+      "loss": 0.2641,
+      "step": 29796
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.993115376219503e-06,
+      "loss": 0.2239,
+      "step": 29797
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.992779540913137e-06,
+      "loss": 0.3143,
+      "step": 29798
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.992443705606772e-06,
+      "loss": 0.299,
+      "step": 29799
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.992107870300407e-06,
+      "loss": 0.4287,
+      "step": 29800
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.99177203499404e-06,
+      "loss": 0.1385,
+      "step": 29801
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.991436199687674e-06,
+      "loss": 0.2946,
+      "step": 29802
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.991100364381309e-06,
+      "loss": 0.3105,
+      "step": 29803
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.990764529074942e-06,
+      "loss": 0.3088,
+      "step": 29804
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.990428693768576e-06,
+      "loss": 0.1526,
+      "step": 29805
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.990092858462211e-06,
+      "loss": 0.4089,
+      "step": 29806
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.989757023155846e-06,
+      "loss": 0.3965,
+      "step": 29807
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.98942118784948e-06,
+      "loss": 0.4371,
+      "step": 29808
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.989085352543115e-06,
+      "loss": 0.1307,
+      "step": 29809
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.988749517236748e-06,
+      "loss": 0.31,
+      "step": 29810
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.988413681930383e-06,
+      "loss": 0.29,
+      "step": 29811
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.988077846624017e-06,
+      "loss": 0.3924,
+      "step": 29812
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.98774201131765e-06,
+      "loss": 0.2538,
+      "step": 29813
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.987406176011285e-06,
+      "loss": 0.284,
+      "step": 29814
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.98707034070492e-06,
+      "loss": 0.4885,
+      "step": 29815
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.986734505398552e-06,
+      "loss": 0.1485,
+      "step": 29816
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.986398670092187e-06,
+      "loss": 0.2835,
+      "step": 29817
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.986062834785822e-06,
+      "loss": 0.3549,
+      "step": 29818
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.985726999479456e-06,
+      "loss": 0.3341,
+      "step": 29819
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.985391164173091e-06,
+      "loss": 0.382,
+      "step": 29820
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.985055328866726e-06,
+      "loss": 0.2299,
+      "step": 29821
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.984719493560359e-06,
+      "loss": 0.1981,
+      "step": 29822
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.984383658253993e-06,
+      "loss": 0.3856,
+      "step": 29823
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.984047822947626e-06,
+      "loss": 0.2684,
+      "step": 29824
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.983711987641261e-06,
+      "loss": 0.2266,
+      "step": 29825
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.983376152334896e-06,
+      "loss": 0.388,
+      "step": 29826
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.98304031702853e-06,
+      "loss": 0.3545,
+      "step": 29827
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.982704481722165e-06,
+      "loss": 0.1867,
+      "step": 29828
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.9823686464158e-06,
+      "loss": 0.1996,
+      "step": 29829
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.982032811109432e-06,
+      "loss": 0.3135,
+      "step": 29830
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.981696975803067e-06,
+      "loss": 0.3702,
+      "step": 29831
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.981361140496702e-06,
+      "loss": 0.3936,
+      "step": 29832
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.981025305190335e-06,
+      "loss": 0.1394,
+      "step": 29833
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.98068946988397e-06,
+      "loss": 0.317,
+      "step": 29834
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.980353634577604e-06,
+      "loss": 0.281,
+      "step": 29835
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.980017799271237e-06,
+      "loss": 0.2836,
+      "step": 29836
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.979681963964872e-06,
+      "loss": 0.2598,
+      "step": 29837
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.979346128658506e-06,
+      "loss": 0.1734,
+      "step": 29838
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.979010293352141e-06,
+      "loss": 0.2324,
+      "step": 29839
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.978674458045776e-06,
+      "loss": 0.2752,
+      "step": 29840
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.97833862273941e-06,
+      "loss": 0.2981,
+      "step": 29841
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.978002787433043e-06,
+      "loss": 0.2406,
+      "step": 29842
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.977666952126678e-06,
+      "loss": 0.3602,
+      "step": 29843
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.977331116820312e-06,
+      "loss": 0.3588,
+      "step": 29844
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.976995281513945e-06,
+      "loss": 0.2663,
+      "step": 29845
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.97665944620758e-06,
+      "loss": 0.2971,
+      "step": 29846
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.976323610901215e-06,
+      "loss": 0.2654,
+      "step": 29847
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.97598777559485e-06,
+      "loss": 0.1981,
+      "step": 29848
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.975651940288484e-06,
+      "loss": 0.2102,
+      "step": 29849
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.975316104982119e-06,
+      "loss": 0.4859,
+      "step": 29850
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.974980269675752e-06,
+      "loss": 0.2268,
+      "step": 29851
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.974644434369386e-06,
+      "loss": 0.3458,
+      "step": 29852
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.974308599063021e-06,
+      "loss": 0.3213,
+      "step": 29853
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.973972763756654e-06,
+      "loss": 0.3564,
+      "step": 29854
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.973636928450288e-06,
+      "loss": 0.2397,
+      "step": 29855
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.973301093143923e-06,
+      "loss": 0.2486,
+      "step": 29856
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.972965257837556e-06,
+      "loss": 0.4256,
+      "step": 29857
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.97262942253119e-06,
+      "loss": 0.3516,
+      "step": 29858
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.972293587224825e-06,
+      "loss": 0.2268,
+      "step": 29859
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.97195775191846e-06,
+      "loss": 0.4131,
+      "step": 29860
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.971621916612095e-06,
+      "loss": 0.2843,
+      "step": 29861
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.97128608130573e-06,
+      "loss": 0.3844,
+      "step": 29862
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.970950245999362e-06,
+      "loss": 0.3364,
+      "step": 29863
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.970614410692997e-06,
+      "loss": 0.3683,
+      "step": 29864
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.970278575386632e-06,
+      "loss": 0.3097,
+      "step": 29865
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.969942740080264e-06,
+      "loss": 0.2958,
+      "step": 29866
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.969606904773899e-06,
+      "loss": 0.3208,
+      "step": 29867
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.969271069467534e-06,
+      "loss": 0.2566,
+      "step": 29868
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.968935234161168e-06,
+      "loss": 0.2704,
+      "step": 29869
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.968599398854803e-06,
+      "loss": 0.2869,
+      "step": 29870
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.968263563548438e-06,
+      "loss": 0.2354,
+      "step": 29871
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.96792772824207e-06,
+      "loss": 0.2628,
+      "step": 29872
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.967591892935705e-06,
+      "loss": 0.289,
+      "step": 29873
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.96725605762934e-06,
+      "loss": 0.3651,
+      "step": 29874
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 9.966920222322973e-06,
+      "loss": 0.321,
+      "step": 29875
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.966584387016608e-06,
+      "loss": 0.4105,
+      "step": 29876
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.966248551710242e-06,
+      "loss": 0.3026,
+      "step": 29877
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.965912716403877e-06,
+      "loss": 0.1832,
+      "step": 29878
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.96557688109751e-06,
+      "loss": 0.3376,
+      "step": 29879
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.965241045791144e-06,
+      "loss": 0.3414,
+      "step": 29880
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.964905210484779e-06,
+      "loss": 0.4714,
+      "step": 29881
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.964569375178414e-06,
+      "loss": 0.3906,
+      "step": 29882
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.964233539872048e-06,
+      "loss": 0.291,
+      "step": 29883
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.963897704565681e-06,
+      "loss": 0.231,
+      "step": 29884
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.963561869259316e-06,
+      "loss": 0.1869,
+      "step": 29885
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.96322603395295e-06,
+      "loss": 0.3119,
+      "step": 29886
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.962890198646584e-06,
+      "loss": 0.5021,
+      "step": 29887
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.962554363340218e-06,
+      "loss": 0.1608,
+      "step": 29888
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.962218528033853e-06,
+      "loss": 0.2893,
+      "step": 29889
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.961882692727488e-06,
+      "loss": 0.2293,
+      "step": 29890
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.961546857421122e-06,
+      "loss": 0.2265,
+      "step": 29891
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.961211022114757e-06,
+      "loss": 0.3159,
+      "step": 29892
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.96087518680839e-06,
+      "loss": 0.2841,
+      "step": 29893
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.960539351502024e-06,
+      "loss": 0.2714,
+      "step": 29894
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.960203516195659e-06,
+      "loss": 0.1581,
+      "step": 29895
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.959867680889292e-06,
+      "loss": 0.2259,
+      "step": 29896
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.959531845582927e-06,
+      "loss": 0.2741,
+      "step": 29897
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.959196010276561e-06,
+      "loss": 0.324,
+      "step": 29898
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.958860174970196e-06,
+      "loss": 0.2941,
+      "step": 29899
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.95852433966383e-06,
+      "loss": 0.4002,
+      "step": 29900
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.958188504357464e-06,
+      "loss": 0.2488,
+      "step": 29901
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.957852669051098e-06,
+      "loss": 0.1535,
+      "step": 29902
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.957516833744733e-06,
+      "loss": 0.3395,
+      "step": 29903
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.957180998438368e-06,
+      "loss": 0.1736,
+      "step": 29904
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.956845163132e-06,
+      "loss": 0.3152,
+      "step": 29905
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.956509327825635e-06,
+      "loss": 0.3433,
+      "step": 29906
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.95617349251927e-06,
+      "loss": 0.1424,
+      "step": 29907
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.955837657212903e-06,
+      "loss": 0.2714,
+      "step": 29908
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.955501821906537e-06,
+      "loss": 0.2731,
+      "step": 29909
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.955165986600172e-06,
+      "loss": 0.3401,
+      "step": 29910
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.954830151293807e-06,
+      "loss": 0.208,
+      "step": 29911
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.954494315987441e-06,
+      "loss": 0.3268,
+      "step": 29912
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.954158480681074e-06,
+      "loss": 0.1781,
+      "step": 29913
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.953822645374709e-06,
+      "loss": 0.2789,
+      "step": 29914
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.953486810068344e-06,
+      "loss": 0.1972,
+      "step": 29915
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.953150974761977e-06,
+      "loss": 0.2608,
+      "step": 29916
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.952815139455611e-06,
+      "loss": 0.3242,
+      "step": 29917
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.952479304149246e-06,
+      "loss": 0.386,
+      "step": 29918
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.95214346884288e-06,
+      "loss": 0.2781,
+      "step": 29919
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.951807633536515e-06,
+      "loss": 0.3269,
+      "step": 29920
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.95147179823015e-06,
+      "loss": 0.3033,
+      "step": 29921
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.951135962923783e-06,
+      "loss": 0.3268,
+      "step": 29922
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.950800127617417e-06,
+      "loss": 0.2642,
+      "step": 29923
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.950464292311052e-06,
+      "loss": 0.2655,
+      "step": 29924
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.950128457004685e-06,
+      "loss": 0.2902,
+      "step": 29925
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.94979262169832e-06,
+      "loss": 0.2827,
+      "step": 29926
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.949456786391954e-06,
+      "loss": 0.1607,
+      "step": 29927
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.949120951085587e-06,
+      "loss": 0.4675,
+      "step": 29928
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.948785115779222e-06,
+      "loss": 0.2509,
+      "step": 29929
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.948449280472857e-06,
+      "loss": 0.2588,
+      "step": 29930
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.948113445166491e-06,
+      "loss": 0.27,
+      "step": 29931
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.947777609860126e-06,
+      "loss": 0.369,
+      "step": 29932
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.94744177455376e-06,
+      "loss": 0.2724,
+      "step": 29933
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.947105939247393e-06,
+      "loss": 0.4043,
+      "step": 29934
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.946770103941028e-06,
+      "loss": 0.3007,
+      "step": 29935
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.946434268634663e-06,
+      "loss": 0.3417,
+      "step": 29936
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.946098433328296e-06,
+      "loss": 0.2813,
+      "step": 29937
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.94576259802193e-06,
+      "loss": 0.3001,
+      "step": 29938
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.945426762715565e-06,
+      "loss": 0.5934,
+      "step": 29939
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.9450909274092e-06,
+      "loss": 0.4194,
+      "step": 29940
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.944755092102834e-06,
+      "loss": 0.1459,
+      "step": 29941
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.944419256796469e-06,
+      "loss": 0.3464,
+      "step": 29942
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.944083421490102e-06,
+      "loss": 0.2218,
+      "step": 29943
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.943747586183736e-06,
+      "loss": 0.345,
+      "step": 29944
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.943411750877371e-06,
+      "loss": 0.3257,
+      "step": 29945
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.943075915571004e-06,
+      "loss": 0.3529,
+      "step": 29946
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.942740080264639e-06,
+      "loss": 0.3927,
+      "step": 29947
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.942404244958273e-06,
+      "loss": 0.2678,
+      "step": 29948
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.942068409651906e-06,
+      "loss": 0.232,
+      "step": 29949
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.941732574345541e-06,
+      "loss": 0.235,
+      "step": 29950
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.941396739039176e-06,
+      "loss": 0.4434,
+      "step": 29951
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.94106090373281e-06,
+      "loss": 0.3328,
+      "step": 29952
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.940725068426445e-06,
+      "loss": 0.3429,
+      "step": 29953
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.94038923312008e-06,
+      "loss": 0.265,
+      "step": 29954
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.940053397813713e-06,
+      "loss": 0.3917,
+      "step": 29955
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.939717562507347e-06,
+      "loss": 0.2463,
+      "step": 29956
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.939381727200982e-06,
+      "loss": 0.2813,
+      "step": 29957
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.939045891894615e-06,
+      "loss": 0.2759,
+      "step": 29958
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.93871005658825e-06,
+      "loss": 0.3982,
+      "step": 29959
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.938374221281884e-06,
+      "loss": 0.3845,
+      "step": 29960
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.938038385975519e-06,
+      "loss": 0.2462,
+      "step": 29961
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.937702550669153e-06,
+      "loss": 0.2278,
+      "step": 29962
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.937366715362788e-06,
+      "loss": 0.4631,
+      "step": 29963
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.937030880056421e-06,
+      "loss": 0.2388,
+      "step": 29964
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.936695044750056e-06,
+      "loss": 0.1838,
+      "step": 29965
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.93635920944369e-06,
+      "loss": 0.2188,
+      "step": 29966
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.936023374137323e-06,
+      "loss": 0.3215,
+      "step": 29967
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.935687538830958e-06,
+      "loss": 0.2411,
+      "step": 29968
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.935351703524593e-06,
+      "loss": 0.3174,
+      "step": 29969
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.935015868218225e-06,
+      "loss": 0.2894,
+      "step": 29970
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.93468003291186e-06,
+      "loss": 0.2168,
+      "step": 29971
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.934344197605495e-06,
+      "loss": 0.3352,
+      "step": 29972
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.93400836229913e-06,
+      "loss": 0.1996,
+      "step": 29973
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.933672526992764e-06,
+      "loss": 0.3031,
+      "step": 29974
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.933336691686399e-06,
+      "loss": 0.209,
+      "step": 29975
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.933000856380032e-06,
+      "loss": 0.3237,
+      "step": 29976
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.932665021073666e-06,
+      "loss": 0.2345,
+      "step": 29977
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.932329185767301e-06,
+      "loss": 0.3287,
+      "step": 29978
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.931993350460934e-06,
+      "loss": 0.0997,
+      "step": 29979
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.931657515154569e-06,
+      "loss": 0.3208,
+      "step": 29980
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.931321679848203e-06,
+      "loss": 0.3514,
+      "step": 29981
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.930985844541838e-06,
+      "loss": 0.4146,
+      "step": 29982
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.930650009235472e-06,
+      "loss": 0.3896,
+      "step": 29983
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.930314173929107e-06,
+      "loss": 0.3863,
+      "step": 29984
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.92997833862274e-06,
+      "loss": 0.2376,
+      "step": 29985
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.929642503316375e-06,
+      "loss": 0.3598,
+      "step": 29986
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.92930666801001e-06,
+      "loss": 0.2969,
+      "step": 29987
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.928970832703642e-06,
+      "loss": 0.5086,
+      "step": 29988
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.928634997397277e-06,
+      "loss": 0.2793,
+      "step": 29989
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.928299162090912e-06,
+      "loss": 0.199,
+      "step": 29990
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.927963326784546e-06,
+      "loss": 0.3203,
+      "step": 29991
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.92762749147818e-06,
+      "loss": 0.2718,
+      "step": 29992
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.927291656171814e-06,
+      "loss": 0.2982,
+      "step": 29993
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.926955820865449e-06,
+      "loss": 0.2688,
+      "step": 29994
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.926619985559083e-06,
+      "loss": 0.3669,
+      "step": 29995
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.926284150252718e-06,
+      "loss": 0.3346,
+      "step": 29996
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.92594831494635e-06,
+      "loss": 0.3238,
+      "step": 29997
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.925612479639985e-06,
+      "loss": 0.526,
+      "step": 29998
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.92527664433362e-06,
+      "loss": 0.2626,
+      "step": 29999
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.924940809027253e-06,
+      "loss": 0.2664,
+      "step": 30000
+    },
+    {
+      "epoch": 1.51,
+      "eval_accuracy": 0.8819452097062129,
+      "eval_f1_microaverage": 0.8819452097062129,
+      "eval_loss": 0.3089483976364136,
+      "eval_precision_macroaverage": 0.767376662032236,
+      "eval_precision_microaverage": 0.8819452097062129,
+      "eval_recall_macroaverage": 0.6765430968461141,
+      "eval_recall_microaverage": 0.8819452097062129,
+      "eval_runtime": 1173.19,
+      "eval_samples_per_second": 232.049,
+      "eval_steps_per_second": 3.626,
+      "step": 30000
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.924604973720888e-06,
+      "loss": 0.2701,
+      "step": 30001
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.924269138414522e-06,
+      "loss": 0.4299,
+      "step": 30002
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.923933303108157e-06,
+      "loss": 0.2294,
+      "step": 30003
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.923597467801792e-06,
+      "loss": 0.3837,
+      "step": 30004
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.923261632495425e-06,
+      "loss": 0.4004,
+      "step": 30005
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.92292579718906e-06,
+      "loss": 0.3121,
+      "step": 30006
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.922589961882694e-06,
+      "loss": 0.2575,
+      "step": 30007
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.922254126576327e-06,
+      "loss": 0.2775,
+      "step": 30008
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.921918291269961e-06,
+      "loss": 0.3247,
+      "step": 30009
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.921582455963596e-06,
+      "loss": 0.2882,
+      "step": 30010
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.92124662065723e-06,
+      "loss": 0.3444,
+      "step": 30011
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.920910785350865e-06,
+      "loss": 0.2442,
+      "step": 30012
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.9205749500445e-06,
+      "loss": 0.257,
+      "step": 30013
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.920239114738133e-06,
+      "loss": 0.332,
+      "step": 30014
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.919903279431768e-06,
+      "loss": 0.245,
+      "step": 30015
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.919567444125402e-06,
+      "loss": 0.233,
+      "step": 30016
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.919231608819035e-06,
+      "loss": 0.2401,
+      "step": 30017
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.91889577351267e-06,
+      "loss": 0.4203,
+      "step": 30018
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.918559938206305e-06,
+      "loss": 0.1811,
+      "step": 30019
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.918224102899938e-06,
+      "loss": 0.3064,
+      "step": 30020
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.917888267593572e-06,
+      "loss": 0.3805,
+      "step": 30021
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.917552432287207e-06,
+      "loss": 0.4216,
+      "step": 30022
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.917216596980841e-06,
+      "loss": 0.1862,
+      "step": 30023
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.916880761674476e-06,
+      "loss": 0.1975,
+      "step": 30024
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.91654492636811e-06,
+      "loss": 0.2196,
+      "step": 30025
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.916209091061744e-06,
+      "loss": 0.2449,
+      "step": 30026
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.915873255755378e-06,
+      "loss": 0.3764,
+      "step": 30027
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.915537420449013e-06,
+      "loss": 0.208,
+      "step": 30028
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.915201585142646e-06,
+      "loss": 0.4403,
+      "step": 30029
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.91486574983628e-06,
+      "loss": 0.3374,
+      "step": 30030
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.914529914529915e-06,
+      "loss": 0.3155,
+      "step": 30031
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.91419407922355e-06,
+      "loss": 0.2254,
+      "step": 30032
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.913858243917185e-06,
+      "loss": 0.3106,
+      "step": 30033
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.91352240861082e-06,
+      "loss": 0.2647,
+      "step": 30034
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.913186573304452e-06,
+      "loss": 0.2651,
+      "step": 30035
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.912850737998087e-06,
+      "loss": 0.2363,
+      "step": 30036
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.912514902691721e-06,
+      "loss": 0.2238,
+      "step": 30037
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.912179067385354e-06,
+      "loss": 0.2754,
+      "step": 30038
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.911843232078989e-06,
+      "loss": 0.3201,
+      "step": 30039
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.911507396772624e-06,
+      "loss": 0.3192,
+      "step": 30040
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.911171561466257e-06,
+      "loss": 0.1975,
+      "step": 30041
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.910835726159891e-06,
+      "loss": 0.3732,
+      "step": 30042
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.910499890853526e-06,
+      "loss": 0.2913,
+      "step": 30043
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.91016405554716e-06,
+      "loss": 0.4344,
+      "step": 30044
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.909828220240795e-06,
+      "loss": 0.4787,
+      "step": 30045
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.90949238493443e-06,
+      "loss": 0.3735,
+      "step": 30046
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.909156549628063e-06,
+      "loss": 0.2283,
+      "step": 30047
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.908820714321697e-06,
+      "loss": 0.2047,
+      "step": 30048
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.908484879015332e-06,
+      "loss": 0.1852,
+      "step": 30049
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.908149043708965e-06,
+      "loss": 0.328,
+      "step": 30050
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.9078132084026e-06,
+      "loss": 0.3261,
+      "step": 30051
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.907477373096234e-06,
+      "loss": 0.3339,
+      "step": 30052
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.907141537789869e-06,
+      "loss": 0.2042,
+      "step": 30053
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.906805702483504e-06,
+      "loss": 0.4072,
+      "step": 30054
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.906469867177138e-06,
+      "loss": 0.4186,
+      "step": 30055
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.906134031870771e-06,
+      "loss": 0.2645,
+      "step": 30056
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.905798196564406e-06,
+      "loss": 0.317,
+      "step": 30057
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.90546236125804e-06,
+      "loss": 0.1916,
+      "step": 30058
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.905126525951674e-06,
+      "loss": 0.4399,
+      "step": 30059
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.904790690645308e-06,
+      "loss": 0.4305,
+      "step": 30060
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.904454855338943e-06,
+      "loss": 0.3523,
+      "step": 30061
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.904119020032576e-06,
+      "loss": 0.311,
+      "step": 30062
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.90378318472621e-06,
+      "loss": 0.3365,
+      "step": 30063
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.903447349419845e-06,
+      "loss": 0.1597,
+      "step": 30064
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.90311151411348e-06,
+      "loss": 0.2593,
+      "step": 30065
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.902775678807114e-06,
+      "loss": 0.6004,
+      "step": 30066
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.902439843500749e-06,
+      "loss": 0.3404,
+      "step": 30067
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.902104008194382e-06,
+      "loss": 0.3183,
+      "step": 30068
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.901768172888017e-06,
+      "loss": 0.2858,
+      "step": 30069
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.901432337581651e-06,
+      "loss": 0.2244,
+      "step": 30070
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.901096502275284e-06,
+      "loss": 0.4038,
+      "step": 30071
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.900760666968919e-06,
+      "loss": 0.1918,
+      "step": 30072
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.900424831662553e-06,
+      "loss": 0.2566,
+      "step": 30073
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 9.900088996356188e-06,
+      "loss": 0.334,
+      "step": 30074
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.899753161049823e-06,
+      "loss": 0.2552,
+      "step": 30075
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.899417325743457e-06,
+      "loss": 0.3767,
+      "step": 30076
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.89908149043709e-06,
+      "loss": 0.3242,
+      "step": 30077
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.898745655130725e-06,
+      "loss": 0.3286,
+      "step": 30078
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.89840981982436e-06,
+      "loss": 0.3763,
+      "step": 30079
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.898073984517993e-06,
+      "loss": 0.1597,
+      "step": 30080
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.897738149211627e-06,
+      "loss": 0.2613,
+      "step": 30081
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.897402313905262e-06,
+      "loss": 0.3391,
+      "step": 30082
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.897066478598895e-06,
+      "loss": 0.2234,
+      "step": 30083
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.89673064329253e-06,
+      "loss": 0.2816,
+      "step": 30084
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.896394807986164e-06,
+      "loss": 0.457,
+      "step": 30085
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.896058972679799e-06,
+      "loss": 0.3188,
+      "step": 30086
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.895723137373433e-06,
+      "loss": 0.3941,
+      "step": 30087
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.895387302067068e-06,
+      "loss": 0.1802,
+      "step": 30088
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.895051466760701e-06,
+      "loss": 0.4106,
+      "step": 30089
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.894715631454336e-06,
+      "loss": 0.3527,
+      "step": 30090
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.89437979614797e-06,
+      "loss": 0.2628,
+      "step": 30091
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.894043960841603e-06,
+      "loss": 0.3403,
+      "step": 30092
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.893708125535238e-06,
+      "loss": 0.2827,
+      "step": 30093
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.893372290228873e-06,
+      "loss": 0.2782,
+      "step": 30094
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.893036454922507e-06,
+      "loss": 0.228,
+      "step": 30095
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.892700619616142e-06,
+      "loss": 0.3334,
+      "step": 30096
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.892364784309775e-06,
+      "loss": 0.2154,
+      "step": 30097
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.89202894900341e-06,
+      "loss": 0.4998,
+      "step": 30098
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.891693113697044e-06,
+      "loss": 0.4101,
+      "step": 30099
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.891357278390677e-06,
+      "loss": 0.2552,
+      "step": 30100
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.891021443084312e-06,
+      "loss": 0.2973,
+      "step": 30101
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.890685607777946e-06,
+      "loss": 0.207,
+      "step": 30102
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.890349772471581e-06,
+      "loss": 0.2259,
+      "step": 30103
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.890013937165214e-06,
+      "loss": 0.4638,
+      "step": 30104
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.889678101858849e-06,
+      "loss": 0.3104,
+      "step": 30105
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.889342266552483e-06,
+      "loss": 0.2867,
+      "step": 30106
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.889006431246118e-06,
+      "loss": 0.2234,
+      "step": 30107
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.888670595939753e-06,
+      "loss": 0.3212,
+      "step": 30108
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.888334760633386e-06,
+      "loss": 0.353,
+      "step": 30109
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.88799892532702e-06,
+      "loss": 0.2443,
+      "step": 30110
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.887663090020655e-06,
+      "loss": 0.289,
+      "step": 30111
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.887327254714288e-06,
+      "loss": 0.2516,
+      "step": 30112
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.886991419407922e-06,
+      "loss": 0.2203,
+      "step": 30113
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.886655584101557e-06,
+      "loss": 0.3497,
+      "step": 30114
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.886319748795192e-06,
+      "loss": 0.2997,
+      "step": 30115
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.885983913488826e-06,
+      "loss": 0.3819,
+      "step": 30116
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.885648078182461e-06,
+      "loss": 0.3425,
+      "step": 30117
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.885312242876094e-06,
+      "loss": 0.242,
+      "step": 30118
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.884976407569729e-06,
+      "loss": 0.2155,
+      "step": 30119
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.884640572263363e-06,
+      "loss": 0.288,
+      "step": 30120
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.884304736956996e-06,
+      "loss": 0.2279,
+      "step": 30121
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.883968901650631e-06,
+      "loss": 0.2168,
+      "step": 30122
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.883633066344266e-06,
+      "loss": 0.3005,
+      "step": 30123
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.8832972310379e-06,
+      "loss": 0.2423,
+      "step": 30124
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.882961395731535e-06,
+      "loss": 0.2769,
+      "step": 30125
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.88262556042517e-06,
+      "loss": 0.2902,
+      "step": 30126
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.882289725118802e-06,
+      "loss": 0.3383,
+      "step": 30127
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.881953889812437e-06,
+      "loss": 0.224,
+      "step": 30128
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.881618054506072e-06,
+      "loss": 0.1487,
+      "step": 30129
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.881282219199705e-06,
+      "loss": 0.5053,
+      "step": 30130
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.88094638389334e-06,
+      "loss": 0.2357,
+      "step": 30131
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.880610548586974e-06,
+      "loss": 0.3061,
+      "step": 30132
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.880274713280607e-06,
+      "loss": 0.3526,
+      "step": 30133
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.879938877974242e-06,
+      "loss": 0.248,
+      "step": 30134
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.879603042667876e-06,
+      "loss": 0.4145,
+      "step": 30135
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.879267207361511e-06,
+      "loss": 0.3425,
+      "step": 30136
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.878931372055146e-06,
+      "loss": 0.352,
+      "step": 30137
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.87859553674878e-06,
+      "loss": 0.2198,
+      "step": 30138
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.878259701442413e-06,
+      "loss": 0.2425,
+      "step": 30139
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.877923866136048e-06,
+      "loss": 0.3552,
+      "step": 30140
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.877588030829682e-06,
+      "loss": 0.3152,
+      "step": 30141
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.877252195523315e-06,
+      "loss": 0.2394,
+      "step": 30142
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.87691636021695e-06,
+      "loss": 0.2231,
+      "step": 30143
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.876580524910585e-06,
+      "loss": 0.3444,
+      "step": 30144
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.87624468960422e-06,
+      "loss": 0.5796,
+      "step": 30145
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.875908854297854e-06,
+      "loss": 0.4665,
+      "step": 30146
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.875573018991489e-06,
+      "loss": 0.2214,
+      "step": 30147
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.875237183685122e-06,
+      "loss": 0.412,
+      "step": 30148
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.874901348378756e-06,
+      "loss": 0.1971,
+      "step": 30149
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.87456551307239e-06,
+      "loss": 0.2198,
+      "step": 30150
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.874229677766024e-06,
+      "loss": 0.279,
+      "step": 30151
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.873893842459658e-06,
+      "loss": 0.2775,
+      "step": 30152
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.873558007153293e-06,
+      "loss": 0.2021,
+      "step": 30153
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.873222171846926e-06,
+      "loss": 0.2432,
+      "step": 30154
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.87288633654056e-06,
+      "loss": 0.3016,
+      "step": 30155
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.872550501234195e-06,
+      "loss": 0.2335,
+      "step": 30156
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.87221466592783e-06,
+      "loss": 0.3016,
+      "step": 30157
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.871878830621465e-06,
+      "loss": 0.3957,
+      "step": 30158
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.8715429953151e-06,
+      "loss": 0.2659,
+      "step": 30159
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.871207160008732e-06,
+      "loss": 0.354,
+      "step": 30160
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.870871324702367e-06,
+      "loss": 0.2322,
+      "step": 30161
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.870535489396002e-06,
+      "loss": 0.2308,
+      "step": 30162
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.870199654089634e-06,
+      "loss": 0.296,
+      "step": 30163
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.869863818783269e-06,
+      "loss": 0.2308,
+      "step": 30164
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.869527983476904e-06,
+      "loss": 0.3951,
+      "step": 30165
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.869192148170538e-06,
+      "loss": 0.2743,
+      "step": 30166
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.868856312864173e-06,
+      "loss": 0.2865,
+      "step": 30167
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.868520477557808e-06,
+      "loss": 0.4032,
+      "step": 30168
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.86818464225144e-06,
+      "loss": 0.3118,
+      "step": 30169
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.867848806945075e-06,
+      "loss": 0.3755,
+      "step": 30170
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.86751297163871e-06,
+      "loss": 0.3306,
+      "step": 30171
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.867177136332343e-06,
+      "loss": 0.149,
+      "step": 30172
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.866841301025978e-06,
+      "loss": 0.2865,
+      "step": 30173
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.866505465719612e-06,
+      "loss": 0.3196,
+      "step": 30174
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.866169630413245e-06,
+      "loss": 0.3653,
+      "step": 30175
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.86583379510688e-06,
+      "loss": 0.3398,
+      "step": 30176
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.865497959800514e-06,
+      "loss": 0.3034,
+      "step": 30177
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.865162124494149e-06,
+      "loss": 0.2893,
+      "step": 30178
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.864826289187784e-06,
+      "loss": 0.3529,
+      "step": 30179
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.864490453881418e-06,
+      "loss": 0.3406,
+      "step": 30180
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.864154618575051e-06,
+      "loss": 0.3257,
+      "step": 30181
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.863818783268686e-06,
+      "loss": 0.378,
+      "step": 30182
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.863482947962319e-06,
+      "loss": 0.4561,
+      "step": 30183
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.863147112655954e-06,
+      "loss": 0.2323,
+      "step": 30184
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.862811277349588e-06,
+      "loss": 0.3384,
+      "step": 30185
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.862475442043223e-06,
+      "loss": 0.2116,
+      "step": 30186
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.862139606736858e-06,
+      "loss": 0.4944,
+      "step": 30187
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.861803771430492e-06,
+      "loss": 0.5405,
+      "step": 30188
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.861467936124125e-06,
+      "loss": 0.3039,
+      "step": 30189
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.86113210081776e-06,
+      "loss": 0.3837,
+      "step": 30190
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.860796265511394e-06,
+      "loss": 0.2981,
+      "step": 30191
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.860460430205027e-06,
+      "loss": 0.26,
+      "step": 30192
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.860124594898662e-06,
+      "loss": 0.3583,
+      "step": 30193
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.859788759592297e-06,
+      "loss": 0.3865,
+      "step": 30194
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.85945292428593e-06,
+      "loss": 0.3379,
+      "step": 30195
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.859117088979564e-06,
+      "loss": 0.2708,
+      "step": 30196
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.858781253673199e-06,
+      "loss": 0.2986,
+      "step": 30197
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.858445418366834e-06,
+      "loss": 0.1643,
+      "step": 30198
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.858109583060468e-06,
+      "loss": 0.3486,
+      "step": 30199
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.857773747754103e-06,
+      "loss": 0.2995,
+      "step": 30200
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.857437912447736e-06,
+      "loss": 0.2564,
+      "step": 30201
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.85710207714137e-06,
+      "loss": 0.2105,
+      "step": 30202
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.856766241835005e-06,
+      "loss": 0.3665,
+      "step": 30203
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.856430406528638e-06,
+      "loss": 0.4665,
+      "step": 30204
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.856094571222273e-06,
+      "loss": 0.2483,
+      "step": 30205
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.855758735915907e-06,
+      "loss": 0.2013,
+      "step": 30206
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.855422900609542e-06,
+      "loss": 0.3625,
+      "step": 30207
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.855087065303177e-06,
+      "loss": 0.2013,
+      "step": 30208
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.854751229996811e-06,
+      "loss": 0.2235,
+      "step": 30209
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.854415394690444e-06,
+      "loss": 0.204,
+      "step": 30210
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.854079559384079e-06,
+      "loss": 0.246,
+      "step": 30211
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.853743724077714e-06,
+      "loss": 0.2769,
+      "step": 30212
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.853407888771347e-06,
+      "loss": 0.3209,
+      "step": 30213
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.853072053464981e-06,
+      "loss": 0.4329,
+      "step": 30214
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.852736218158616e-06,
+      "loss": 0.2166,
+      "step": 30215
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.85240038285225e-06,
+      "loss": 0.1823,
+      "step": 30216
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.852064547545883e-06,
+      "loss": 0.3885,
+      "step": 30217
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.851728712239518e-06,
+      "loss": 0.1885,
+      "step": 30218
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.851392876933153e-06,
+      "loss": 0.3766,
+      "step": 30219
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.851057041626787e-06,
+      "loss": 0.175,
+      "step": 30220
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.850721206320422e-06,
+      "loss": 0.2424,
+      "step": 30221
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.850385371014055e-06,
+      "loss": 0.2444,
+      "step": 30222
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.85004953570769e-06,
+      "loss": 0.2564,
+      "step": 30223
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.849713700401324e-06,
+      "loss": 0.4048,
+      "step": 30224
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.849377865094957e-06,
+      "loss": 0.2175,
+      "step": 30225
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.849042029788592e-06,
+      "loss": 0.303,
+      "step": 30226
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.848706194482227e-06,
+      "loss": 0.179,
+      "step": 30227
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.848370359175861e-06,
+      "loss": 0.344,
+      "step": 30228
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.848034523869496e-06,
+      "loss": 0.3554,
+      "step": 30229
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.84769868856313e-06,
+      "loss": 0.3146,
+      "step": 30230
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.847362853256763e-06,
+      "loss": 0.2136,
+      "step": 30231
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.847027017950398e-06,
+      "loss": 0.2313,
+      "step": 30232
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.846691182644033e-06,
+      "loss": 0.2648,
+      "step": 30233
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.846355347337666e-06,
+      "loss": 0.2562,
+      "step": 30234
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.8460195120313e-06,
+      "loss": 0.3645,
+      "step": 30235
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.845683676724935e-06,
+      "loss": 0.3712,
+      "step": 30236
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.84534784141857e-06,
+      "loss": 0.2451,
+      "step": 30237
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.845012006112204e-06,
+      "loss": 0.4521,
+      "step": 30238
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.844676170805839e-06,
+      "loss": 0.3097,
+      "step": 30239
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.844340335499472e-06,
+      "loss": 0.4483,
+      "step": 30240
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.844004500193106e-06,
+      "loss": 0.2347,
+      "step": 30241
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.843668664886741e-06,
+      "loss": 0.2925,
+      "step": 30242
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.843332829580374e-06,
+      "loss": 0.2248,
+      "step": 30243
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.842996994274009e-06,
+      "loss": 0.3376,
+      "step": 30244
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.842661158967643e-06,
+      "loss": 0.2123,
+      "step": 30245
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.842325323661276e-06,
+      "loss": 0.2505,
+      "step": 30246
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.841989488354911e-06,
+      "loss": 0.3907,
+      "step": 30247
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.841653653048546e-06,
+      "loss": 0.2693,
+      "step": 30248
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.84131781774218e-06,
+      "loss": 0.2968,
+      "step": 30249
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.840981982435815e-06,
+      "loss": 0.2418,
+      "step": 30250
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.84064614712945e-06,
+      "loss": 0.3421,
+      "step": 30251
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.840310311823083e-06,
+      "loss": 0.3192,
+      "step": 30252
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.839974476516717e-06,
+      "loss": 0.1932,
+      "step": 30253
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.839638641210352e-06,
+      "loss": 0.3379,
+      "step": 30254
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.839302805903985e-06,
+      "loss": 0.2788,
+      "step": 30255
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.83896697059762e-06,
+      "loss": 0.2743,
+      "step": 30256
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.838631135291254e-06,
+      "loss": 0.3388,
+      "step": 30257
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.838295299984889e-06,
+      "loss": 0.3748,
+      "step": 30258
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.837959464678523e-06,
+      "loss": 0.2581,
+      "step": 30259
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.837623629372158e-06,
+      "loss": 0.4697,
+      "step": 30260
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.837287794065791e-06,
+      "loss": 0.3096,
+      "step": 30261
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.836951958759426e-06,
+      "loss": 0.359,
+      "step": 30262
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.83661612345306e-06,
+      "loss": 0.4668,
+      "step": 30263
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.836280288146693e-06,
+      "loss": 0.2674,
+      "step": 30264
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.835944452840328e-06,
+      "loss": 0.3452,
+      "step": 30265
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.835608617533963e-06,
+      "loss": 0.3522,
+      "step": 30266
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.835272782227595e-06,
+      "loss": 0.295,
+      "step": 30267
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.83493694692123e-06,
+      "loss": 0.3882,
+      "step": 30268
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.834601111614865e-06,
+      "loss": 0.3792,
+      "step": 30269
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.8342652763085e-06,
+      "loss": 0.3432,
+      "step": 30270
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.833929441002134e-06,
+      "loss": 0.283,
+      "step": 30271
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 9.833593605695769e-06,
+      "loss": 0.2638,
+      "step": 30272
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.833257770389402e-06,
+      "loss": 0.1377,
+      "step": 30273
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.832921935083036e-06,
+      "loss": 0.4686,
+      "step": 30274
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.83258609977667e-06,
+      "loss": 0.3892,
+      "step": 30275
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.832250264470304e-06,
+      "loss": 0.3021,
+      "step": 30276
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.831914429163939e-06,
+      "loss": 0.2501,
+      "step": 30277
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.831578593857573e-06,
+      "loss": 0.3111,
+      "step": 30278
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.831242758551208e-06,
+      "loss": 0.2189,
+      "step": 30279
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.830906923244842e-06,
+      "loss": 0.3703,
+      "step": 30280
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.830571087938475e-06,
+      "loss": 0.4712,
+      "step": 30281
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.83023525263211e-06,
+      "loss": 0.3172,
+      "step": 30282
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.829899417325745e-06,
+      "loss": 0.3854,
+      "step": 30283
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.829563582019378e-06,
+      "loss": 0.2119,
+      "step": 30284
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.829227746713012e-06,
+      "loss": 0.2728,
+      "step": 30285
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.828891911406647e-06,
+      "loss": 0.365,
+      "step": 30286
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.82855607610028e-06,
+      "loss": 0.3207,
+      "step": 30287
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.828220240793915e-06,
+      "loss": 0.1637,
+      "step": 30288
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.82788440548755e-06,
+      "loss": 0.1744,
+      "step": 30289
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.827548570181184e-06,
+      "loss": 0.3494,
+      "step": 30290
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.827212734874819e-06,
+      "loss": 0.2918,
+      "step": 30291
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.826876899568453e-06,
+      "loss": 0.3133,
+      "step": 30292
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.826541064262086e-06,
+      "loss": 0.1907,
+      "step": 30293
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.82620522895572e-06,
+      "loss": 0.3282,
+      "step": 30294
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.825869393649355e-06,
+      "loss": 0.3085,
+      "step": 30295
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.825533558342988e-06,
+      "loss": 0.2536,
+      "step": 30296
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.825197723036623e-06,
+      "loss": 0.3426,
+      "step": 30297
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.824861887730258e-06,
+      "loss": 0.3012,
+      "step": 30298
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.824526052423892e-06,
+      "loss": 0.3323,
+      "step": 30299
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.824190217117527e-06,
+      "loss": 0.2174,
+      "step": 30300
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.823854381811162e-06,
+      "loss": 0.2009,
+      "step": 30301
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.823518546504795e-06,
+      "loss": 0.2204,
+      "step": 30302
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.82318271119843e-06,
+      "loss": 0.2751,
+      "step": 30303
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.822846875892064e-06,
+      "loss": 0.302,
+      "step": 30304
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.822511040585697e-06,
+      "loss": 0.2621,
+      "step": 30305
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.822175205279331e-06,
+      "loss": 0.2366,
+      "step": 30306
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.821839369972966e-06,
+      "loss": 0.135,
+      "step": 30307
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.821503534666599e-06,
+      "loss": 0.1355,
+      "step": 30308
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.821167699360234e-06,
+      "loss": 0.3716,
+      "step": 30309
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.820831864053868e-06,
+      "loss": 0.1494,
+      "step": 30310
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.820496028747503e-06,
+      "loss": 0.2648,
+      "step": 30311
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.820160193441138e-06,
+      "loss": 0.3079,
+      "step": 30312
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.819824358134772e-06,
+      "loss": 0.5869,
+      "step": 30313
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.819488522828405e-06,
+      "loss": 0.3313,
+      "step": 30314
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.81915268752204e-06,
+      "loss": 0.439,
+      "step": 30315
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.818816852215675e-06,
+      "loss": 0.3,
+      "step": 30316
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.818481016909308e-06,
+      "loss": 0.2656,
+      "step": 30317
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.818145181602942e-06,
+      "loss": 0.233,
+      "step": 30318
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.817809346296577e-06,
+      "loss": 0.3553,
+      "step": 30319
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.817473510990211e-06,
+      "loss": 0.1973,
+      "step": 30320
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.817137675683846e-06,
+      "loss": 0.1477,
+      "step": 30321
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.81680184037748e-06,
+      "loss": 0.1831,
+      "step": 30322
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.816466005071114e-06,
+      "loss": 0.2143,
+      "step": 30323
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.816130169764748e-06,
+      "loss": 0.3746,
+      "step": 30324
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.815794334458383e-06,
+      "loss": 0.3134,
+      "step": 30325
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.815458499152016e-06,
+      "loss": 0.2151,
+      "step": 30326
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.81512266384565e-06,
+      "loss": 0.3169,
+      "step": 30327
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.814786828539285e-06,
+      "loss": 0.3422,
+      "step": 30328
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.81445099323292e-06,
+      "loss": 0.5121,
+      "step": 30329
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.814115157926553e-06,
+      "loss": 0.2743,
+      "step": 30330
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.813779322620187e-06,
+      "loss": 0.2343,
+      "step": 30331
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.813443487313822e-06,
+      "loss": 0.3477,
+      "step": 30332
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.813107652007457e-06,
+      "loss": 0.3294,
+      "step": 30333
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.812771816701091e-06,
+      "loss": 0.2789,
+      "step": 30334
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.812435981394724e-06,
+      "loss": 0.1974,
+      "step": 30335
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.812100146088359e-06,
+      "loss": 0.3167,
+      "step": 30336
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.811764310781994e-06,
+      "loss": 0.2525,
+      "step": 30337
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.811428475475627e-06,
+      "loss": 0.3697,
+      "step": 30338
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.811092640169261e-06,
+      "loss": 0.2997,
+      "step": 30339
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.810756804862896e-06,
+      "loss": 0.2724,
+      "step": 30340
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.81042096955653e-06,
+      "loss": 0.4245,
+      "step": 30341
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.810085134250165e-06,
+      "loss": 0.3265,
+      "step": 30342
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.8097492989438e-06,
+      "loss": 0.3123,
+      "step": 30343
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.809413463637433e-06,
+      "loss": 0.2389,
+      "step": 30344
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.809077628331067e-06,
+      "loss": 0.3316,
+      "step": 30345
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.808741793024702e-06,
+      "loss": 0.2875,
+      "step": 30346
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.808405957718335e-06,
+      "loss": 0.3854,
+      "step": 30347
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.80807012241197e-06,
+      "loss": 0.224,
+      "step": 30348
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.807734287105604e-06,
+      "loss": 0.1293,
+      "step": 30349
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.807398451799239e-06,
+      "loss": 0.2662,
+      "step": 30350
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.807062616492874e-06,
+      "loss": 0.1469,
+      "step": 30351
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.806726781186507e-06,
+      "loss": 0.2338,
+      "step": 30352
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.806390945880141e-06,
+      "loss": 0.3512,
+      "step": 30353
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.806055110573776e-06,
+      "loss": 0.4096,
+      "step": 30354
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.80571927526741e-06,
+      "loss": 0.3323,
+      "step": 30355
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.805383439961044e-06,
+      "loss": 0.5878,
+      "step": 30356
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.805047604654678e-06,
+      "loss": 0.3254,
+      "step": 30357
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.804711769348313e-06,
+      "loss": 0.3997,
+      "step": 30358
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.804375934041946e-06,
+      "loss": 0.4859,
+      "step": 30359
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.80404009873558e-06,
+      "loss": 0.2024,
+      "step": 30360
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.803704263429215e-06,
+      "loss": 0.3142,
+      "step": 30361
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.80336842812285e-06,
+      "loss": 0.4329,
+      "step": 30362
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.803032592816484e-06,
+      "loss": 0.2309,
+      "step": 30363
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.802696757510117e-06,
+      "loss": 0.2542,
+      "step": 30364
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.802360922203752e-06,
+      "loss": 0.1886,
+      "step": 30365
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.802025086897387e-06,
+      "loss": 0.2132,
+      "step": 30366
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.80168925159102e-06,
+      "loss": 0.3667,
+      "step": 30367
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.801353416284654e-06,
+      "loss": 0.4303,
+      "step": 30368
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.801017580978289e-06,
+      "loss": 0.2111,
+      "step": 30369
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.800681745671923e-06,
+      "loss": 0.1735,
+      "step": 30370
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.800345910365558e-06,
+      "loss": 0.3735,
+      "step": 30371
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.800010075059193e-06,
+      "loss": 0.247,
+      "step": 30372
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.799674239752826e-06,
+      "loss": 0.3201,
+      "step": 30373
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.79933840444646e-06,
+      "loss": 0.4493,
+      "step": 30374
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.799002569140095e-06,
+      "loss": 0.312,
+      "step": 30375
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.798666733833728e-06,
+      "loss": 0.2079,
+      "step": 30376
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.798330898527363e-06,
+      "loss": 0.2469,
+      "step": 30377
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.797995063220997e-06,
+      "loss": 0.3173,
+      "step": 30378
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.79765922791463e-06,
+      "loss": 0.3029,
+      "step": 30379
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.797323392608265e-06,
+      "loss": 0.3249,
+      "step": 30380
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.7969875573019e-06,
+      "loss": 0.3389,
+      "step": 30381
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.796651721995534e-06,
+      "loss": 0.2709,
+      "step": 30382
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.796315886689169e-06,
+      "loss": 0.3329,
+      "step": 30383
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.795980051382803e-06,
+      "loss": 0.4008,
+      "step": 30384
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.795644216076436e-06,
+      "loss": 0.4041,
+      "step": 30385
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.795308380770071e-06,
+      "loss": 0.269,
+      "step": 30386
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.794972545463706e-06,
+      "loss": 0.2654,
+      "step": 30387
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.794636710157339e-06,
+      "loss": 0.3378,
+      "step": 30388
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.794300874850973e-06,
+      "loss": 0.5249,
+      "step": 30389
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.793965039544608e-06,
+      "loss": 0.3293,
+      "step": 30390
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.793629204238243e-06,
+      "loss": 0.3758,
+      "step": 30391
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.793293368931877e-06,
+      "loss": 0.274,
+      "step": 30392
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.792957533625512e-06,
+      "loss": 0.2757,
+      "step": 30393
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.792621698319145e-06,
+      "loss": 0.3303,
+      "step": 30394
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.79228586301278e-06,
+      "loss": 0.2787,
+      "step": 30395
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.791950027706414e-06,
+      "loss": 0.4537,
+      "step": 30396
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.791614192400047e-06,
+      "loss": 0.5323,
+      "step": 30397
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.791278357093682e-06,
+      "loss": 0.2167,
+      "step": 30398
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.790942521787316e-06,
+      "loss": 0.3242,
+      "step": 30399
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.79060668648095e-06,
+      "loss": 0.4924,
+      "step": 30400
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.790270851174584e-06,
+      "loss": 0.334,
+      "step": 30401
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.789935015868219e-06,
+      "loss": 0.2027,
+      "step": 30402
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.789599180561853e-06,
+      "loss": 0.158,
+      "step": 30403
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.789263345255488e-06,
+      "loss": 0.2633,
+      "step": 30404
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.788927509949123e-06,
+      "loss": 0.332,
+      "step": 30405
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.788591674642756e-06,
+      "loss": 0.3029,
+      "step": 30406
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.78825583933639e-06,
+      "loss": 0.2895,
+      "step": 30407
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.787920004030025e-06,
+      "loss": 0.4579,
+      "step": 30408
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.787584168723658e-06,
+      "loss": 0.2391,
+      "step": 30409
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.787248333417292e-06,
+      "loss": 0.4067,
+      "step": 30410
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.786912498110927e-06,
+      "loss": 0.2431,
+      "step": 30411
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.786576662804562e-06,
+      "loss": 0.4311,
+      "step": 30412
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.786240827498196e-06,
+      "loss": 0.496,
+      "step": 30413
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.785904992191831e-06,
+      "loss": 0.2604,
+      "step": 30414
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.785569156885464e-06,
+      "loss": 0.3789,
+      "step": 30415
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.785233321579099e-06,
+      "loss": 0.217,
+      "step": 30416
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.784897486272733e-06,
+      "loss": 0.2891,
+      "step": 30417
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.784561650966366e-06,
+      "loss": 0.4507,
+      "step": 30418
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.784225815660001e-06,
+      "loss": 0.3092,
+      "step": 30419
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.783889980353636e-06,
+      "loss": 0.3598,
+      "step": 30420
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.783554145047268e-06,
+      "loss": 0.2936,
+      "step": 30421
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.783218309740903e-06,
+      "loss": 0.2698,
+      "step": 30422
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.782882474434538e-06,
+      "loss": 0.2017,
+      "step": 30423
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.782546639128172e-06,
+      "loss": 0.3402,
+      "step": 30424
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.782210803821807e-06,
+      "loss": 0.2251,
+      "step": 30425
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.781874968515442e-06,
+      "loss": 0.2121,
+      "step": 30426
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.781539133209075e-06,
+      "loss": 0.2483,
+      "step": 30427
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.78120329790271e-06,
+      "loss": 0.2536,
+      "step": 30428
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.780867462596344e-06,
+      "loss": 0.2453,
+      "step": 30429
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.780531627289977e-06,
+      "loss": 0.2651,
+      "step": 30430
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.780195791983612e-06,
+      "loss": 0.3004,
+      "step": 30431
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.779859956677246e-06,
+      "loss": 0.2871,
+      "step": 30432
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.779524121370881e-06,
+      "loss": 0.1999,
+      "step": 30433
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.779188286064516e-06,
+      "loss": 0.1675,
+      "step": 30434
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.77885245075815e-06,
+      "loss": 0.2037,
+      "step": 30435
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.778516615451783e-06,
+      "loss": 0.4232,
+      "step": 30436
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.778180780145418e-06,
+      "loss": 0.3143,
+      "step": 30437
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.777844944839052e-06,
+      "loss": 0.2122,
+      "step": 30438
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.777509109532685e-06,
+      "loss": 0.3392,
+      "step": 30439
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.77717327422632e-06,
+      "loss": 0.2633,
+      "step": 30440
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.776837438919955e-06,
+      "loss": 0.2372,
+      "step": 30441
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.77650160361359e-06,
+      "loss": 0.1797,
+      "step": 30442
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.776165768307222e-06,
+      "loss": 0.2998,
+      "step": 30443
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.775829933000857e-06,
+      "loss": 0.2511,
+      "step": 30444
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.775494097694492e-06,
+      "loss": 0.3542,
+      "step": 30445
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.775158262388126e-06,
+      "loss": 0.3872,
+      "step": 30446
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.77482242708176e-06,
+      "loss": 0.3523,
+      "step": 30447
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.774486591775394e-06,
+      "loss": 0.3903,
+      "step": 30448
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.774150756469028e-06,
+      "loss": 0.1331,
+      "step": 30449
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.773814921162663e-06,
+      "loss": 0.2803,
+      "step": 30450
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.773479085856296e-06,
+      "loss": 0.434,
+      "step": 30451
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.77314325054993e-06,
+      "loss": 0.3015,
+      "step": 30452
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.772807415243565e-06,
+      "loss": 0.134,
+      "step": 30453
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.7724715799372e-06,
+      "loss": 0.3225,
+      "step": 30454
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.772135744630835e-06,
+      "loss": 0.3452,
+      "step": 30455
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.771799909324468e-06,
+      "loss": 0.2341,
+      "step": 30456
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.771464074018102e-06,
+      "loss": 0.2185,
+      "step": 30457
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.771128238711737e-06,
+      "loss": 0.3793,
+      "step": 30458
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.77079240340537e-06,
+      "loss": 0.2409,
+      "step": 30459
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.770456568099004e-06,
+      "loss": 0.2962,
+      "step": 30460
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.770120732792639e-06,
+      "loss": 0.3324,
+      "step": 30461
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.769784897486274e-06,
+      "loss": 0.2189,
+      "step": 30462
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.769449062179908e-06,
+      "loss": 0.3799,
+      "step": 30463
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.769113226873543e-06,
+      "loss": 0.4155,
+      "step": 30464
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.768777391567176e-06,
+      "loss": 0.3743,
+      "step": 30465
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.76844155626081e-06,
+      "loss": 0.2912,
+      "step": 30466
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.768105720954445e-06,
+      "loss": 0.2243,
+      "step": 30467
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.767769885648078e-06,
+      "loss": 0.3587,
+      "step": 30468
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.767434050341713e-06,
+      "loss": 0.2326,
+      "step": 30469
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.767098215035348e-06,
+      "loss": 0.2368,
+      "step": 30470
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 9.76676237972898e-06,
+      "loss": 0.3455,
+      "step": 30471
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.766426544422615e-06,
+      "loss": 0.4095,
+      "step": 30472
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.76609070911625e-06,
+      "loss": 0.1186,
+      "step": 30473
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.765754873809884e-06,
+      "loss": 0.2533,
+      "step": 30474
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.765419038503519e-06,
+      "loss": 0.1867,
+      "step": 30475
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.765083203197154e-06,
+      "loss": 0.2262,
+      "step": 30476
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.764747367890787e-06,
+      "loss": 0.2937,
+      "step": 30477
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.764411532584421e-06,
+      "loss": 0.2283,
+      "step": 30478
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.764075697278056e-06,
+      "loss": 0.3282,
+      "step": 30479
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.763739861971689e-06,
+      "loss": 0.2417,
+      "step": 30480
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.763404026665324e-06,
+      "loss": 0.2122,
+      "step": 30481
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.763068191358958e-06,
+      "loss": 0.1949,
+      "step": 30482
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.762732356052593e-06,
+      "loss": 0.174,
+      "step": 30483
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.762396520746228e-06,
+      "loss": 0.1989,
+      "step": 30484
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.762060685439862e-06,
+      "loss": 0.2363,
+      "step": 30485
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.761724850133495e-06,
+      "loss": 0.3184,
+      "step": 30486
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.76138901482713e-06,
+      "loss": 0.4112,
+      "step": 30487
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.761053179520764e-06,
+      "loss": 0.3628,
+      "step": 30488
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.760717344214397e-06,
+      "loss": 0.2523,
+      "step": 30489
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.760381508908032e-06,
+      "loss": 0.3777,
+      "step": 30490
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.760045673601667e-06,
+      "loss": 0.3871,
+      "step": 30491
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.7597098382953e-06,
+      "loss": 0.3667,
+      "step": 30492
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.759374002988934e-06,
+      "loss": 0.3077,
+      "step": 30493
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.759038167682569e-06,
+      "loss": 0.2652,
+      "step": 30494
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.758702332376204e-06,
+      "loss": 0.3083,
+      "step": 30495
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.758366497069838e-06,
+      "loss": 0.2894,
+      "step": 30496
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.758030661763473e-06,
+      "loss": 0.2488,
+      "step": 30497
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.757694826457106e-06,
+      "loss": 0.2821,
+      "step": 30498
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.75735899115074e-06,
+      "loss": 0.3265,
+      "step": 30499
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.757023155844375e-06,
+      "loss": 0.2484,
+      "step": 30500
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.756687320538008e-06,
+      "loss": 0.296,
+      "step": 30501
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.756351485231643e-06,
+      "loss": 0.3837,
+      "step": 30502
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.756015649925277e-06,
+      "loss": 0.3706,
+      "step": 30503
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.755679814618912e-06,
+      "loss": 0.2998,
+      "step": 30504
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.755343979312547e-06,
+      "loss": 0.2015,
+      "step": 30505
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.755008144006181e-06,
+      "loss": 0.3495,
+      "step": 30506
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.754672308699814e-06,
+      "loss": 0.374,
+      "step": 30507
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.754336473393449e-06,
+      "loss": 0.328,
+      "step": 30508
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.754000638087084e-06,
+      "loss": 0.2118,
+      "step": 30509
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.753664802780717e-06,
+      "loss": 0.1796,
+      "step": 30510
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.753328967474351e-06,
+      "loss": 0.444,
+      "step": 30511
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.752993132167986e-06,
+      "loss": 0.3211,
+      "step": 30512
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.752657296861619e-06,
+      "loss": 0.4716,
+      "step": 30513
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.752321461555253e-06,
+      "loss": 0.4162,
+      "step": 30514
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.751985626248888e-06,
+      "loss": 0.2247,
+      "step": 30515
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.751649790942523e-06,
+      "loss": 0.2173,
+      "step": 30516
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.751313955636157e-06,
+      "loss": 0.3431,
+      "step": 30517
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.750978120329792e-06,
+      "loss": 0.4836,
+      "step": 30518
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.750642285023425e-06,
+      "loss": 0.3137,
+      "step": 30519
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.75030644971706e-06,
+      "loss": 0.2759,
+      "step": 30520
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.749970614410694e-06,
+      "loss": 0.1529,
+      "step": 30521
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.749634779104327e-06,
+      "loss": 0.2853,
+      "step": 30522
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.749298943797962e-06,
+      "loss": 0.3717,
+      "step": 30523
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.748963108491597e-06,
+      "loss": 0.2054,
+      "step": 30524
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.748627273185231e-06,
+      "loss": 0.2924,
+      "step": 30525
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.748291437878866e-06,
+      "loss": 0.3567,
+      "step": 30526
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.7479556025725e-06,
+      "loss": 0.4781,
+      "step": 30527
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.747619767266133e-06,
+      "loss": 0.346,
+      "step": 30528
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.747283931959768e-06,
+      "loss": 0.5557,
+      "step": 30529
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.746948096653403e-06,
+      "loss": 0.2127,
+      "step": 30530
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.746612261347036e-06,
+      "loss": 0.2063,
+      "step": 30531
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.74627642604067e-06,
+      "loss": 0.237,
+      "step": 30532
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.745940590734305e-06,
+      "loss": 0.3379,
+      "step": 30533
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.745604755427938e-06,
+      "loss": 0.3387,
+      "step": 30534
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.745268920121573e-06,
+      "loss": 0.3485,
+      "step": 30535
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.744933084815207e-06,
+      "loss": 0.3441,
+      "step": 30536
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.744597249508842e-06,
+      "loss": 0.337,
+      "step": 30537
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.744261414202476e-06,
+      "loss": 0.3563,
+      "step": 30538
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.743925578896111e-06,
+      "loss": 0.2036,
+      "step": 30539
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.743589743589744e-06,
+      "loss": 0.3407,
+      "step": 30540
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.743253908283379e-06,
+      "loss": 0.3025,
+      "step": 30541
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.742918072977013e-06,
+      "loss": 0.3465,
+      "step": 30542
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.742582237670646e-06,
+      "loss": 0.3774,
+      "step": 30543
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.742246402364281e-06,
+      "loss": 0.3107,
+      "step": 30544
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.741910567057916e-06,
+      "loss": 0.4846,
+      "step": 30545
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.74157473175155e-06,
+      "loss": 0.399,
+      "step": 30546
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.741238896445185e-06,
+      "loss": 0.2887,
+      "step": 30547
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.740903061138818e-06,
+      "loss": 0.19,
+      "step": 30548
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.740567225832453e-06,
+      "loss": 0.3132,
+      "step": 30549
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.740231390526087e-06,
+      "loss": 0.2569,
+      "step": 30550
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.73989555521972e-06,
+      "loss": 0.193,
+      "step": 30551
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.739559719913355e-06,
+      "loss": 0.2788,
+      "step": 30552
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.73922388460699e-06,
+      "loss": 0.3136,
+      "step": 30553
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.738888049300624e-06,
+      "loss": 0.2407,
+      "step": 30554
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.738552213994259e-06,
+      "loss": 0.2116,
+      "step": 30555
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.738216378687892e-06,
+      "loss": 0.2804,
+      "step": 30556
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.737880543381526e-06,
+      "loss": 0.195,
+      "step": 30557
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.737544708075161e-06,
+      "loss": 0.1483,
+      "step": 30558
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.737208872768796e-06,
+      "loss": 0.2365,
+      "step": 30559
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.736873037462429e-06,
+      "loss": 0.3021,
+      "step": 30560
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.736537202156063e-06,
+      "loss": 0.482,
+      "step": 30561
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.736201366849698e-06,
+      "loss": 0.3394,
+      "step": 30562
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.73586553154333e-06,
+      "loss": 0.2644,
+      "step": 30563
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.735529696236965e-06,
+      "loss": 0.3227,
+      "step": 30564
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.7351938609306e-06,
+      "loss": 0.2089,
+      "step": 30565
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.734858025624235e-06,
+      "loss": 0.2484,
+      "step": 30566
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.73452219031787e-06,
+      "loss": 0.1635,
+      "step": 30567
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.734186355011504e-06,
+      "loss": 0.2892,
+      "step": 30568
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.733850519705137e-06,
+      "loss": 0.2768,
+      "step": 30569
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.733514684398772e-06,
+      "loss": 0.2629,
+      "step": 30570
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.733178849092406e-06,
+      "loss": 0.3115,
+      "step": 30571
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.73284301378604e-06,
+      "loss": 0.2977,
+      "step": 30572
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.732507178479674e-06,
+      "loss": 0.0746,
+      "step": 30573
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.732171343173309e-06,
+      "loss": 0.2916,
+      "step": 30574
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.731835507866943e-06,
+      "loss": 0.326,
+      "step": 30575
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.731499672560578e-06,
+      "loss": 0.1309,
+      "step": 30576
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.731163837254212e-06,
+      "loss": 0.2813,
+      "step": 30577
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.730828001947845e-06,
+      "loss": 0.3497,
+      "step": 30578
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.73049216664148e-06,
+      "loss": 0.2399,
+      "step": 30579
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.730156331335115e-06,
+      "loss": 0.4201,
+      "step": 30580
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.729820496028748e-06,
+      "loss": 0.286,
+      "step": 30581
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.729484660722382e-06,
+      "loss": 0.5337,
+      "step": 30582
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.729148825416017e-06,
+      "loss": 0.3555,
+      "step": 30583
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.72881299010965e-06,
+      "loss": 0.3795,
+      "step": 30584
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.728477154803285e-06,
+      "loss": 0.2584,
+      "step": 30585
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.72814131949692e-06,
+      "loss": 0.2992,
+      "step": 30586
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.727805484190554e-06,
+      "loss": 0.276,
+      "step": 30587
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.727469648884189e-06,
+      "loss": 0.173,
+      "step": 30588
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.727133813577823e-06,
+      "loss": 0.2266,
+      "step": 30589
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.726797978271456e-06,
+      "loss": 0.2355,
+      "step": 30590
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.72646214296509e-06,
+      "loss": 0.1787,
+      "step": 30591
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.726126307658725e-06,
+      "loss": 0.1795,
+      "step": 30592
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.725790472352358e-06,
+      "loss": 0.4071,
+      "step": 30593
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.725454637045993e-06,
+      "loss": 0.2163,
+      "step": 30594
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.725118801739628e-06,
+      "loss": 0.3168,
+      "step": 30595
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.724782966433262e-06,
+      "loss": 0.2686,
+      "step": 30596
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.724447131126897e-06,
+      "loss": 0.225,
+      "step": 30597
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.724111295820532e-06,
+      "loss": 0.3753,
+      "step": 30598
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.723775460514165e-06,
+      "loss": 0.2314,
+      "step": 30599
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.7234396252078e-06,
+      "loss": 0.2503,
+      "step": 30600
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.723103789901434e-06,
+      "loss": 0.2085,
+      "step": 30601
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.722767954595067e-06,
+      "loss": 0.4689,
+      "step": 30602
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.722432119288701e-06,
+      "loss": 0.5315,
+      "step": 30603
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.722096283982336e-06,
+      "loss": 0.248,
+      "step": 30604
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.721760448675969e-06,
+      "loss": 0.3365,
+      "step": 30605
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.721424613369604e-06,
+      "loss": 0.2548,
+      "step": 30606
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.721088778063238e-06,
+      "loss": 0.292,
+      "step": 30607
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.720752942756873e-06,
+      "loss": 0.1979,
+      "step": 30608
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.720417107450508e-06,
+      "loss": 0.2795,
+      "step": 30609
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.720081272144142e-06,
+      "loss": 0.2157,
+      "step": 30610
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.719745436837775e-06,
+      "loss": 0.2202,
+      "step": 30611
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.71940960153141e-06,
+      "loss": 0.266,
+      "step": 30612
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.719073766225045e-06,
+      "loss": 0.4739,
+      "step": 30613
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.718737930918677e-06,
+      "loss": 0.1894,
+      "step": 30614
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.718402095612312e-06,
+      "loss": 0.2341,
+      "step": 30615
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.718066260305947e-06,
+      "loss": 0.3009,
+      "step": 30616
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.717730424999581e-06,
+      "loss": 0.2261,
+      "step": 30617
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.717394589693216e-06,
+      "loss": 0.2658,
+      "step": 30618
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.71705875438685e-06,
+      "loss": 0.4397,
+      "step": 30619
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.716722919080484e-06,
+      "loss": 0.2639,
+      "step": 30620
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.716387083774118e-06,
+      "loss": 0.2433,
+      "step": 30621
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.716051248467753e-06,
+      "loss": 0.1568,
+      "step": 30622
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.715715413161386e-06,
+      "loss": 0.3618,
+      "step": 30623
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.71537957785502e-06,
+      "loss": 0.2854,
+      "step": 30624
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.715043742548655e-06,
+      "loss": 0.3483,
+      "step": 30625
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.714707907242288e-06,
+      "loss": 0.2821,
+      "step": 30626
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.714372071935923e-06,
+      "loss": 0.3639,
+      "step": 30627
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.714036236629557e-06,
+      "loss": 0.4552,
+      "step": 30628
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.713700401323192e-06,
+      "loss": 0.2697,
+      "step": 30629
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.713364566016827e-06,
+      "loss": 0.1825,
+      "step": 30630
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.713028730710461e-06,
+      "loss": 0.2338,
+      "step": 30631
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.712692895404094e-06,
+      "loss": 0.3164,
+      "step": 30632
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.712357060097729e-06,
+      "loss": 0.3612,
+      "step": 30633
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.712021224791362e-06,
+      "loss": 0.3988,
+      "step": 30634
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.711685389484997e-06,
+      "loss": 0.2618,
+      "step": 30635
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.711349554178631e-06,
+      "loss": 0.3562,
+      "step": 30636
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.711013718872266e-06,
+      "loss": 0.2141,
+      "step": 30637
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.7106778835659e-06,
+      "loss": 0.43,
+      "step": 30638
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.710342048259535e-06,
+      "loss": 0.2413,
+      "step": 30639
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.710006212953168e-06,
+      "loss": 0.1876,
+      "step": 30640
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.709670377646803e-06,
+      "loss": 0.324,
+      "step": 30641
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.709334542340437e-06,
+      "loss": 0.2352,
+      "step": 30642
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.70899870703407e-06,
+      "loss": 0.2522,
+      "step": 30643
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.708662871727705e-06,
+      "loss": 0.3051,
+      "step": 30644
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.70832703642134e-06,
+      "loss": 0.2305,
+      "step": 30645
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.707991201114973e-06,
+      "loss": 0.4194,
+      "step": 30646
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.707655365808607e-06,
+      "loss": 0.5043,
+      "step": 30647
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.707319530502242e-06,
+      "loss": 0.292,
+      "step": 30648
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.706983695195877e-06,
+      "loss": 0.5027,
+      "step": 30649
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.706647859889511e-06,
+      "loss": 0.1792,
+      "step": 30650
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.706312024583146e-06,
+      "loss": 0.2823,
+      "step": 30651
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.705976189276779e-06,
+      "loss": 0.3509,
+      "step": 30652
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.705640353970413e-06,
+      "loss": 0.2194,
+      "step": 30653
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.705304518664048e-06,
+      "loss": 0.3785,
+      "step": 30654
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.704968683357681e-06,
+      "loss": 0.3793,
+      "step": 30655
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.704632848051316e-06,
+      "loss": 0.2316,
+      "step": 30656
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.70429701274495e-06,
+      "loss": 0.2762,
+      "step": 30657
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.703961177438585e-06,
+      "loss": 0.4223,
+      "step": 30658
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.70362534213222e-06,
+      "loss": 0.2674,
+      "step": 30659
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.703289506825854e-06,
+      "loss": 0.3037,
+      "step": 30660
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.702953671519487e-06,
+      "loss": 0.3298,
+      "step": 30661
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.702617836213122e-06,
+      "loss": 0.2465,
+      "step": 30662
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.702282000906757e-06,
+      "loss": 0.1987,
+      "step": 30663
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.70194616560039e-06,
+      "loss": 0.4082,
+      "step": 30664
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.701610330294024e-06,
+      "loss": 0.4878,
+      "step": 30665
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.701274494987659e-06,
+      "loss": 0.3998,
+      "step": 30666
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.700938659681293e-06,
+      "loss": 0.1709,
+      "step": 30667
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.700602824374928e-06,
+      "loss": 0.2475,
+      "step": 30668
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 9.700266989068561e-06,
+      "loss": 0.2459,
+      "step": 30669
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.699931153762196e-06,
+      "loss": 0.3664,
+      "step": 30670
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.69959531845583e-06,
+      "loss": 0.4401,
+      "step": 30671
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.699259483149465e-06,
+      "loss": 0.4132,
+      "step": 30672
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.698923647843098e-06,
+      "loss": 0.2793,
+      "step": 30673
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.698587812536733e-06,
+      "loss": 0.248,
+      "step": 30674
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.698251977230367e-06,
+      "loss": 0.2639,
+      "step": 30675
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.697916141924e-06,
+      "loss": 0.2893,
+      "step": 30676
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.697580306617635e-06,
+      "loss": 0.2489,
+      "step": 30677
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.69724447131127e-06,
+      "loss": 0.3677,
+      "step": 30678
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.696908636004904e-06,
+      "loss": 0.3108,
+      "step": 30679
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.696572800698539e-06,
+      "loss": 0.3706,
+      "step": 30680
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.696236965392173e-06,
+      "loss": 0.1795,
+      "step": 30681
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.695901130085806e-06,
+      "loss": 0.2627,
+      "step": 30682
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.695565294779441e-06,
+      "loss": 0.3794,
+      "step": 30683
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.695229459473076e-06,
+      "loss": 0.1859,
+      "step": 30684
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.694893624166709e-06,
+      "loss": 0.2572,
+      "step": 30685
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.694557788860343e-06,
+      "loss": 0.323,
+      "step": 30686
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.694221953553978e-06,
+      "loss": 0.3524,
+      "step": 30687
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.693886118247613e-06,
+      "loss": 0.2063,
+      "step": 30688
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.693550282941247e-06,
+      "loss": 0.3844,
+      "step": 30689
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.693214447634882e-06,
+      "loss": 0.256,
+      "step": 30690
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.692878612328515e-06,
+      "loss": 0.4009,
+      "step": 30691
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.69254277702215e-06,
+      "loss": 0.4188,
+      "step": 30692
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.692206941715784e-06,
+      "loss": 0.349,
+      "step": 30693
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.691871106409417e-06,
+      "loss": 0.1768,
+      "step": 30694
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.691535271103052e-06,
+      "loss": 0.204,
+      "step": 30695
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.691199435796686e-06,
+      "loss": 0.2915,
+      "step": 30696
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.69086360049032e-06,
+      "loss": 0.3434,
+      "step": 30697
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.690527765183954e-06,
+      "loss": 0.2405,
+      "step": 30698
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.690191929877589e-06,
+      "loss": 0.3531,
+      "step": 30699
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.689856094571223e-06,
+      "loss": 0.5225,
+      "step": 30700
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.689520259264858e-06,
+      "loss": 0.2257,
+      "step": 30701
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.689184423958493e-06,
+      "loss": 0.2972,
+      "step": 30702
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.688848588652126e-06,
+      "loss": 0.2735,
+      "step": 30703
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.68851275334576e-06,
+      "loss": 0.338,
+      "step": 30704
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.688176918039395e-06,
+      "loss": 0.258,
+      "step": 30705
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.687841082733028e-06,
+      "loss": 0.3277,
+      "step": 30706
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.687505247426662e-06,
+      "loss": 0.6157,
+      "step": 30707
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.687169412120297e-06,
+      "loss": 0.3349,
+      "step": 30708
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.686833576813932e-06,
+      "loss": 0.1557,
+      "step": 30709
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.686497741507566e-06,
+      "loss": 0.3225,
+      "step": 30710
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.686161906201201e-06,
+      "loss": 0.3005,
+      "step": 30711
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.685826070894834e-06,
+      "loss": 0.3188,
+      "step": 30712
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.685490235588469e-06,
+      "loss": 0.4266,
+      "step": 30713
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.685154400282103e-06,
+      "loss": 0.3135,
+      "step": 30714
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.684818564975736e-06,
+      "loss": 0.245,
+      "step": 30715
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.684482729669371e-06,
+      "loss": 0.4016,
+      "step": 30716
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.684146894363006e-06,
+      "loss": 0.3724,
+      "step": 30717
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.683811059056638e-06,
+      "loss": 0.3019,
+      "step": 30718
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.683475223750273e-06,
+      "loss": 0.4448,
+      "step": 30719
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.683139388443908e-06,
+      "loss": 0.382,
+      "step": 30720
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.682803553137542e-06,
+      "loss": 0.148,
+      "step": 30721
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.682467717831177e-06,
+      "loss": 0.2876,
+      "step": 30722
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.682131882524812e-06,
+      "loss": 0.1843,
+      "step": 30723
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.681796047218445e-06,
+      "loss": 0.2599,
+      "step": 30724
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.68146021191208e-06,
+      "loss": 0.3579,
+      "step": 30725
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.681124376605712e-06,
+      "loss": 0.2056,
+      "step": 30726
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.680788541299347e-06,
+      "loss": 0.236,
+      "step": 30727
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.680452705992982e-06,
+      "loss": 0.28,
+      "step": 30728
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.680116870686616e-06,
+      "loss": 0.3557,
+      "step": 30729
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.679781035380251e-06,
+      "loss": 0.2507,
+      "step": 30730
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.679445200073886e-06,
+      "loss": 0.1882,
+      "step": 30731
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.679109364767518e-06,
+      "loss": 0.3822,
+      "step": 30732
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.678773529461153e-06,
+      "loss": 0.3184,
+      "step": 30733
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.678437694154788e-06,
+      "loss": 0.3543,
+      "step": 30734
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.67810185884842e-06,
+      "loss": 0.3746,
+      "step": 30735
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.677766023542055e-06,
+      "loss": 0.3114,
+      "step": 30736
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.67743018823569e-06,
+      "loss": 0.4251,
+      "step": 30737
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.677094352929323e-06,
+      "loss": 0.274,
+      "step": 30738
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.676758517622958e-06,
+      "loss": 0.2338,
+      "step": 30739
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.676422682316592e-06,
+      "loss": 0.3905,
+      "step": 30740
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.676086847010227e-06,
+      "loss": 0.4326,
+      "step": 30741
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.675751011703862e-06,
+      "loss": 0.3693,
+      "step": 30742
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.675415176397496e-06,
+      "loss": 0.2113,
+      "step": 30743
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.675079341091129e-06,
+      "loss": 0.3574,
+      "step": 30744
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.674743505784764e-06,
+      "loss": 0.2731,
+      "step": 30745
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.674407670478398e-06,
+      "loss": 0.3481,
+      "step": 30746
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.674071835172031e-06,
+      "loss": 0.3141,
+      "step": 30747
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.673735999865666e-06,
+      "loss": 0.2041,
+      "step": 30748
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.6734001645593e-06,
+      "loss": 0.3137,
+      "step": 30749
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.673064329252935e-06,
+      "loss": 0.2452,
+      "step": 30750
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.67272849394657e-06,
+      "loss": 0.2307,
+      "step": 30751
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.672392658640205e-06,
+      "loss": 0.1828,
+      "step": 30752
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.672056823333838e-06,
+      "loss": 0.2715,
+      "step": 30753
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.671720988027472e-06,
+      "loss": 0.3771,
+      "step": 30754
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.671385152721107e-06,
+      "loss": 0.3551,
+      "step": 30755
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.67104931741474e-06,
+      "loss": 0.3394,
+      "step": 30756
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.670713482108374e-06,
+      "loss": 0.3244,
+      "step": 30757
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.670377646802009e-06,
+      "loss": 0.2435,
+      "step": 30758
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.670041811495642e-06,
+      "loss": 0.15,
+      "step": 30759
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.669705976189277e-06,
+      "loss": 0.3733,
+      "step": 30760
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.669370140882911e-06,
+      "loss": 0.3077,
+      "step": 30761
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.669034305576546e-06,
+      "loss": 0.1583,
+      "step": 30762
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.66869847027018e-06,
+      "loss": 0.3238,
+      "step": 30763
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.668362634963815e-06,
+      "loss": 0.4077,
+      "step": 30764
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.668026799657448e-06,
+      "loss": 0.2402,
+      "step": 30765
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.667690964351083e-06,
+      "loss": 0.1713,
+      "step": 30766
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.667355129044718e-06,
+      "loss": 0.3831,
+      "step": 30767
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.66701929373835e-06,
+      "loss": 0.2796,
+      "step": 30768
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.666683458431985e-06,
+      "loss": 0.2266,
+      "step": 30769
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.66634762312562e-06,
+      "loss": 0.1847,
+      "step": 30770
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.666011787819254e-06,
+      "loss": 0.2448,
+      "step": 30771
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.665675952512889e-06,
+      "loss": 0.2571,
+      "step": 30772
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.665340117206524e-06,
+      "loss": 0.2441,
+      "step": 30773
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.665004281900157e-06,
+      "loss": 0.3647,
+      "step": 30774
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.664668446593791e-06,
+      "loss": 0.2288,
+      "step": 30775
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.664332611287426e-06,
+      "loss": 0.2078,
+      "step": 30776
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.663996775981059e-06,
+      "loss": 0.2236,
+      "step": 30777
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.663660940674694e-06,
+      "loss": 0.3879,
+      "step": 30778
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.663325105368328e-06,
+      "loss": 0.1468,
+      "step": 30779
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.662989270061963e-06,
+      "loss": 0.2871,
+      "step": 30780
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.662653434755598e-06,
+      "loss": 0.2908,
+      "step": 30781
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.66231759944923e-06,
+      "loss": 0.1987,
+      "step": 30782
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.661981764142865e-06,
+      "loss": 0.3279,
+      "step": 30783
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.6616459288365e-06,
+      "loss": 0.1855,
+      "step": 30784
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.661310093530134e-06,
+      "loss": 0.1623,
+      "step": 30785
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.660974258223767e-06,
+      "loss": 0.3781,
+      "step": 30786
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.660638422917402e-06,
+      "loss": 0.3082,
+      "step": 30787
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.660302587611037e-06,
+      "loss": 0.307,
+      "step": 30788
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.65996675230467e-06,
+      "loss": 0.2573,
+      "step": 30789
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.659630916998304e-06,
+      "loss": 0.1467,
+      "step": 30790
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.659295081691939e-06,
+      "loss": 0.2167,
+      "step": 30791
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.658959246385574e-06,
+      "loss": 0.1866,
+      "step": 30792
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.658623411079208e-06,
+      "loss": 0.2982,
+      "step": 30793
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.658287575772843e-06,
+      "loss": 0.3274,
+      "step": 30794
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.657951740466476e-06,
+      "loss": 0.2851,
+      "step": 30795
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.65761590516011e-06,
+      "loss": 0.5091,
+      "step": 30796
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.657280069853745e-06,
+      "loss": 0.3162,
+      "step": 30797
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.656944234547378e-06,
+      "loss": 0.2992,
+      "step": 30798
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.656608399241013e-06,
+      "loss": 0.3198,
+      "step": 30799
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.656272563934647e-06,
+      "loss": 0.1677,
+      "step": 30800
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.655936728628282e-06,
+      "loss": 0.3686,
+      "step": 30801
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.655600893321917e-06,
+      "loss": 0.1918,
+      "step": 30802
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.655265058015551e-06,
+      "loss": 0.3301,
+      "step": 30803
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.654929222709184e-06,
+      "loss": 0.2317,
+      "step": 30804
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.654593387402819e-06,
+      "loss": 0.229,
+      "step": 30805
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.654257552096454e-06,
+      "loss": 0.2466,
+      "step": 30806
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.653921716790087e-06,
+      "loss": 0.3059,
+      "step": 30807
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.653585881483721e-06,
+      "loss": 0.3111,
+      "step": 30808
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.653250046177356e-06,
+      "loss": 0.3213,
+      "step": 30809
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.652914210870989e-06,
+      "loss": 0.4041,
+      "step": 30810
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.652578375564623e-06,
+      "loss": 0.3292,
+      "step": 30811
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.652242540258258e-06,
+      "loss": 0.3006,
+      "step": 30812
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.651906704951893e-06,
+      "loss": 0.3021,
+      "step": 30813
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.651570869645527e-06,
+      "loss": 0.3067,
+      "step": 30814
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.65123503433916e-06,
+      "loss": 0.2612,
+      "step": 30815
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.650899199032795e-06,
+      "loss": 0.2925,
+      "step": 30816
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.65056336372643e-06,
+      "loss": 0.4225,
+      "step": 30817
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.650227528420063e-06,
+      "loss": 0.4081,
+      "step": 30818
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.649891693113697e-06,
+      "loss": 0.3007,
+      "step": 30819
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.649555857807332e-06,
+      "loss": 0.4851,
+      "step": 30820
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.649220022500966e-06,
+      "loss": 0.2087,
+      "step": 30821
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.648884187194601e-06,
+      "loss": 0.3381,
+      "step": 30822
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.648548351888236e-06,
+      "loss": 0.2846,
+      "step": 30823
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.648212516581869e-06,
+      "loss": 0.2377,
+      "step": 30824
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.647876681275503e-06,
+      "loss": 0.3071,
+      "step": 30825
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.647540845969138e-06,
+      "loss": 0.3876,
+      "step": 30826
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.647205010662771e-06,
+      "loss": 0.3698,
+      "step": 30827
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.646869175356406e-06,
+      "loss": 0.3645,
+      "step": 30828
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.64653334005004e-06,
+      "loss": 0.2672,
+      "step": 30829
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.646197504743673e-06,
+      "loss": 0.2935,
+      "step": 30830
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.645861669437308e-06,
+      "loss": 0.3727,
+      "step": 30831
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.645525834130943e-06,
+      "loss": 0.1752,
+      "step": 30832
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.645189998824577e-06,
+      "loss": 0.3289,
+      "step": 30833
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.644854163518212e-06,
+      "loss": 0.463,
+      "step": 30834
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.644518328211846e-06,
+      "loss": 0.3063,
+      "step": 30835
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.64418249290548e-06,
+      "loss": 0.4154,
+      "step": 30836
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.643846657599114e-06,
+      "loss": 0.3547,
+      "step": 30837
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.643510822292749e-06,
+      "loss": 0.3088,
+      "step": 30838
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.643174986986382e-06,
+      "loss": 0.2429,
+      "step": 30839
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.642839151680016e-06,
+      "loss": 0.3368,
+      "step": 30840
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.642503316373651e-06,
+      "loss": 0.2272,
+      "step": 30841
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.642167481067286e-06,
+      "loss": 0.2868,
+      "step": 30842
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.64183164576092e-06,
+      "loss": 0.2006,
+      "step": 30843
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.641495810454555e-06,
+      "loss": 0.2828,
+      "step": 30844
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.641159975148188e-06,
+      "loss": 0.2446,
+      "step": 30845
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.640824139841823e-06,
+      "loss": 0.3671,
+      "step": 30846
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.640488304535457e-06,
+      "loss": 0.2398,
+      "step": 30847
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.64015246922909e-06,
+      "loss": 0.2776,
+      "step": 30848
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.639816633922725e-06,
+      "loss": 0.3317,
+      "step": 30849
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.63948079861636e-06,
+      "loss": 0.3405,
+      "step": 30850
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.639144963309992e-06,
+      "loss": 0.3404,
+      "step": 30851
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.638809128003627e-06,
+      "loss": 0.2719,
+      "step": 30852
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.638473292697262e-06,
+      "loss": 0.328,
+      "step": 30853
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.638137457390896e-06,
+      "loss": 0.2492,
+      "step": 30854
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.637801622084531e-06,
+      "loss": 0.3819,
+      "step": 30855
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.637465786778166e-06,
+      "loss": 0.2612,
+      "step": 30856
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.637129951471799e-06,
+      "loss": 0.1978,
+      "step": 30857
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.636794116165433e-06,
+      "loss": 0.234,
+      "step": 30858
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.636458280859068e-06,
+      "loss": 0.2607,
+      "step": 30859
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.6361224455527e-06,
+      "loss": 0.3249,
+      "step": 30860
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.635786610246335e-06,
+      "loss": 0.3109,
+      "step": 30861
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.63545077493997e-06,
+      "loss": 0.2467,
+      "step": 30862
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.635114939633605e-06,
+      "loss": 0.4646,
+      "step": 30863
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.63477910432724e-06,
+      "loss": 0.3458,
+      "step": 30864
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.634443269020874e-06,
+      "loss": 0.1798,
+      "step": 30865
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.634107433714507e-06,
+      "loss": 0.4167,
+      "step": 30866
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.633771598408142e-06,
+      "loss": 0.3488,
+      "step": 30867
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 9.633435763101776e-06,
+      "loss": 0.2978,
+      "step": 30868
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.63309992779541e-06,
+      "loss": 0.4279,
+      "step": 30869
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.632764092489044e-06,
+      "loss": 0.2273,
+      "step": 30870
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.632428257182679e-06,
+      "loss": 0.1771,
+      "step": 30871
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.632092421876311e-06,
+      "loss": 0.3645,
+      "step": 30872
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.631756586569946e-06,
+      "loss": 0.3802,
+      "step": 30873
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.63142075126358e-06,
+      "loss": 0.3327,
+      "step": 30874
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.631084915957215e-06,
+      "loss": 0.3961,
+      "step": 30875
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.63074908065085e-06,
+      "loss": 0.4162,
+      "step": 30876
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.630413245344485e-06,
+      "loss": 0.3651,
+      "step": 30877
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.630077410038118e-06,
+      "loss": 0.2788,
+      "step": 30878
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.629741574731752e-06,
+      "loss": 0.2961,
+      "step": 30879
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.629405739425387e-06,
+      "loss": 0.2313,
+      "step": 30880
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.62906990411902e-06,
+      "loss": 0.3541,
+      "step": 30881
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.628734068812655e-06,
+      "loss": 0.2928,
+      "step": 30882
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.62839823350629e-06,
+      "loss": 0.2749,
+      "step": 30883
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.628062398199924e-06,
+      "loss": 0.2764,
+      "step": 30884
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.627726562893559e-06,
+      "loss": 0.3102,
+      "step": 30885
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.627390727587193e-06,
+      "loss": 0.2909,
+      "step": 30886
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.627054892280826e-06,
+      "loss": 0.2562,
+      "step": 30887
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.62671905697446e-06,
+      "loss": 0.4036,
+      "step": 30888
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.626383221668095e-06,
+      "loss": 0.3725,
+      "step": 30889
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.626047386361728e-06,
+      "loss": 0.19,
+      "step": 30890
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.625711551055363e-06,
+      "loss": 0.2212,
+      "step": 30891
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.625375715748998e-06,
+      "loss": 0.4032,
+      "step": 30892
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.625039880442632e-06,
+      "loss": 0.4879,
+      "step": 30893
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.624704045136265e-06,
+      "loss": 0.2181,
+      "step": 30894
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.6243682098299e-06,
+      "loss": 0.178,
+      "step": 30895
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.624032374523535e-06,
+      "loss": 0.4426,
+      "step": 30896
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.62369653921717e-06,
+      "loss": 0.2404,
+      "step": 30897
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.623360703910804e-06,
+      "loss": 0.3357,
+      "step": 30898
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.623024868604437e-06,
+      "loss": 0.2558,
+      "step": 30899
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.622689033298071e-06,
+      "loss": 0.1993,
+      "step": 30900
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.622353197991706e-06,
+      "loss": 0.1974,
+      "step": 30901
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.622017362685339e-06,
+      "loss": 0.1413,
+      "step": 30902
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.621681527378974e-06,
+      "loss": 0.253,
+      "step": 30903
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.621345692072608e-06,
+      "loss": 0.2335,
+      "step": 30904
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.621009856766243e-06,
+      "loss": 0.4525,
+      "step": 30905
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.620674021459878e-06,
+      "loss": 0.2903,
+      "step": 30906
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.62033818615351e-06,
+      "loss": 0.3743,
+      "step": 30907
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.620002350847145e-06,
+      "loss": 0.1998,
+      "step": 30908
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.61966651554078e-06,
+      "loss": 0.3499,
+      "step": 30909
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.619330680234413e-06,
+      "loss": 0.1942,
+      "step": 30910
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.618994844928047e-06,
+      "loss": 0.4362,
+      "step": 30911
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.618659009621682e-06,
+      "loss": 0.3506,
+      "step": 30912
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.618323174315317e-06,
+      "loss": 0.3073,
+      "step": 30913
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.617987339008951e-06,
+      "loss": 0.2413,
+      "step": 30914
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.617651503702586e-06,
+      "loss": 0.3466,
+      "step": 30915
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.617315668396219e-06,
+      "loss": 0.2512,
+      "step": 30916
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.616979833089854e-06,
+      "loss": 0.4131,
+      "step": 30917
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.616643997783488e-06,
+      "loss": 0.2548,
+      "step": 30918
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.616308162477121e-06,
+      "loss": 0.2762,
+      "step": 30919
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.615972327170756e-06,
+      "loss": 0.2206,
+      "step": 30920
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.61563649186439e-06,
+      "loss": 0.3764,
+      "step": 30921
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.615300656558024e-06,
+      "loss": 0.3167,
+      "step": 30922
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.614964821251658e-06,
+      "loss": 0.2817,
+      "step": 30923
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.614628985945293e-06,
+      "loss": 0.1594,
+      "step": 30924
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.614293150638927e-06,
+      "loss": 0.2522,
+      "step": 30925
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.613957315332562e-06,
+      "loss": 0.2297,
+      "step": 30926
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.613621480026197e-06,
+      "loss": 0.1961,
+      "step": 30927
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.61328564471983e-06,
+      "loss": 0.3143,
+      "step": 30928
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.612949809413464e-06,
+      "loss": 0.2934,
+      "step": 30929
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.612613974107099e-06,
+      "loss": 0.5584,
+      "step": 30930
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.612278138800732e-06,
+      "loss": 0.1683,
+      "step": 30931
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.611942303494367e-06,
+      "loss": 0.3398,
+      "step": 30932
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.611606468188001e-06,
+      "loss": 0.1684,
+      "step": 30933
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.611270632881636e-06,
+      "loss": 0.2342,
+      "step": 30934
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.61093479757527e-06,
+      "loss": 0.4131,
+      "step": 30935
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.610598962268905e-06,
+      "loss": 0.2967,
+      "step": 30936
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.610263126962538e-06,
+      "loss": 0.2554,
+      "step": 30937
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.609927291656173e-06,
+      "loss": 0.284,
+      "step": 30938
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.609591456349807e-06,
+      "loss": 0.181,
+      "step": 30939
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.60925562104344e-06,
+      "loss": 0.2232,
+      "step": 30940
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.608919785737075e-06,
+      "loss": 0.2706,
+      "step": 30941
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.60858395043071e-06,
+      "loss": 0.432,
+      "step": 30942
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.608248115124343e-06,
+      "loss": 0.206,
+      "step": 30943
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.607912279817977e-06,
+      "loss": 0.1588,
+      "step": 30944
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.607576444511612e-06,
+      "loss": 0.2071,
+      "step": 30945
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.607240609205247e-06,
+      "loss": 0.4129,
+      "step": 30946
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.606904773898881e-06,
+      "loss": 0.4375,
+      "step": 30947
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.606568938592516e-06,
+      "loss": 0.4266,
+      "step": 30948
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.606233103286149e-06,
+      "loss": 0.1341,
+      "step": 30949
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.605897267979783e-06,
+      "loss": 0.2885,
+      "step": 30950
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.605561432673418e-06,
+      "loss": 0.3149,
+      "step": 30951
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.605225597367051e-06,
+      "loss": 0.4145,
+      "step": 30952
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.604889762060686e-06,
+      "loss": 0.2759,
+      "step": 30953
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.60455392675432e-06,
+      "loss": 0.3362,
+      "step": 30954
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.604218091447955e-06,
+      "loss": 0.2829,
+      "step": 30955
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.60388225614159e-06,
+      "loss": 0.2544,
+      "step": 30956
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.603546420835224e-06,
+      "loss": 0.3141,
+      "step": 30957
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.603210585528857e-06,
+      "loss": 0.4265,
+      "step": 30958
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.602874750222492e-06,
+      "loss": 0.2815,
+      "step": 30959
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.602538914916127e-06,
+      "loss": 0.3263,
+      "step": 30960
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.60220307960976e-06,
+      "loss": 0.4355,
+      "step": 30961
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.601867244303394e-06,
+      "loss": 0.2565,
+      "step": 30962
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.601531408997029e-06,
+      "loss": 0.2704,
+      "step": 30963
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.601195573690662e-06,
+      "loss": 0.1822,
+      "step": 30964
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.600859738384296e-06,
+      "loss": 0.2461,
+      "step": 30965
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.600523903077931e-06,
+      "loss": 0.2677,
+      "step": 30966
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.600188067771566e-06,
+      "loss": 0.2534,
+      "step": 30967
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.5998522324652e-06,
+      "loss": 0.3608,
+      "step": 30968
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.599516397158835e-06,
+      "loss": 0.3194,
+      "step": 30969
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.599180561852468e-06,
+      "loss": 0.3317,
+      "step": 30970
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.598844726546103e-06,
+      "loss": 0.3189,
+      "step": 30971
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.598508891239737e-06,
+      "loss": 0.3258,
+      "step": 30972
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.59817305593337e-06,
+      "loss": 0.2893,
+      "step": 30973
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.597837220627005e-06,
+      "loss": 0.3404,
+      "step": 30974
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.59750138532064e-06,
+      "loss": 0.2183,
+      "step": 30975
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.597165550014274e-06,
+      "loss": 0.3841,
+      "step": 30976
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.596829714707909e-06,
+      "loss": 0.2595,
+      "step": 30977
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.596493879401543e-06,
+      "loss": 0.4236,
+      "step": 30978
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.596158044095176e-06,
+      "loss": 0.3048,
+      "step": 30979
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.595822208788811e-06,
+      "loss": 0.369,
+      "step": 30980
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.595486373482446e-06,
+      "loss": 0.2367,
+      "step": 30981
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.595150538176079e-06,
+      "loss": 0.2704,
+      "step": 30982
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.594814702869713e-06,
+      "loss": 0.2187,
+      "step": 30983
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.594478867563348e-06,
+      "loss": 0.2056,
+      "step": 30984
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.594143032256981e-06,
+      "loss": 0.4125,
+      "step": 30985
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.593807196950616e-06,
+      "loss": 0.2944,
+      "step": 30986
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.59347136164425e-06,
+      "loss": 0.2114,
+      "step": 30987
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.593135526337885e-06,
+      "loss": 0.3661,
+      "step": 30988
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.59279969103152e-06,
+      "loss": 0.2288,
+      "step": 30989
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.592463855725154e-06,
+      "loss": 0.2618,
+      "step": 30990
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.592128020418787e-06,
+      "loss": 0.4622,
+      "step": 30991
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.591792185112422e-06,
+      "loss": 0.3494,
+      "step": 30992
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.591456349806056e-06,
+      "loss": 0.3884,
+      "step": 30993
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.59112051449969e-06,
+      "loss": 0.1528,
+      "step": 30994
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.590784679193324e-06,
+      "loss": 0.3482,
+      "step": 30995
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.590448843886959e-06,
+      "loss": 0.2486,
+      "step": 30996
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.590113008580593e-06,
+      "loss": 0.1876,
+      "step": 30997
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.589777173274228e-06,
+      "loss": 0.2219,
+      "step": 30998
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.589441337967861e-06,
+      "loss": 0.3793,
+      "step": 30999
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.589105502661496e-06,
+      "loss": 0.3163,
+      "step": 31000
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.58876966735513e-06,
+      "loss": 0.226,
+      "step": 31001
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.588433832048763e-06,
+      "loss": 0.4133,
+      "step": 31002
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.588097996742398e-06,
+      "loss": 0.1684,
+      "step": 31003
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.587762161436032e-06,
+      "loss": 0.307,
+      "step": 31004
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.587426326129667e-06,
+      "loss": 0.1748,
+      "step": 31005
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.587090490823302e-06,
+      "loss": 0.3846,
+      "step": 31006
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.586754655516935e-06,
+      "loss": 0.3995,
+      "step": 31007
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.58641882021057e-06,
+      "loss": 0.2698,
+      "step": 31008
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.586082984904204e-06,
+      "loss": 0.4164,
+      "step": 31009
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.585747149597839e-06,
+      "loss": 0.3545,
+      "step": 31010
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.585411314291472e-06,
+      "loss": 0.2213,
+      "step": 31011
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.585075478985106e-06,
+      "loss": 0.1858,
+      "step": 31012
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.584739643678741e-06,
+      "loss": 0.1985,
+      "step": 31013
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.584403808372374e-06,
+      "loss": 0.1928,
+      "step": 31014
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.584067973066008e-06,
+      "loss": 0.2595,
+      "step": 31015
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.583732137759643e-06,
+      "loss": 0.2741,
+      "step": 31016
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.583396302453278e-06,
+      "loss": 0.2656,
+      "step": 31017
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.583060467146912e-06,
+      "loss": 0.367,
+      "step": 31018
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.582724631840547e-06,
+      "loss": 0.2834,
+      "step": 31019
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.58238879653418e-06,
+      "loss": 0.4066,
+      "step": 31020
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.582052961227815e-06,
+      "loss": 0.1927,
+      "step": 31021
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.58171712592145e-06,
+      "loss": 0.2361,
+      "step": 31022
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.581381290615082e-06,
+      "loss": 0.3344,
+      "step": 31023
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.581045455308717e-06,
+      "loss": 0.3556,
+      "step": 31024
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.580709620002352e-06,
+      "loss": 0.2953,
+      "step": 31025
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.580373784695986e-06,
+      "loss": 0.4229,
+      "step": 31026
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.580037949389621e-06,
+      "loss": 0.2788,
+      "step": 31027
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.579702114083255e-06,
+      "loss": 0.2712,
+      "step": 31028
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.579366278776888e-06,
+      "loss": 0.2705,
+      "step": 31029
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.579030443470523e-06,
+      "loss": 0.2715,
+      "step": 31030
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.578694608164158e-06,
+      "loss": 0.2424,
+      "step": 31031
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.57835877285779e-06,
+      "loss": 0.3545,
+      "step": 31032
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.578022937551425e-06,
+      "loss": 0.4131,
+      "step": 31033
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.57768710224506e-06,
+      "loss": 0.5478,
+      "step": 31034
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.577351266938693e-06,
+      "loss": 0.2343,
+      "step": 31035
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.577015431632328e-06,
+      "loss": 0.3183,
+      "step": 31036
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.576679596325962e-06,
+      "loss": 0.3986,
+      "step": 31037
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.576343761019597e-06,
+      "loss": 0.1541,
+      "step": 31038
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.576007925713232e-06,
+      "loss": 0.4278,
+      "step": 31039
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.575672090406866e-06,
+      "loss": 0.3717,
+      "step": 31040
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.575336255100499e-06,
+      "loss": 0.3248,
+      "step": 31041
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.575000419794134e-06,
+      "loss": 0.1972,
+      "step": 31042
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.574664584487768e-06,
+      "loss": 0.1916,
+      "step": 31043
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.574328749181401e-06,
+      "loss": 0.3815,
+      "step": 31044
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.573992913875036e-06,
+      "loss": 0.2078,
+      "step": 31045
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.57365707856867e-06,
+      "loss": 0.2271,
+      "step": 31046
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.573321243262305e-06,
+      "loss": 0.3309,
+      "step": 31047
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.57298540795594e-06,
+      "loss": 0.3732,
+      "step": 31048
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.572649572649575e-06,
+      "loss": 0.2838,
+      "step": 31049
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.572313737343208e-06,
+      "loss": 0.3483,
+      "step": 31050
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.571977902036842e-06,
+      "loss": 0.3687,
+      "step": 31051
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.571642066730477e-06,
+      "loss": 0.3358,
+      "step": 31052
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.57130623142411e-06,
+      "loss": 0.4737,
+      "step": 31053
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.570970396117744e-06,
+      "loss": 0.2198,
+      "step": 31054
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.570634560811379e-06,
+      "loss": 0.2913,
+      "step": 31055
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.570298725505012e-06,
+      "loss": 0.3161,
+      "step": 31056
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.569962890198647e-06,
+      "loss": 0.2763,
+      "step": 31057
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.569627054892281e-06,
+      "loss": 0.2929,
+      "step": 31058
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.569291219585916e-06,
+      "loss": 0.3108,
+      "step": 31059
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.56895538427955e-06,
+      "loss": 0.4373,
+      "step": 31060
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.568619548973185e-06,
+      "loss": 0.2413,
+      "step": 31061
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.568283713666818e-06,
+      "loss": 0.2958,
+      "step": 31062
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.567947878360453e-06,
+      "loss": 0.2827,
+      "step": 31063
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.567612043054088e-06,
+      "loss": 0.2411,
+      "step": 31064
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.56727620774772e-06,
+      "loss": 0.4324,
+      "step": 31065
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 9.566940372441355e-06,
+      "loss": 0.2021,
+      "step": 31066
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.56660453713499e-06,
+      "loss": 0.2254,
+      "step": 31067
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.566268701828624e-06,
+      "loss": 0.386,
+      "step": 31068
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.565932866522259e-06,
+      "loss": 0.2852,
+      "step": 31069
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.565597031215894e-06,
+      "loss": 0.2878,
+      "step": 31070
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.565261195909527e-06,
+      "loss": 0.1785,
+      "step": 31071
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.564925360603161e-06,
+      "loss": 0.3593,
+      "step": 31072
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.564589525296796e-06,
+      "loss": 0.334,
+      "step": 31073
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.564253689990429e-06,
+      "loss": 0.2928,
+      "step": 31074
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.563917854684064e-06,
+      "loss": 0.2787,
+      "step": 31075
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.563582019377698e-06,
+      "loss": 0.4488,
+      "step": 31076
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.563246184071331e-06,
+      "loss": 0.4314,
+      "step": 31077
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.562910348764966e-06,
+      "loss": 0.3968,
+      "step": 31078
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.5625745134586e-06,
+      "loss": 0.4885,
+      "step": 31079
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.562238678152235e-06,
+      "loss": 0.2407,
+      "step": 31080
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.56190284284587e-06,
+      "loss": 0.2882,
+      "step": 31081
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.561567007539504e-06,
+      "loss": 0.375,
+      "step": 31082
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.561231172233137e-06,
+      "loss": 0.3642,
+      "step": 31083
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.560895336926772e-06,
+      "loss": 0.1896,
+      "step": 31084
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.560559501620405e-06,
+      "loss": 0.2612,
+      "step": 31085
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.56022366631404e-06,
+      "loss": 0.3091,
+      "step": 31086
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.559887831007674e-06,
+      "loss": 0.3217,
+      "step": 31087
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.559551995701309e-06,
+      "loss": 0.3057,
+      "step": 31088
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.559216160394944e-06,
+      "loss": 0.1804,
+      "step": 31089
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.558880325088578e-06,
+      "loss": 0.2533,
+      "step": 31090
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.558544489782211e-06,
+      "loss": 0.2721,
+      "step": 31091
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.558208654475846e-06,
+      "loss": 0.2723,
+      "step": 31092
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.55787281916948e-06,
+      "loss": 0.3049,
+      "step": 31093
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.557536983863113e-06,
+      "loss": 0.3116,
+      "step": 31094
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.557201148556748e-06,
+      "loss": 0.4069,
+      "step": 31095
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.556865313250383e-06,
+      "loss": 0.3173,
+      "step": 31096
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.556529477944017e-06,
+      "loss": 0.3314,
+      "step": 31097
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.55619364263765e-06,
+      "loss": 0.2752,
+      "step": 31098
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.555857807331285e-06,
+      "loss": 0.2483,
+      "step": 31099
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.55552197202492e-06,
+      "loss": 0.3938,
+      "step": 31100
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.555186136718554e-06,
+      "loss": 0.2937,
+      "step": 31101
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.554850301412189e-06,
+      "loss": 0.3814,
+      "step": 31102
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.554514466105822e-06,
+      "loss": 0.1569,
+      "step": 31103
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.554178630799457e-06,
+      "loss": 0.1902,
+      "step": 31104
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.553842795493091e-06,
+      "loss": 0.3394,
+      "step": 31105
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.553506960186724e-06,
+      "loss": 0.228,
+      "step": 31106
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.553171124880359e-06,
+      "loss": 0.4065,
+      "step": 31107
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.552835289573993e-06,
+      "loss": 0.2793,
+      "step": 31108
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.552499454267628e-06,
+      "loss": 0.196,
+      "step": 31109
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.552163618961263e-06,
+      "loss": 0.2747,
+      "step": 31110
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.551827783654897e-06,
+      "loss": 0.2214,
+      "step": 31111
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.55149194834853e-06,
+      "loss": 0.3307,
+      "step": 31112
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.551156113042165e-06,
+      "loss": 0.3837,
+      "step": 31113
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.5508202777358e-06,
+      "loss": 0.2362,
+      "step": 31114
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.550484442429433e-06,
+      "loss": 0.2303,
+      "step": 31115
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.550148607123067e-06,
+      "loss": 0.2404,
+      "step": 31116
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.549812771816702e-06,
+      "loss": 0.3599,
+      "step": 31117
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.549476936510336e-06,
+      "loss": 0.2046,
+      "step": 31118
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.549141101203971e-06,
+      "loss": 0.3417,
+      "step": 31119
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.548805265897604e-06,
+      "loss": 0.4181,
+      "step": 31120
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.548469430591239e-06,
+      "loss": 0.2418,
+      "step": 31121
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.548133595284873e-06,
+      "loss": 0.336,
+      "step": 31122
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.547797759978508e-06,
+      "loss": 0.2705,
+      "step": 31123
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.547461924672141e-06,
+      "loss": 0.208,
+      "step": 31124
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.547126089365776e-06,
+      "loss": 0.3977,
+      "step": 31125
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.54679025405941e-06,
+      "loss": 0.3328,
+      "step": 31126
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.546454418753043e-06,
+      "loss": 0.3766,
+      "step": 31127
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.546118583446678e-06,
+      "loss": 0.3684,
+      "step": 31128
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.545782748140313e-06,
+      "loss": 0.2609,
+      "step": 31129
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.545446912833947e-06,
+      "loss": 0.263,
+      "step": 31130
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.545111077527582e-06,
+      "loss": 0.2556,
+      "step": 31131
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.544775242221216e-06,
+      "loss": 0.3094,
+      "step": 31132
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.54443940691485e-06,
+      "loss": 0.2626,
+      "step": 31133
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.544103571608484e-06,
+      "loss": 0.3416,
+      "step": 31134
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.543767736302119e-06,
+      "loss": 0.2606,
+      "step": 31135
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.543431900995752e-06,
+      "loss": 0.2914,
+      "step": 31136
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.543096065689386e-06,
+      "loss": 0.3595,
+      "step": 31137
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.542760230383021e-06,
+      "loss": 0.3352,
+      "step": 31138
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.542424395076656e-06,
+      "loss": 0.2964,
+      "step": 31139
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.54208855977029e-06,
+      "loss": 0.2085,
+      "step": 31140
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.541752724463925e-06,
+      "loss": 0.3279,
+      "step": 31141
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.541416889157558e-06,
+      "loss": 0.31,
+      "step": 31142
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.541081053851193e-06,
+      "loss": 0.27,
+      "step": 31143
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.540745218544827e-06,
+      "loss": 0.2656,
+      "step": 31144
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.54040938323846e-06,
+      "loss": 0.1483,
+      "step": 31145
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.540073547932095e-06,
+      "loss": 0.2396,
+      "step": 31146
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.53973771262573e-06,
+      "loss": 0.1569,
+      "step": 31147
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.539401877319362e-06,
+      "loss": 0.207,
+      "step": 31148
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.539066042012997e-06,
+      "loss": 0.3096,
+      "step": 31149
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.538730206706632e-06,
+      "loss": 0.3032,
+      "step": 31150
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.538394371400266e-06,
+      "loss": 0.1487,
+      "step": 31151
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.538058536093901e-06,
+      "loss": 0.4394,
+      "step": 31152
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.537722700787536e-06,
+      "loss": 0.177,
+      "step": 31153
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.537386865481169e-06,
+      "loss": 0.3782,
+      "step": 31154
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.537051030174803e-06,
+      "loss": 0.3669,
+      "step": 31155
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.536715194868438e-06,
+      "loss": 0.2221,
+      "step": 31156
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.53637935956207e-06,
+      "loss": 0.3101,
+      "step": 31157
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.536043524255705e-06,
+      "loss": 0.2123,
+      "step": 31158
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.53570768894934e-06,
+      "loss": 0.4164,
+      "step": 31159
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.535371853642975e-06,
+      "loss": 0.2145,
+      "step": 31160
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.53503601833661e-06,
+      "loss": 0.3822,
+      "step": 31161
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.534700183030244e-06,
+      "loss": 0.1681,
+      "step": 31162
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.534364347723877e-06,
+      "loss": 0.4735,
+      "step": 31163
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.534028512417512e-06,
+      "loss": 0.2811,
+      "step": 31164
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.533692677111146e-06,
+      "loss": 0.2101,
+      "step": 31165
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.53335684180478e-06,
+      "loss": 0.3679,
+      "step": 31166
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.533021006498414e-06,
+      "loss": 0.2268,
+      "step": 31167
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.532685171192049e-06,
+      "loss": 0.4096,
+      "step": 31168
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.532349335885681e-06,
+      "loss": 0.2089,
+      "step": 31169
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.532013500579316e-06,
+      "loss": 0.4344,
+      "step": 31170
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.53167766527295e-06,
+      "loss": 0.3802,
+      "step": 31171
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.531341829966585e-06,
+      "loss": 0.2244,
+      "step": 31172
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.53100599466022e-06,
+      "loss": 0.3233,
+      "step": 31173
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.530670159353855e-06,
+      "loss": 0.2316,
+      "step": 31174
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.530334324047488e-06,
+      "loss": 0.2455,
+      "step": 31175
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.529998488741122e-06,
+      "loss": 0.2132,
+      "step": 31176
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.529662653434755e-06,
+      "loss": 0.1792,
+      "step": 31177
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.52932681812839e-06,
+      "loss": 0.2139,
+      "step": 31178
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.528990982822025e-06,
+      "loss": 0.2612,
+      "step": 31179
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.52865514751566e-06,
+      "loss": 0.3578,
+      "step": 31180
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.528319312209294e-06,
+      "loss": 0.2784,
+      "step": 31181
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.527983476902929e-06,
+      "loss": 0.2364,
+      "step": 31182
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.527647641596561e-06,
+      "loss": 0.184,
+      "step": 31183
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.527311806290196e-06,
+      "loss": 0.2791,
+      "step": 31184
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.52697597098383e-06,
+      "loss": 0.1953,
+      "step": 31185
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.526640135677464e-06,
+      "loss": 0.1738,
+      "step": 31186
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.526304300371098e-06,
+      "loss": 0.1598,
+      "step": 31187
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.525968465064733e-06,
+      "loss": 0.2283,
+      "step": 31188
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.525632629758366e-06,
+      "loss": 0.2179,
+      "step": 31189
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.525296794452e-06,
+      "loss": 0.296,
+      "step": 31190
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.524960959145635e-06,
+      "loss": 0.3235,
+      "step": 31191
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.52462512383927e-06,
+      "loss": 0.2298,
+      "step": 31192
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.524289288532905e-06,
+      "loss": 0.4018,
+      "step": 31193
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.52395345322654e-06,
+      "loss": 0.1663,
+      "step": 31194
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.523617617920172e-06,
+      "loss": 0.2751,
+      "step": 31195
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.523281782613807e-06,
+      "loss": 0.2965,
+      "step": 31196
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.522945947307441e-06,
+      "loss": 0.2896,
+      "step": 31197
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.522610112001074e-06,
+      "loss": 0.1979,
+      "step": 31198
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.522274276694709e-06,
+      "loss": 0.3249,
+      "step": 31199
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.521938441388344e-06,
+      "loss": 0.2234,
+      "step": 31200
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.521602606081978e-06,
+      "loss": 0.418,
+      "step": 31201
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.521266770775613e-06,
+      "loss": 0.1654,
+      "step": 31202
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.520930935469248e-06,
+      "loss": 0.234,
+      "step": 31203
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.52059510016288e-06,
+      "loss": 0.5493,
+      "step": 31204
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.520259264856515e-06,
+      "loss": 0.2548,
+      "step": 31205
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.51992342955015e-06,
+      "loss": 0.2376,
+      "step": 31206
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.519587594243783e-06,
+      "loss": 0.3409,
+      "step": 31207
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.519251758937417e-06,
+      "loss": 0.3809,
+      "step": 31208
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.518915923631052e-06,
+      "loss": 0.2189,
+      "step": 31209
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.518580088324687e-06,
+      "loss": 0.1594,
+      "step": 31210
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.51824425301832e-06,
+      "loss": 0.2487,
+      "step": 31211
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.517908417711954e-06,
+      "loss": 0.1752,
+      "step": 31212
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.517572582405589e-06,
+      "loss": 0.253,
+      "step": 31213
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.517236747099224e-06,
+      "loss": 0.3615,
+      "step": 31214
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.516900911792858e-06,
+      "loss": 0.2353,
+      "step": 31215
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.516565076486491e-06,
+      "loss": 0.2775,
+      "step": 31216
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.516229241180126e-06,
+      "loss": 0.327,
+      "step": 31217
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.51589340587376e-06,
+      "loss": 0.226,
+      "step": 31218
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.515557570567394e-06,
+      "loss": 0.3696,
+      "step": 31219
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.515221735261028e-06,
+      "loss": 0.2762,
+      "step": 31220
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.514885899954663e-06,
+      "loss": 0.5771,
+      "step": 31221
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.514550064648297e-06,
+      "loss": 0.2742,
+      "step": 31222
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.514214229341932e-06,
+      "loss": 0.2594,
+      "step": 31223
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.513878394035567e-06,
+      "loss": 0.274,
+      "step": 31224
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.5135425587292e-06,
+      "loss": 0.1864,
+      "step": 31225
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.513206723422834e-06,
+      "loss": 0.4529,
+      "step": 31226
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.512870888116469e-06,
+      "loss": 0.1879,
+      "step": 31227
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.512535052810102e-06,
+      "loss": 0.3642,
+      "step": 31228
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.512199217503737e-06,
+      "loss": 0.2345,
+      "step": 31229
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.511863382197371e-06,
+      "loss": 0.2836,
+      "step": 31230
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.511527546891006e-06,
+      "loss": 0.2359,
+      "step": 31231
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.51119171158464e-06,
+      "loss": 0.3025,
+      "step": 31232
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.510855876278274e-06,
+      "loss": 0.181,
+      "step": 31233
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.510520040971908e-06,
+      "loss": 0.3226,
+      "step": 31234
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.510184205665543e-06,
+      "loss": 0.3212,
+      "step": 31235
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.509848370359177e-06,
+      "loss": 0.3006,
+      "step": 31236
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.50951253505281e-06,
+      "loss": 0.3948,
+      "step": 31237
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.509176699746445e-06,
+      "loss": 0.2573,
+      "step": 31238
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.50884086444008e-06,
+      "loss": 0.4073,
+      "step": 31239
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.508505029133713e-06,
+      "loss": 0.295,
+      "step": 31240
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.508169193827347e-06,
+      "loss": 0.4058,
+      "step": 31241
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.507833358520982e-06,
+      "loss": 0.3124,
+      "step": 31242
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.507497523214617e-06,
+      "loss": 0.2883,
+      "step": 31243
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.507161687908251e-06,
+      "loss": 0.4059,
+      "step": 31244
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.506825852601886e-06,
+      "loss": 0.1978,
+      "step": 31245
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.506490017295519e-06,
+      "loss": 0.3283,
+      "step": 31246
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.506154181989153e-06,
+      "loss": 0.3147,
+      "step": 31247
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.505818346682788e-06,
+      "loss": 0.1398,
+      "step": 31248
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.505482511376421e-06,
+      "loss": 0.368,
+      "step": 31249
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.505146676070056e-06,
+      "loss": 0.1683,
+      "step": 31250
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.50481084076369e-06,
+      "loss": 0.2828,
+      "step": 31251
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.504475005457325e-06,
+      "loss": 0.3548,
+      "step": 31252
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.50413917015096e-06,
+      "loss": 0.3045,
+      "step": 31253
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.503803334844594e-06,
+      "loss": 0.3356,
+      "step": 31254
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.503467499538227e-06,
+      "loss": 0.241,
+      "step": 31255
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.503131664231862e-06,
+      "loss": 0.4445,
+      "step": 31256
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.502795828925497e-06,
+      "loss": 0.2444,
+      "step": 31257
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.50245999361913e-06,
+      "loss": 0.3083,
+      "step": 31258
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.502124158312764e-06,
+      "loss": 0.3774,
+      "step": 31259
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.501788323006399e-06,
+      "loss": 0.2373,
+      "step": 31260
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.501452487700032e-06,
+      "loss": 0.2338,
+      "step": 31261
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.501116652393666e-06,
+      "loss": 0.4583,
+      "step": 31262
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.500780817087301e-06,
+      "loss": 0.4536,
+      "step": 31263
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.500444981780936e-06,
+      "loss": 0.2652,
+      "step": 31264
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 9.50010914647457e-06,
+      "loss": 0.2255,
+      "step": 31265
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.499773311168203e-06,
+      "loss": 0.29,
+      "step": 31266
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.499437475861838e-06,
+      "loss": 0.2862,
+      "step": 31267
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.499101640555473e-06,
+      "loss": 0.3448,
+      "step": 31268
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.498765805249106e-06,
+      "loss": 0.2448,
+      "step": 31269
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.49842996994274e-06,
+      "loss": 0.3944,
+      "step": 31270
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.498094134636375e-06,
+      "loss": 0.4371,
+      "step": 31271
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.49775829933001e-06,
+      "loss": 0.3282,
+      "step": 31272
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.497422464023644e-06,
+      "loss": 0.19,
+      "step": 31273
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.497086628717279e-06,
+      "loss": 0.1887,
+      "step": 31274
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.496750793410912e-06,
+      "loss": 0.3394,
+      "step": 31275
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.496414958104546e-06,
+      "loss": 0.3086,
+      "step": 31276
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.496079122798181e-06,
+      "loss": 0.2771,
+      "step": 31277
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.495743287491814e-06,
+      "loss": 0.1824,
+      "step": 31278
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.495407452185449e-06,
+      "loss": 0.2306,
+      "step": 31279
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.495071616879083e-06,
+      "loss": 0.2202,
+      "step": 31280
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.494735781572716e-06,
+      "loss": 0.2982,
+      "step": 31281
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.494399946266351e-06,
+      "loss": 0.3847,
+      "step": 31282
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.494064110959986e-06,
+      "loss": 0.5011,
+      "step": 31283
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.49372827565362e-06,
+      "loss": 0.2933,
+      "step": 31284
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.493392440347255e-06,
+      "loss": 0.3023,
+      "step": 31285
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.49305660504089e-06,
+      "loss": 0.3132,
+      "step": 31286
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.492720769734522e-06,
+      "loss": 0.2638,
+      "step": 31287
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.492384934428157e-06,
+      "loss": 0.2446,
+      "step": 31288
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.492049099121792e-06,
+      "loss": 0.1987,
+      "step": 31289
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.491713263815425e-06,
+      "loss": 0.2066,
+      "step": 31290
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.49137742850906e-06,
+      "loss": 0.2608,
+      "step": 31291
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.491041593202694e-06,
+      "loss": 0.1559,
+      "step": 31292
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.490705757896329e-06,
+      "loss": 0.3173,
+      "step": 31293
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.490369922589963e-06,
+      "loss": 0.2831,
+      "step": 31294
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.490034087283598e-06,
+      "loss": 0.4487,
+      "step": 31295
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.489698251977231e-06,
+      "loss": 0.1468,
+      "step": 31296
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.489362416670866e-06,
+      "loss": 0.2165,
+      "step": 31297
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.4890265813645e-06,
+      "loss": 0.3115,
+      "step": 31298
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.488690746058133e-06,
+      "loss": 0.2213,
+      "step": 31299
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.488354910751768e-06,
+      "loss": 0.3399,
+      "step": 31300
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.488019075445402e-06,
+      "loss": 0.2236,
+      "step": 31301
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.487683240139035e-06,
+      "loss": 0.3912,
+      "step": 31302
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.48734740483267e-06,
+      "loss": 0.2268,
+      "step": 31303
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.487011569526305e-06,
+      "loss": 0.2054,
+      "step": 31304
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.48667573421994e-06,
+      "loss": 0.4951,
+      "step": 31305
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.486339898913574e-06,
+      "loss": 0.2522,
+      "step": 31306
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.486004063607209e-06,
+      "loss": 0.3532,
+      "step": 31307
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.485668228300842e-06,
+      "loss": 0.2503,
+      "step": 31308
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.485332392994476e-06,
+      "loss": 0.2953,
+      "step": 31309
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.484996557688111e-06,
+      "loss": 0.1161,
+      "step": 31310
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.484660722381744e-06,
+      "loss": 0.4298,
+      "step": 31311
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.484324887075378e-06,
+      "loss": 0.3662,
+      "step": 31312
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.483989051769013e-06,
+      "loss": 0.2281,
+      "step": 31313
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.483653216462648e-06,
+      "loss": 0.2101,
+      "step": 31314
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.483317381156282e-06,
+      "loss": 0.2526,
+      "step": 31315
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.482981545849917e-06,
+      "loss": 0.3594,
+      "step": 31316
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.48264571054355e-06,
+      "loss": 0.2415,
+      "step": 31317
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.482309875237185e-06,
+      "loss": 0.3832,
+      "step": 31318
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.48197403993082e-06,
+      "loss": 0.4206,
+      "step": 31319
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.481638204624452e-06,
+      "loss": 0.3395,
+      "step": 31320
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.481302369318087e-06,
+      "loss": 0.2354,
+      "step": 31321
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.480966534011722e-06,
+      "loss": 0.2153,
+      "step": 31322
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.480630698705356e-06,
+      "loss": 0.2502,
+      "step": 31323
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.480294863398989e-06,
+      "loss": 0.2368,
+      "step": 31324
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.479959028092624e-06,
+      "loss": 0.304,
+      "step": 31325
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.479623192786258e-06,
+      "loss": 0.3553,
+      "step": 31326
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.479287357479893e-06,
+      "loss": 0.3739,
+      "step": 31327
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.478951522173528e-06,
+      "loss": 0.2163,
+      "step": 31328
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.47861568686716e-06,
+      "loss": 0.3663,
+      "step": 31329
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.478279851560795e-06,
+      "loss": 0.2719,
+      "step": 31330
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.47794401625443e-06,
+      "loss": 0.2502,
+      "step": 31331
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.477608180948063e-06,
+      "loss": 0.4862,
+      "step": 31332
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.477272345641698e-06,
+      "loss": 0.4529,
+      "step": 31333
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.476936510335332e-06,
+      "loss": 0.3248,
+      "step": 31334
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.476600675028967e-06,
+      "loss": 0.2768,
+      "step": 31335
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.476264839722602e-06,
+      "loss": 0.2098,
+      "step": 31336
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.475929004416236e-06,
+      "loss": 0.2925,
+      "step": 31337
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.475593169109869e-06,
+      "loss": 0.2692,
+      "step": 31338
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.475257333803504e-06,
+      "loss": 0.2843,
+      "step": 31339
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.474921498497138e-06,
+      "loss": 0.2662,
+      "step": 31340
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.474585663190771e-06,
+      "loss": 0.1883,
+      "step": 31341
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.474249827884406e-06,
+      "loss": 0.2873,
+      "step": 31342
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.47391399257804e-06,
+      "loss": 0.2014,
+      "step": 31343
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.473578157271675e-06,
+      "loss": 0.177,
+      "step": 31344
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.47324232196531e-06,
+      "loss": 0.4247,
+      "step": 31345
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.472906486658943e-06,
+      "loss": 0.2715,
+      "step": 31346
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.472570651352578e-06,
+      "loss": 0.3479,
+      "step": 31347
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.472234816046212e-06,
+      "loss": 0.3684,
+      "step": 31348
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.471898980739847e-06,
+      "loss": 0.3142,
+      "step": 31349
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.47156314543348e-06,
+      "loss": 0.3284,
+      "step": 31350
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.471227310127114e-06,
+      "loss": 0.1621,
+      "step": 31351
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.470891474820749e-06,
+      "loss": 0.4083,
+      "step": 31352
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.470555639514382e-06,
+      "loss": 0.2684,
+      "step": 31353
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.470219804208017e-06,
+      "loss": 0.3611,
+      "step": 31354
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.469883968901651e-06,
+      "loss": 0.2079,
+      "step": 31355
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.469548133595286e-06,
+      "loss": 0.3909,
+      "step": 31356
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.46921229828892e-06,
+      "loss": 0.3492,
+      "step": 31357
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.468876462982554e-06,
+      "loss": 0.4475,
+      "step": 31358
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.468540627676188e-06,
+      "loss": 0.3926,
+      "step": 31359
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.468204792369823e-06,
+      "loss": 0.4861,
+      "step": 31360
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.467868957063456e-06,
+      "loss": 0.3062,
+      "step": 31361
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.46753312175709e-06,
+      "loss": 0.2051,
+      "step": 31362
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.467197286450725e-06,
+      "loss": 0.4094,
+      "step": 31363
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.46686145114436e-06,
+      "loss": 0.2912,
+      "step": 31364
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.466525615837994e-06,
+      "loss": 0.2226,
+      "step": 31365
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.466189780531629e-06,
+      "loss": 0.3149,
+      "step": 31366
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.465853945225262e-06,
+      "loss": 0.3195,
+      "step": 31367
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.465518109918897e-06,
+      "loss": 0.2383,
+      "step": 31368
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.465182274612531e-06,
+      "loss": 0.2251,
+      "step": 31369
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.464846439306164e-06,
+      "loss": 0.3043,
+      "step": 31370
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.464510603999799e-06,
+      "loss": 0.2179,
+      "step": 31371
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.464174768693434e-06,
+      "loss": 0.3328,
+      "step": 31372
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.463838933387067e-06,
+      "loss": 0.1965,
+      "step": 31373
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.463503098080701e-06,
+      "loss": 0.3356,
+      "step": 31374
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.463167262774336e-06,
+      "loss": 0.4397,
+      "step": 31375
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.46283142746797e-06,
+      "loss": 0.3535,
+      "step": 31376
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.462495592161605e-06,
+      "loss": 0.2684,
+      "step": 31377
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.46215975685524e-06,
+      "loss": 0.162,
+      "step": 31378
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.461823921548873e-06,
+      "loss": 0.1574,
+      "step": 31379
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.461488086242507e-06,
+      "loss": 0.3107,
+      "step": 31380
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.461152250936142e-06,
+      "loss": 0.2371,
+      "step": 31381
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.460816415629775e-06,
+      "loss": 0.2954,
+      "step": 31382
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.46048058032341e-06,
+      "loss": 0.2266,
+      "step": 31383
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.460144745017044e-06,
+      "loss": 0.3451,
+      "step": 31384
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.459808909710679e-06,
+      "loss": 0.313,
+      "step": 31385
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.459473074404314e-06,
+      "loss": 0.3466,
+      "step": 31386
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.459137239097948e-06,
+      "loss": 0.3024,
+      "step": 31387
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.458801403791581e-06,
+      "loss": 0.3465,
+      "step": 31388
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.458465568485216e-06,
+      "loss": 0.281,
+      "step": 31389
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.45812973317885e-06,
+      "loss": 0.3531,
+      "step": 31390
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.457793897872483e-06,
+      "loss": 0.4215,
+      "step": 31391
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.457458062566118e-06,
+      "loss": 0.2726,
+      "step": 31392
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.457122227259753e-06,
+      "loss": 0.1923,
+      "step": 31393
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.456786391953386e-06,
+      "loss": 0.3277,
+      "step": 31394
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.45645055664702e-06,
+      "loss": 0.2336,
+      "step": 31395
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.456114721340655e-06,
+      "loss": 0.3065,
+      "step": 31396
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.45577888603429e-06,
+      "loss": 0.2258,
+      "step": 31397
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.455443050727924e-06,
+      "loss": 0.2109,
+      "step": 31398
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.455107215421559e-06,
+      "loss": 0.3296,
+      "step": 31399
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.454771380115192e-06,
+      "loss": 0.2594,
+      "step": 31400
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.454435544808827e-06,
+      "loss": 0.2286,
+      "step": 31401
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.454099709502461e-06,
+      "loss": 0.2587,
+      "step": 31402
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.453763874196094e-06,
+      "loss": 0.3595,
+      "step": 31403
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.453428038889729e-06,
+      "loss": 0.1979,
+      "step": 31404
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.453092203583363e-06,
+      "loss": 0.3715,
+      "step": 31405
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.452756368276998e-06,
+      "loss": 0.3053,
+      "step": 31406
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.452420532970633e-06,
+      "loss": 0.4929,
+      "step": 31407
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.452084697664267e-06,
+      "loss": 0.2597,
+      "step": 31408
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.4517488623579e-06,
+      "loss": 0.3026,
+      "step": 31409
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.451413027051535e-06,
+      "loss": 0.4802,
+      "step": 31410
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.45107719174517e-06,
+      "loss": 0.2743,
+      "step": 31411
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.450741356438803e-06,
+      "loss": 0.3681,
+      "step": 31412
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.450405521132437e-06,
+      "loss": 0.3002,
+      "step": 31413
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.450069685826072e-06,
+      "loss": 0.315,
+      "step": 31414
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.449733850519705e-06,
+      "loss": 0.1524,
+      "step": 31415
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.44939801521334e-06,
+      "loss": 0.3372,
+      "step": 31416
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.449062179906974e-06,
+      "loss": 0.3027,
+      "step": 31417
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.448726344600609e-06,
+      "loss": 0.2517,
+      "step": 31418
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.448390509294243e-06,
+      "loss": 0.1567,
+      "step": 31419
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.448054673987878e-06,
+      "loss": 0.3747,
+      "step": 31420
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.447718838681511e-06,
+      "loss": 0.2969,
+      "step": 31421
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.447383003375146e-06,
+      "loss": 0.2824,
+      "step": 31422
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.44704716806878e-06,
+      "loss": 0.3003,
+      "step": 31423
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.446711332762413e-06,
+      "loss": 0.3544,
+      "step": 31424
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.446375497456048e-06,
+      "loss": 0.2757,
+      "step": 31425
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.446039662149683e-06,
+      "loss": 0.5518,
+      "step": 31426
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.445703826843317e-06,
+      "loss": 0.4366,
+      "step": 31427
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.445367991536952e-06,
+      "loss": 0.2982,
+      "step": 31428
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.445032156230586e-06,
+      "loss": 0.2716,
+      "step": 31429
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.44469632092422e-06,
+      "loss": 0.3399,
+      "step": 31430
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.444360485617854e-06,
+      "loss": 0.2751,
+      "step": 31431
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.444024650311489e-06,
+      "loss": 0.4127,
+      "step": 31432
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.443688815005122e-06,
+      "loss": 0.2339,
+      "step": 31433
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.443352979698756e-06,
+      "loss": 0.3947,
+      "step": 31434
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.443017144392391e-06,
+      "loss": 0.2028,
+      "step": 31435
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.442681309086024e-06,
+      "loss": 0.1622,
+      "step": 31436
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.442345473779659e-06,
+      "loss": 0.2682,
+      "step": 31437
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.442009638473293e-06,
+      "loss": 0.3483,
+      "step": 31438
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.441673803166928e-06,
+      "loss": 0.3795,
+      "step": 31439
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.441337967860563e-06,
+      "loss": 0.2033,
+      "step": 31440
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.441002132554197e-06,
+      "loss": 0.3015,
+      "step": 31441
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.44066629724783e-06,
+      "loss": 0.1841,
+      "step": 31442
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.440330461941465e-06,
+      "loss": 0.2832,
+      "step": 31443
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.4399946266351e-06,
+      "loss": 0.3189,
+      "step": 31444
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.439658791328732e-06,
+      "loss": 0.1736,
+      "step": 31445
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.439322956022367e-06,
+      "loss": 0.4507,
+      "step": 31446
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.438987120716002e-06,
+      "loss": 0.2606,
+      "step": 31447
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.438651285409636e-06,
+      "loss": 0.3252,
+      "step": 31448
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.438315450103271e-06,
+      "loss": 0.349,
+      "step": 31449
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.437979614796904e-06,
+      "loss": 0.3708,
+      "step": 31450
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.437643779490539e-06,
+      "loss": 0.2828,
+      "step": 31451
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.437307944184173e-06,
+      "loss": 0.2162,
+      "step": 31452
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.436972108877806e-06,
+      "loss": 0.2131,
+      "step": 31453
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.43663627357144e-06,
+      "loss": 0.3676,
+      "step": 31454
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.436300438265075e-06,
+      "loss": 0.347,
+      "step": 31455
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.43596460295871e-06,
+      "loss": 0.2404,
+      "step": 31456
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.435628767652345e-06,
+      "loss": 0.1749,
+      "step": 31457
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.43529293234598e-06,
+      "loss": 0.3661,
+      "step": 31458
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.434957097039612e-06,
+      "loss": 0.3083,
+      "step": 31459
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.434621261733247e-06,
+      "loss": 0.3673,
+      "step": 31460
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.434285426426882e-06,
+      "loss": 0.1641,
+      "step": 31461
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.433949591120515e-06,
+      "loss": 0.374,
+      "step": 31462
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 9.43361375581415e-06,
+      "loss": 0.2904,
+      "step": 31463
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.433277920507784e-06,
+      "loss": 0.1434,
+      "step": 31464
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.432942085201417e-06,
+      "loss": 0.5367,
+      "step": 31465
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.432606249895051e-06,
+      "loss": 0.1697,
+      "step": 31466
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.432270414588686e-06,
+      "loss": 0.2145,
+      "step": 31467
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.43193457928232e-06,
+      "loss": 0.3676,
+      "step": 31468
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.431598743975955e-06,
+      "loss": 0.2948,
+      "step": 31469
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.43126290866959e-06,
+      "loss": 0.2146,
+      "step": 31470
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.430927073363223e-06,
+      "loss": 0.2738,
+      "step": 31471
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.430591238056858e-06,
+      "loss": 0.2555,
+      "step": 31472
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.430255402750492e-06,
+      "loss": 0.221,
+      "step": 31473
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.429919567444125e-06,
+      "loss": 0.2481,
+      "step": 31474
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.42958373213776e-06,
+      "loss": 0.2201,
+      "step": 31475
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.429247896831395e-06,
+      "loss": 0.2184,
+      "step": 31476
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.42891206152503e-06,
+      "loss": 0.3052,
+      "step": 31477
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.428576226218664e-06,
+      "loss": 0.1832,
+      "step": 31478
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.428240390912299e-06,
+      "loss": 0.2958,
+      "step": 31479
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.427904555605931e-06,
+      "loss": 0.4368,
+      "step": 31480
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.427568720299566e-06,
+      "loss": 0.2229,
+      "step": 31481
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.4272328849932e-06,
+      "loss": 0.3302,
+      "step": 31482
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.426897049686834e-06,
+      "loss": 0.2565,
+      "step": 31483
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.426561214380468e-06,
+      "loss": 0.2334,
+      "step": 31484
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.426225379074103e-06,
+      "loss": 0.24,
+      "step": 31485
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.425889543767736e-06,
+      "loss": 0.3272,
+      "step": 31486
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.42555370846137e-06,
+      "loss": 0.3091,
+      "step": 31487
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.425217873155005e-06,
+      "loss": 0.2542,
+      "step": 31488
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.42488203784864e-06,
+      "loss": 0.3814,
+      "step": 31489
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.424546202542275e-06,
+      "loss": 0.2349,
+      "step": 31490
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.42421036723591e-06,
+      "loss": 0.1388,
+      "step": 31491
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.423874531929542e-06,
+      "loss": 0.2489,
+      "step": 31492
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.423538696623177e-06,
+      "loss": 0.2584,
+      "step": 31493
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.423202861316811e-06,
+      "loss": 0.2183,
+      "step": 31494
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.422867026010444e-06,
+      "loss": 0.2679,
+      "step": 31495
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.422531190704079e-06,
+      "loss": 0.4893,
+      "step": 31496
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.422195355397714e-06,
+      "loss": 0.4632,
+      "step": 31497
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.421859520091348e-06,
+      "loss": 0.386,
+      "step": 31498
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.421523684784983e-06,
+      "loss": 0.3386,
+      "step": 31499
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.421187849478618e-06,
+      "loss": 0.1988,
+      "step": 31500
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.42085201417225e-06,
+      "loss": 0.4376,
+      "step": 31501
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.420516178865885e-06,
+      "loss": 0.3581,
+      "step": 31502
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.42018034355952e-06,
+      "loss": 0.4168,
+      "step": 31503
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.419844508253153e-06,
+      "loss": 0.2858,
+      "step": 31504
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.419508672946787e-06,
+      "loss": 0.2843,
+      "step": 31505
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.419172837640422e-06,
+      "loss": 0.1888,
+      "step": 31506
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.418837002334055e-06,
+      "loss": 0.2359,
+      "step": 31507
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.41850116702769e-06,
+      "loss": 0.2592,
+      "step": 31508
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.418165331721324e-06,
+      "loss": 0.2072,
+      "step": 31509
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.417829496414959e-06,
+      "loss": 0.2529,
+      "step": 31510
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.417493661108594e-06,
+      "loss": 0.2763,
+      "step": 31511
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.417157825802228e-06,
+      "loss": 0.2168,
+      "step": 31512
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.416821990495861e-06,
+      "loss": 0.1625,
+      "step": 31513
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.416486155189496e-06,
+      "loss": 0.2453,
+      "step": 31514
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.41615031988313e-06,
+      "loss": 0.3256,
+      "step": 31515
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.415814484576764e-06,
+      "loss": 0.3292,
+      "step": 31516
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.415478649270398e-06,
+      "loss": 0.3161,
+      "step": 31517
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.415142813964033e-06,
+      "loss": 0.2943,
+      "step": 31518
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.414806978657667e-06,
+      "loss": 0.3673,
+      "step": 31519
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.414471143351302e-06,
+      "loss": 0.2493,
+      "step": 31520
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.414135308044937e-06,
+      "loss": 0.1881,
+      "step": 31521
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.41379947273857e-06,
+      "loss": 0.215,
+      "step": 31522
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.413463637432204e-06,
+      "loss": 0.1506,
+      "step": 31523
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.413127802125839e-06,
+      "loss": 0.3751,
+      "step": 31524
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.412791966819472e-06,
+      "loss": 0.3383,
+      "step": 31525
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.412456131513107e-06,
+      "loss": 0.2613,
+      "step": 31526
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.412120296206741e-06,
+      "loss": 0.315,
+      "step": 31527
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.411784460900374e-06,
+      "loss": 0.1649,
+      "step": 31528
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.411448625594009e-06,
+      "loss": 0.2967,
+      "step": 31529
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.411112790287644e-06,
+      "loss": 0.1762,
+      "step": 31530
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.410776954981278e-06,
+      "loss": 0.2961,
+      "step": 31531
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.410441119674913e-06,
+      "loss": 0.1462,
+      "step": 31532
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.410105284368547e-06,
+      "loss": 0.1616,
+      "step": 31533
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.40976944906218e-06,
+      "loss": 0.2831,
+      "step": 31534
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.409433613755815e-06,
+      "loss": 0.3771,
+      "step": 31535
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.40909777844945e-06,
+      "loss": 0.3098,
+      "step": 31536
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.408761943143083e-06,
+      "loss": 0.4124,
+      "step": 31537
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.408426107836717e-06,
+      "loss": 0.1914,
+      "step": 31538
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.408090272530352e-06,
+      "loss": 0.307,
+      "step": 31539
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.407754437223987e-06,
+      "loss": 0.3106,
+      "step": 31540
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.407418601917621e-06,
+      "loss": 0.2283,
+      "step": 31541
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.407082766611254e-06,
+      "loss": 0.2791,
+      "step": 31542
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.406746931304889e-06,
+      "loss": 0.3432,
+      "step": 31543
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.406411095998523e-06,
+      "loss": 0.4485,
+      "step": 31544
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.406075260692156e-06,
+      "loss": 0.3863,
+      "step": 31545
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.405739425385791e-06,
+      "loss": 0.3281,
+      "step": 31546
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.405403590079426e-06,
+      "loss": 0.3057,
+      "step": 31547
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.40506775477306e-06,
+      "loss": 0.4763,
+      "step": 31548
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.404731919466693e-06,
+      "loss": 0.3246,
+      "step": 31549
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.404396084160328e-06,
+      "loss": 0.2592,
+      "step": 31550
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.404060248853963e-06,
+      "loss": 0.2007,
+      "step": 31551
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.403724413547597e-06,
+      "loss": 0.3865,
+      "step": 31552
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.403388578241232e-06,
+      "loss": 0.2438,
+      "step": 31553
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.403052742934865e-06,
+      "loss": 0.3173,
+      "step": 31554
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.4027169076285e-06,
+      "loss": 0.2493,
+      "step": 31555
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.402381072322134e-06,
+      "loss": 0.2593,
+      "step": 31556
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.402045237015767e-06,
+      "loss": 0.3576,
+      "step": 31557
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.401709401709402e-06,
+      "loss": 0.3633,
+      "step": 31558
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.401373566403036e-06,
+      "loss": 0.2205,
+      "step": 31559
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.401037731096671e-06,
+      "loss": 0.2797,
+      "step": 31560
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.400701895790306e-06,
+      "loss": 0.2135,
+      "step": 31561
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.40036606048394e-06,
+      "loss": 0.2564,
+      "step": 31562
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.400030225177573e-06,
+      "loss": 0.2559,
+      "step": 31563
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.399694389871208e-06,
+      "loss": 0.3154,
+      "step": 31564
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.399358554564843e-06,
+      "loss": 0.2947,
+      "step": 31565
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.399022719258476e-06,
+      "loss": 0.2098,
+      "step": 31566
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.39868688395211e-06,
+      "loss": 0.2507,
+      "step": 31567
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.398351048645745e-06,
+      "loss": 0.2621,
+      "step": 31568
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.39801521333938e-06,
+      "loss": 0.114,
+      "step": 31569
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.397679378033014e-06,
+      "loss": 0.3644,
+      "step": 31570
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.397343542726647e-06,
+      "loss": 0.241,
+      "step": 31571
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.397007707420282e-06,
+      "loss": 0.2135,
+      "step": 31572
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.396671872113916e-06,
+      "loss": 0.3058,
+      "step": 31573
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.396336036807551e-06,
+      "loss": 0.2061,
+      "step": 31574
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.396000201501184e-06,
+      "loss": 0.4087,
+      "step": 31575
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.395664366194819e-06,
+      "loss": 0.3309,
+      "step": 31576
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.395328530888453e-06,
+      "loss": 0.2199,
+      "step": 31577
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.394992695582086e-06,
+      "loss": 0.2667,
+      "step": 31578
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.394656860275721e-06,
+      "loss": 0.3159,
+      "step": 31579
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.394321024969356e-06,
+      "loss": 0.2742,
+      "step": 31580
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.39398518966299e-06,
+      "loss": 0.1575,
+      "step": 31581
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.393649354356625e-06,
+      "loss": 0.4455,
+      "step": 31582
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.39331351905026e-06,
+      "loss": 0.1795,
+      "step": 31583
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.392977683743892e-06,
+      "loss": 0.2748,
+      "step": 31584
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.392641848437527e-06,
+      "loss": 0.2595,
+      "step": 31585
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.392306013131162e-06,
+      "loss": 0.5301,
+      "step": 31586
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.391970177824795e-06,
+      "loss": 0.3713,
+      "step": 31587
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.39163434251843e-06,
+      "loss": 0.3028,
+      "step": 31588
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.391298507212064e-06,
+      "loss": 0.1814,
+      "step": 31589
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.390962671905699e-06,
+      "loss": 0.1423,
+      "step": 31590
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.390626836599333e-06,
+      "loss": 0.452,
+      "step": 31591
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.390291001292968e-06,
+      "loss": 0.0945,
+      "step": 31592
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.389955165986601e-06,
+      "loss": 0.226,
+      "step": 31593
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.389619330680236e-06,
+      "loss": 0.2828,
+      "step": 31594
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.38928349537387e-06,
+      "loss": 0.2458,
+      "step": 31595
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.388947660067503e-06,
+      "loss": 0.3349,
+      "step": 31596
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.388611824761138e-06,
+      "loss": 0.2309,
+      "step": 31597
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.388275989454772e-06,
+      "loss": 0.2734,
+      "step": 31598
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.387940154148405e-06,
+      "loss": 0.3382,
+      "step": 31599
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.38760431884204e-06,
+      "loss": 0.423,
+      "step": 31600
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.387268483535675e-06,
+      "loss": 0.3215,
+      "step": 31601
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.38693264822931e-06,
+      "loss": 0.3432,
+      "step": 31602
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.386596812922944e-06,
+      "loss": 0.2382,
+      "step": 31603
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.386260977616579e-06,
+      "loss": 0.3478,
+      "step": 31604
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.385925142310212e-06,
+      "loss": 0.4551,
+      "step": 31605
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.385589307003846e-06,
+      "loss": 0.3663,
+      "step": 31606
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.385253471697481e-06,
+      "loss": 0.18,
+      "step": 31607
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.384917636391114e-06,
+      "loss": 0.3536,
+      "step": 31608
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.384581801084748e-06,
+      "loss": 0.2628,
+      "step": 31609
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.384245965778383e-06,
+      "loss": 0.2549,
+      "step": 31610
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.383910130472018e-06,
+      "loss": 0.285,
+      "step": 31611
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.383574295165652e-06,
+      "loss": 0.1264,
+      "step": 31612
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.383238459859287e-06,
+      "loss": 0.2816,
+      "step": 31613
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.38290262455292e-06,
+      "loss": 0.2015,
+      "step": 31614
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.382566789246555e-06,
+      "loss": 0.3721,
+      "step": 31615
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.38223095394019e-06,
+      "loss": 0.3956,
+      "step": 31616
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.381895118633822e-06,
+      "loss": 0.1994,
+      "step": 31617
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.381559283327457e-06,
+      "loss": 0.1395,
+      "step": 31618
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.381223448021092e-06,
+      "loss": 0.2642,
+      "step": 31619
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.380887612714724e-06,
+      "loss": 0.3211,
+      "step": 31620
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.380551777408359e-06,
+      "loss": 0.3752,
+      "step": 31621
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.380215942101994e-06,
+      "loss": 0.4168,
+      "step": 31622
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.379880106795628e-06,
+      "loss": 0.2454,
+      "step": 31623
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.379544271489263e-06,
+      "loss": 0.4447,
+      "step": 31624
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.379208436182898e-06,
+      "loss": 0.2545,
+      "step": 31625
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.37887260087653e-06,
+      "loss": 0.4129,
+      "step": 31626
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.378536765570165e-06,
+      "loss": 0.2748,
+      "step": 31627
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.378200930263798e-06,
+      "loss": 0.2323,
+      "step": 31628
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.377865094957433e-06,
+      "loss": 0.293,
+      "step": 31629
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.377529259651068e-06,
+      "loss": 0.3142,
+      "step": 31630
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.377193424344702e-06,
+      "loss": 0.1481,
+      "step": 31631
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.376857589038337e-06,
+      "loss": 0.3344,
+      "step": 31632
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.376521753731972e-06,
+      "loss": 0.1751,
+      "step": 31633
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.376185918425604e-06,
+      "loss": 0.202,
+      "step": 31634
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.375850083119239e-06,
+      "loss": 0.2494,
+      "step": 31635
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.375514247812874e-06,
+      "loss": 0.5629,
+      "step": 31636
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.375178412506507e-06,
+      "loss": 0.2921,
+      "step": 31637
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.374842577200141e-06,
+      "loss": 0.2287,
+      "step": 31638
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.374506741893776e-06,
+      "loss": 0.3908,
+      "step": 31639
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.374170906587409e-06,
+      "loss": 0.2066,
+      "step": 31640
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.373835071281044e-06,
+      "loss": 0.3931,
+      "step": 31641
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.373499235974678e-06,
+      "loss": 0.2093,
+      "step": 31642
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.373163400668313e-06,
+      "loss": 0.2286,
+      "step": 31643
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.372827565361948e-06,
+      "loss": 0.3804,
+      "step": 31644
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.372491730055582e-06,
+      "loss": 0.2821,
+      "step": 31645
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.372155894749215e-06,
+      "loss": 0.2173,
+      "step": 31646
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.37182005944285e-06,
+      "loss": 0.3536,
+      "step": 31647
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.371484224136484e-06,
+      "loss": 0.1395,
+      "step": 31648
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.371148388830117e-06,
+      "loss": 0.4733,
+      "step": 31649
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.370812553523752e-06,
+      "loss": 0.3276,
+      "step": 31650
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.370476718217387e-06,
+      "loss": 0.2616,
+      "step": 31651
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.370140882911021e-06,
+      "loss": 0.3774,
+      "step": 31652
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.369805047604656e-06,
+      "loss": 0.3126,
+      "step": 31653
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.36946921229829e-06,
+      "loss": 0.2066,
+      "step": 31654
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.369133376991924e-06,
+      "loss": 0.3959,
+      "step": 31655
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.368797541685558e-06,
+      "loss": 0.5211,
+      "step": 31656
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.368461706379193e-06,
+      "loss": 0.3267,
+      "step": 31657
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.368125871072826e-06,
+      "loss": 0.2011,
+      "step": 31658
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.36779003576646e-06,
+      "loss": 0.2108,
+      "step": 31659
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.367454200460095e-06,
+      "loss": 0.4104,
+      "step": 31660
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.36711836515373e-06,
+      "loss": 0.22,
+      "step": 31661
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 9.366782529847363e-06,
+      "loss": 0.3735,
+      "step": 31662
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.366446694540997e-06,
+      "loss": 0.257,
+      "step": 31663
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.366110859234632e-06,
+      "loss": 0.2468,
+      "step": 31664
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.365775023928267e-06,
+      "loss": 0.3085,
+      "step": 31665
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.365439188621901e-06,
+      "loss": 0.363,
+      "step": 31666
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.365103353315534e-06,
+      "loss": 0.4253,
+      "step": 31667
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.364767518009169e-06,
+      "loss": 0.237,
+      "step": 31668
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.364431682702804e-06,
+      "loss": 0.4888,
+      "step": 31669
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.364095847396437e-06,
+      "loss": 0.4422,
+      "step": 31670
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.363760012090071e-06,
+      "loss": 0.1905,
+      "step": 31671
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.363424176783706e-06,
+      "loss": 0.1989,
+      "step": 31672
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.36308834147734e-06,
+      "loss": 0.2697,
+      "step": 31673
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.362752506170975e-06,
+      "loss": 0.1491,
+      "step": 31674
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.36241667086461e-06,
+      "loss": 0.1614,
+      "step": 31675
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.362080835558243e-06,
+      "loss": 0.3486,
+      "step": 31676
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.361745000251877e-06,
+      "loss": 0.2644,
+      "step": 31677
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.361409164945512e-06,
+      "loss": 0.2073,
+      "step": 31678
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.361073329639145e-06,
+      "loss": 0.5328,
+      "step": 31679
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.36073749433278e-06,
+      "loss": 0.1715,
+      "step": 31680
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.360401659026414e-06,
+      "loss": 0.2569,
+      "step": 31681
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.360065823720049e-06,
+      "loss": 0.2968,
+      "step": 31682
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.359729988413684e-06,
+      "loss": 0.2512,
+      "step": 31683
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.359394153107317e-06,
+      "loss": 0.2405,
+      "step": 31684
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.359058317800951e-06,
+      "loss": 0.3068,
+      "step": 31685
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.358722482494586e-06,
+      "loss": 0.3036,
+      "step": 31686
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.35838664718822e-06,
+      "loss": 0.2123,
+      "step": 31687
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.358050811881853e-06,
+      "loss": 0.2197,
+      "step": 31688
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.357714976575488e-06,
+      "loss": 0.2032,
+      "step": 31689
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.357379141269123e-06,
+      "loss": 0.2107,
+      "step": 31690
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.357043305962756e-06,
+      "loss": 0.4039,
+      "step": 31691
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.35670747065639e-06,
+      "loss": 0.2524,
+      "step": 31692
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.356371635350025e-06,
+      "loss": 0.2402,
+      "step": 31693
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.35603580004366e-06,
+      "loss": 0.281,
+      "step": 31694
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.355699964737294e-06,
+      "loss": 0.1736,
+      "step": 31695
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.355364129430929e-06,
+      "loss": 0.2819,
+      "step": 31696
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.355028294124562e-06,
+      "loss": 0.2272,
+      "step": 31697
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.354692458818196e-06,
+      "loss": 0.3889,
+      "step": 31698
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.354356623511831e-06,
+      "loss": 0.4714,
+      "step": 31699
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.354020788205464e-06,
+      "loss": 0.2289,
+      "step": 31700
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.353684952899099e-06,
+      "loss": 0.4282,
+      "step": 31701
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.353349117592733e-06,
+      "loss": 0.2121,
+      "step": 31702
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.353013282286368e-06,
+      "loss": 0.3746,
+      "step": 31703
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.352677446980003e-06,
+      "loss": 0.3271,
+      "step": 31704
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.352341611673637e-06,
+      "loss": 0.3229,
+      "step": 31705
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.35200577636727e-06,
+      "loss": 0.372,
+      "step": 31706
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.351669941060905e-06,
+      "loss": 0.2017,
+      "step": 31707
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.35133410575454e-06,
+      "loss": 0.1951,
+      "step": 31708
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.350998270448173e-06,
+      "loss": 0.2099,
+      "step": 31709
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.350662435141807e-06,
+      "loss": 0.2968,
+      "step": 31710
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.350326599835442e-06,
+      "loss": 0.3886,
+      "step": 31711
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.349990764529075e-06,
+      "loss": 0.1755,
+      "step": 31712
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.34965492922271e-06,
+      "loss": 0.1946,
+      "step": 31713
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.349319093916344e-06,
+      "loss": 0.3647,
+      "step": 31714
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.348983258609979e-06,
+      "loss": 0.2638,
+      "step": 31715
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.348647423303613e-06,
+      "loss": 0.2411,
+      "step": 31716
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.348311587997246e-06,
+      "loss": 0.2925,
+      "step": 31717
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.347975752690881e-06,
+      "loss": 0.2898,
+      "step": 31718
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.347639917384516e-06,
+      "loss": 0.2939,
+      "step": 31719
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.347304082078149e-06,
+      "loss": 0.4255,
+      "step": 31720
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.346968246771783e-06,
+      "loss": 0.3525,
+      "step": 31721
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.346632411465418e-06,
+      "loss": 0.2603,
+      "step": 31722
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.346296576159053e-06,
+      "loss": 0.3411,
+      "step": 31723
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.345960740852687e-06,
+      "loss": 0.2005,
+      "step": 31724
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.345624905546322e-06,
+      "loss": 0.3813,
+      "step": 31725
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.345289070239955e-06,
+      "loss": 0.2337,
+      "step": 31726
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.34495323493359e-06,
+      "loss": 0.3313,
+      "step": 31727
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.344617399627224e-06,
+      "loss": 0.303,
+      "step": 31728
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.344281564320857e-06,
+      "loss": 0.366,
+      "step": 31729
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.343945729014492e-06,
+      "loss": 0.3107,
+      "step": 31730
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.343609893708126e-06,
+      "loss": 0.2542,
+      "step": 31731
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.34327405840176e-06,
+      "loss": 0.2675,
+      "step": 31732
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.342938223095394e-06,
+      "loss": 0.2166,
+      "step": 31733
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.342602387789029e-06,
+      "loss": 0.281,
+      "step": 31734
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.342266552482663e-06,
+      "loss": 0.2976,
+      "step": 31735
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.341930717176298e-06,
+      "loss": 0.1929,
+      "step": 31736
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.341594881869933e-06,
+      "loss": 0.2381,
+      "step": 31737
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.341259046563565e-06,
+      "loss": 0.1814,
+      "step": 31738
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.3409232112572e-06,
+      "loss": 0.3474,
+      "step": 31739
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.340587375950835e-06,
+      "loss": 0.2209,
+      "step": 31740
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.340251540644468e-06,
+      "loss": 0.3038,
+      "step": 31741
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.339915705338102e-06,
+      "loss": 0.3175,
+      "step": 31742
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.339579870031737e-06,
+      "loss": 0.3955,
+      "step": 31743
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.339244034725372e-06,
+      "loss": 0.2982,
+      "step": 31744
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.338908199419006e-06,
+      "loss": 0.3471,
+      "step": 31745
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.338572364112641e-06,
+      "loss": 0.2599,
+      "step": 31746
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.338236528806274e-06,
+      "loss": 0.2165,
+      "step": 31747
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.337900693499909e-06,
+      "loss": 0.3846,
+      "step": 31748
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.337564858193543e-06,
+      "loss": 0.2369,
+      "step": 31749
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.337229022887176e-06,
+      "loss": 0.502,
+      "step": 31750
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.33689318758081e-06,
+      "loss": 0.2307,
+      "step": 31751
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.336557352274445e-06,
+      "loss": 0.2376,
+      "step": 31752
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.336221516968078e-06,
+      "loss": 0.2471,
+      "step": 31753
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.335885681661713e-06,
+      "loss": 0.2776,
+      "step": 31754
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.335549846355348e-06,
+      "loss": 0.319,
+      "step": 31755
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.335214011048982e-06,
+      "loss": 0.3266,
+      "step": 31756
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.334878175742617e-06,
+      "loss": 0.3317,
+      "step": 31757
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.334542340436252e-06,
+      "loss": 0.2075,
+      "step": 31758
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.334206505129885e-06,
+      "loss": 0.3206,
+      "step": 31759
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.33387066982352e-06,
+      "loss": 0.2504,
+      "step": 31760
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.333534834517154e-06,
+      "loss": 0.361,
+      "step": 31761
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.333198999210787e-06,
+      "loss": 0.2902,
+      "step": 31762
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.332863163904421e-06,
+      "loss": 0.2405,
+      "step": 31763
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.332527328598056e-06,
+      "loss": 0.3128,
+      "step": 31764
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.33219149329169e-06,
+      "loss": 0.3763,
+      "step": 31765
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.331855657985325e-06,
+      "loss": 0.1583,
+      "step": 31766
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.33151982267896e-06,
+      "loss": 0.2646,
+      "step": 31767
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.331183987372593e-06,
+      "loss": 0.2824,
+      "step": 31768
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.330848152066228e-06,
+      "loss": 0.2081,
+      "step": 31769
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.330512316759862e-06,
+      "loss": 0.3036,
+      "step": 31770
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.330176481453495e-06,
+      "loss": 0.3845,
+      "step": 31771
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.32984064614713e-06,
+      "loss": 0.252,
+      "step": 31772
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.329504810840765e-06,
+      "loss": 0.295,
+      "step": 31773
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.3291689755344e-06,
+      "loss": 0.3982,
+      "step": 31774
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.328833140228032e-06,
+      "loss": 0.3605,
+      "step": 31775
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.328497304921667e-06,
+      "loss": 0.2332,
+      "step": 31776
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.328161469615301e-06,
+      "loss": 0.2818,
+      "step": 31777
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.327825634308936e-06,
+      "loss": 0.3327,
+      "step": 31778
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.32748979900257e-06,
+      "loss": 0.3047,
+      "step": 31779
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.327153963696204e-06,
+      "loss": 0.3172,
+      "step": 31780
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.326818128389838e-06,
+      "loss": 0.3015,
+      "step": 31781
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.326482293083473e-06,
+      "loss": 0.1987,
+      "step": 31782
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.326146457777106e-06,
+      "loss": 0.3951,
+      "step": 31783
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.32581062247074e-06,
+      "loss": 0.2111,
+      "step": 31784
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.325474787164375e-06,
+      "loss": 0.2858,
+      "step": 31785
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.32513895185801e-06,
+      "loss": 0.2559,
+      "step": 31786
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.324803116551645e-06,
+      "loss": 0.2659,
+      "step": 31787
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.32446728124528e-06,
+      "loss": 0.4097,
+      "step": 31788
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.324131445938912e-06,
+      "loss": 0.2002,
+      "step": 31789
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.323795610632547e-06,
+      "loss": 0.2879,
+      "step": 31790
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.323459775326181e-06,
+      "loss": 0.2404,
+      "step": 31791
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.323123940019814e-06,
+      "loss": 0.2244,
+      "step": 31792
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.322788104713449e-06,
+      "loss": 0.289,
+      "step": 31793
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.322452269407084e-06,
+      "loss": 0.4129,
+      "step": 31794
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.322116434100718e-06,
+      "loss": 0.3498,
+      "step": 31795
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.321780598794353e-06,
+      "loss": 0.3702,
+      "step": 31796
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.321444763487986e-06,
+      "loss": 0.2922,
+      "step": 31797
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.32110892818162e-06,
+      "loss": 0.2055,
+      "step": 31798
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.320773092875255e-06,
+      "loss": 0.2305,
+      "step": 31799
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.32043725756889e-06,
+      "loss": 0.3884,
+      "step": 31800
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.320101422262523e-06,
+      "loss": 0.3658,
+      "step": 31801
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.319765586956157e-06,
+      "loss": 0.2552,
+      "step": 31802
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.319429751649792e-06,
+      "loss": 0.3354,
+      "step": 31803
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.319093916343425e-06,
+      "loss": 0.3211,
+      "step": 31804
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.31875808103706e-06,
+      "loss": 0.1265,
+      "step": 31805
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.318422245730694e-06,
+      "loss": 0.4171,
+      "step": 31806
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.318086410424329e-06,
+      "loss": 0.2423,
+      "step": 31807
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.317750575117964e-06,
+      "loss": 0.3518,
+      "step": 31808
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.317414739811597e-06,
+      "loss": 0.4073,
+      "step": 31809
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.317078904505231e-06,
+      "loss": 0.1051,
+      "step": 31810
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.316743069198866e-06,
+      "loss": 0.2506,
+      "step": 31811
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.316407233892499e-06,
+      "loss": 0.2875,
+      "step": 31812
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.316071398586134e-06,
+      "loss": 0.2012,
+      "step": 31813
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.315735563279768e-06,
+      "loss": 0.3464,
+      "step": 31814
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.315399727973403e-06,
+      "loss": 0.3557,
+      "step": 31815
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.315063892667037e-06,
+      "loss": 0.3667,
+      "step": 31816
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.314728057360672e-06,
+      "loss": 0.2326,
+      "step": 31817
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.314392222054305e-06,
+      "loss": 0.4503,
+      "step": 31818
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.31405638674794e-06,
+      "loss": 0.1657,
+      "step": 31819
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.313720551441574e-06,
+      "loss": 0.4522,
+      "step": 31820
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.313384716135207e-06,
+      "loss": 0.3554,
+      "step": 31821
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.313048880828842e-06,
+      "loss": 0.3632,
+      "step": 31822
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.312713045522477e-06,
+      "loss": 0.4377,
+      "step": 31823
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.31237721021611e-06,
+      "loss": 0.2118,
+      "step": 31824
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.312041374909744e-06,
+      "loss": 0.3181,
+      "step": 31825
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.311705539603379e-06,
+      "loss": 0.1672,
+      "step": 31826
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.311369704297013e-06,
+      "loss": 0.3155,
+      "step": 31827
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.311033868990648e-06,
+      "loss": 0.3398,
+      "step": 31828
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.310698033684283e-06,
+      "loss": 0.3292,
+      "step": 31829
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.310362198377916e-06,
+      "loss": 0.1973,
+      "step": 31830
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.31002636307155e-06,
+      "loss": 0.2238,
+      "step": 31831
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.309690527765185e-06,
+      "loss": 0.2073,
+      "step": 31832
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.309354692458818e-06,
+      "loss": 0.3707,
+      "step": 31833
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.309018857152453e-06,
+      "loss": 0.338,
+      "step": 31834
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.308683021846087e-06,
+      "loss": 0.2838,
+      "step": 31835
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.308347186539722e-06,
+      "loss": 0.3441,
+      "step": 31836
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.308011351233357e-06,
+      "loss": 0.1925,
+      "step": 31837
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.307675515926991e-06,
+      "loss": 0.3407,
+      "step": 31838
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.307339680620624e-06,
+      "loss": 0.2189,
+      "step": 31839
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.307003845314259e-06,
+      "loss": 0.2406,
+      "step": 31840
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.306668010007893e-06,
+      "loss": 0.3827,
+      "step": 31841
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.306332174701526e-06,
+      "loss": 0.477,
+      "step": 31842
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.305996339395161e-06,
+      "loss": 0.2776,
+      "step": 31843
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.305660504088796e-06,
+      "loss": 0.4866,
+      "step": 31844
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.305324668782429e-06,
+      "loss": 0.2148,
+      "step": 31845
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.304988833476063e-06,
+      "loss": 0.1891,
+      "step": 31846
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.304652998169698e-06,
+      "loss": 0.3112,
+      "step": 31847
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.304317162863333e-06,
+      "loss": 0.2922,
+      "step": 31848
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.303981327556967e-06,
+      "loss": 0.3848,
+      "step": 31849
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.303645492250602e-06,
+      "loss": 0.2764,
+      "step": 31850
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.303309656944235e-06,
+      "loss": 0.3853,
+      "step": 31851
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.30297382163787e-06,
+      "loss": 0.2279,
+      "step": 31852
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.302637986331504e-06,
+      "loss": 0.278,
+      "step": 31853
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.302302151025137e-06,
+      "loss": 0.2671,
+      "step": 31854
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.301966315718772e-06,
+      "loss": 0.398,
+      "step": 31855
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.301630480412406e-06,
+      "loss": 0.3651,
+      "step": 31856
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.301294645106041e-06,
+      "loss": 0.3139,
+      "step": 31857
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.300958809799676e-06,
+      "loss": 0.3347,
+      "step": 31858
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.30062297449331e-06,
+      "loss": 0.3619,
+      "step": 31859
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 9.300287139186943e-06,
+      "loss": 0.3798,
+      "step": 31860
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.299951303880578e-06,
+      "loss": 0.3108,
+      "step": 31861
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.299615468574213e-06,
+      "loss": 0.3582,
+      "step": 31862
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.299279633267846e-06,
+      "loss": 0.2747,
+      "step": 31863
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.29894379796148e-06,
+      "loss": 0.3533,
+      "step": 31864
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.298607962655115e-06,
+      "loss": 0.3833,
+      "step": 31865
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.298272127348748e-06,
+      "loss": 0.2521,
+      "step": 31866
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.297936292042382e-06,
+      "loss": 0.2852,
+      "step": 31867
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.297600456736017e-06,
+      "loss": 0.2516,
+      "step": 31868
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.297264621429652e-06,
+      "loss": 0.3161,
+      "step": 31869
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.296928786123286e-06,
+      "loss": 0.1899,
+      "step": 31870
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.296592950816921e-06,
+      "loss": 0.3106,
+      "step": 31871
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.296257115510554e-06,
+      "loss": 0.3736,
+      "step": 31872
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.295921280204189e-06,
+      "loss": 0.3474,
+      "step": 31873
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.295585444897823e-06,
+      "loss": 0.3448,
+      "step": 31874
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.295249609591456e-06,
+      "loss": 0.2684,
+      "step": 31875
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.294913774285091e-06,
+      "loss": 0.336,
+      "step": 31876
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.294577938978726e-06,
+      "loss": 0.3647,
+      "step": 31877
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.29424210367236e-06,
+      "loss": 0.2765,
+      "step": 31878
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.293906268365995e-06,
+      "loss": 0.4443,
+      "step": 31879
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.29357043305963e-06,
+      "loss": 0.1871,
+      "step": 31880
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.293234597753262e-06,
+      "loss": 0.3659,
+      "step": 31881
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.292898762446897e-06,
+      "loss": 0.3573,
+      "step": 31882
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.292562927140532e-06,
+      "loss": 0.2749,
+      "step": 31883
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.292227091834165e-06,
+      "loss": 0.3444,
+      "step": 31884
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.2918912565278e-06,
+      "loss": 0.3103,
+      "step": 31885
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.291555421221434e-06,
+      "loss": 0.2481,
+      "step": 31886
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.291219585915069e-06,
+      "loss": 0.2105,
+      "step": 31887
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.290883750608702e-06,
+      "loss": 0.2757,
+      "step": 31888
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.290547915302336e-06,
+      "loss": 0.2697,
+      "step": 31889
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.290212079995971e-06,
+      "loss": 0.2062,
+      "step": 31890
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.289876244689606e-06,
+      "loss": 0.3409,
+      "step": 31891
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.28954040938324e-06,
+      "loss": 0.4083,
+      "step": 31892
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.289204574076873e-06,
+      "loss": 0.31,
+      "step": 31893
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.288868738770508e-06,
+      "loss": 0.4208,
+      "step": 31894
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.288532903464142e-06,
+      "loss": 0.3568,
+      "step": 31895
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.288197068157775e-06,
+      "loss": 0.2475,
+      "step": 31896
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.28786123285141e-06,
+      "loss": 0.2787,
+      "step": 31897
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.287525397545045e-06,
+      "loss": 0.3472,
+      "step": 31898
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.28718956223868e-06,
+      "loss": 0.2869,
+      "step": 31899
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.286853726932314e-06,
+      "loss": 0.1878,
+      "step": 31900
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.286517891625947e-06,
+      "loss": 0.2141,
+      "step": 31901
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.286182056319582e-06,
+      "loss": 0.4712,
+      "step": 31902
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.285846221013216e-06,
+      "loss": 0.1329,
+      "step": 31903
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.28551038570685e-06,
+      "loss": 0.3813,
+      "step": 31904
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.285174550400484e-06,
+      "loss": 0.2215,
+      "step": 31905
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.284838715094118e-06,
+      "loss": 0.2304,
+      "step": 31906
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.284502879787753e-06,
+      "loss": 0.5396,
+      "step": 31907
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.284167044481388e-06,
+      "loss": 0.339,
+      "step": 31908
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.283831209175022e-06,
+      "loss": 0.3826,
+      "step": 31909
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.283495373868655e-06,
+      "loss": 0.3707,
+      "step": 31910
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.28315953856229e-06,
+      "loss": 0.2614,
+      "step": 31911
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.282823703255925e-06,
+      "loss": 0.3042,
+      "step": 31912
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.282487867949558e-06,
+      "loss": 0.4059,
+      "step": 31913
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.282152032643192e-06,
+      "loss": 0.2414,
+      "step": 31914
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.281816197336827e-06,
+      "loss": 0.2301,
+      "step": 31915
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.28148036203046e-06,
+      "loss": 0.3576,
+      "step": 31916
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.281144526724094e-06,
+      "loss": 0.2697,
+      "step": 31917
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.280808691417729e-06,
+      "loss": 0.2676,
+      "step": 31918
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.280472856111364e-06,
+      "loss": 0.1894,
+      "step": 31919
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.280137020804998e-06,
+      "loss": 0.3267,
+      "step": 31920
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.279801185498633e-06,
+      "loss": 0.3496,
+      "step": 31921
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.279465350192266e-06,
+      "loss": 0.3572,
+      "step": 31922
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.2791295148859e-06,
+      "loss": 0.2777,
+      "step": 31923
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.278793679579535e-06,
+      "loss": 0.3285,
+      "step": 31924
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.278457844273168e-06,
+      "loss": 0.4345,
+      "step": 31925
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.278122008966803e-06,
+      "loss": 0.3534,
+      "step": 31926
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.277786173660438e-06,
+      "loss": 0.1922,
+      "step": 31927
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.277450338354072e-06,
+      "loss": 0.2276,
+      "step": 31928
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.277114503047707e-06,
+      "loss": 0.3518,
+      "step": 31929
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.276778667741342e-06,
+      "loss": 0.2988,
+      "step": 31930
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.276442832434974e-06,
+      "loss": 0.3032,
+      "step": 31931
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.276106997128609e-06,
+      "loss": 0.288,
+      "step": 31932
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.275771161822244e-06,
+      "loss": 0.2338,
+      "step": 31933
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.275435326515877e-06,
+      "loss": 0.3689,
+      "step": 31934
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.275099491209511e-06,
+      "loss": 0.2736,
+      "step": 31935
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.274763655903146e-06,
+      "loss": 0.3073,
+      "step": 31936
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.274427820596779e-06,
+      "loss": 0.2893,
+      "step": 31937
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.274091985290414e-06,
+      "loss": 0.3098,
+      "step": 31938
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.273756149984048e-06,
+      "loss": 0.2418,
+      "step": 31939
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.273420314677683e-06,
+      "loss": 0.2195,
+      "step": 31940
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.273084479371318e-06,
+      "loss": 0.2428,
+      "step": 31941
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.272748644064952e-06,
+      "loss": 0.2418,
+      "step": 31942
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.272412808758585e-06,
+      "loss": 0.2885,
+      "step": 31943
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.27207697345222e-06,
+      "loss": 0.2427,
+      "step": 31944
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.271741138145854e-06,
+      "loss": 0.2572,
+      "step": 31945
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.271405302839487e-06,
+      "loss": 0.2791,
+      "step": 31946
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.271069467533122e-06,
+      "loss": 0.1704,
+      "step": 31947
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.270733632226757e-06,
+      "loss": 0.2802,
+      "step": 31948
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.270397796920391e-06,
+      "loss": 0.313,
+      "step": 31949
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.270061961614026e-06,
+      "loss": 0.2195,
+      "step": 31950
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.26972612630766e-06,
+      "loss": 0.2401,
+      "step": 31951
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.269390291001294e-06,
+      "loss": 0.2832,
+      "step": 31952
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.269054455694928e-06,
+      "loss": 0.289,
+      "step": 31953
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.268718620388563e-06,
+      "loss": 0.2198,
+      "step": 31954
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.268382785082196e-06,
+      "loss": 0.2649,
+      "step": 31955
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.26804694977583e-06,
+      "loss": 0.3774,
+      "step": 31956
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.267711114469465e-06,
+      "loss": 0.3832,
+      "step": 31957
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.267375279163098e-06,
+      "loss": 0.3439,
+      "step": 31958
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.267039443856733e-06,
+      "loss": 0.3432,
+      "step": 31959
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.266703608550367e-06,
+      "loss": 0.1675,
+      "step": 31960
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.266367773244002e-06,
+      "loss": 0.3472,
+      "step": 31961
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.266031937937637e-06,
+      "loss": 0.191,
+      "step": 31962
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.265696102631271e-06,
+      "loss": 0.2285,
+      "step": 31963
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.265360267324904e-06,
+      "loss": 0.4377,
+      "step": 31964
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.265024432018539e-06,
+      "loss": 0.1874,
+      "step": 31965
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.264688596712174e-06,
+      "loss": 0.4535,
+      "step": 31966
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.264352761405807e-06,
+      "loss": 0.224,
+      "step": 31967
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.264016926099441e-06,
+      "loss": 0.2385,
+      "step": 31968
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.263681090793076e-06,
+      "loss": 0.2582,
+      "step": 31969
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.26334525548671e-06,
+      "loss": 0.2626,
+      "step": 31970
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.263009420180345e-06,
+      "loss": 0.291,
+      "step": 31971
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.26267358487398e-06,
+      "loss": 0.3643,
+      "step": 31972
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.262337749567613e-06,
+      "loss": 0.3844,
+      "step": 31973
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.262001914261247e-06,
+      "loss": 0.2433,
+      "step": 31974
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.261666078954882e-06,
+      "loss": 0.3412,
+      "step": 31975
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.261330243648515e-06,
+      "loss": 0.2165,
+      "step": 31976
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.26099440834215e-06,
+      "loss": 0.2317,
+      "step": 31977
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.260658573035784e-06,
+      "loss": 0.2951,
+      "step": 31978
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.260322737729417e-06,
+      "loss": 0.3012,
+      "step": 31979
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.259986902423052e-06,
+      "loss": 0.2341,
+      "step": 31980
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.259651067116687e-06,
+      "loss": 0.2313,
+      "step": 31981
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.259315231810321e-06,
+      "loss": 0.4172,
+      "step": 31982
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.258979396503956e-06,
+      "loss": 0.3672,
+      "step": 31983
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.25864356119759e-06,
+      "loss": 0.2752,
+      "step": 31984
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.258307725891223e-06,
+      "loss": 0.3677,
+      "step": 31985
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.257971890584858e-06,
+      "loss": 0.3499,
+      "step": 31986
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.257636055278493e-06,
+      "loss": 0.2582,
+      "step": 31987
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.257300219972126e-06,
+      "loss": 0.2603,
+      "step": 31988
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.25696438466576e-06,
+      "loss": 0.3183,
+      "step": 31989
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.256628549359395e-06,
+      "loss": 0.4392,
+      "step": 31990
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.25629271405303e-06,
+      "loss": 0.3661,
+      "step": 31991
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.255956878746664e-06,
+      "loss": 0.3553,
+      "step": 31992
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.255621043440297e-06,
+      "loss": 0.3102,
+      "step": 31993
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.255285208133932e-06,
+      "loss": 0.2338,
+      "step": 31994
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.254949372827566e-06,
+      "loss": 0.3023,
+      "step": 31995
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.2546135375212e-06,
+      "loss": 0.3405,
+      "step": 31996
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.254277702214834e-06,
+      "loss": 0.2252,
+      "step": 31997
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.253941866908469e-06,
+      "loss": 0.2311,
+      "step": 31998
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.253606031602103e-06,
+      "loss": 0.3442,
+      "step": 31999
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.253270196295738e-06,
+      "loss": 0.4397,
+      "step": 32000
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.252934360989371e-06,
+      "loss": 0.3471,
+      "step": 32001
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.252598525683006e-06,
+      "loss": 0.4253,
+      "step": 32002
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.25226269037664e-06,
+      "loss": 0.1842,
+      "step": 32003
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.251926855070275e-06,
+      "loss": 0.1752,
+      "step": 32004
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.251591019763908e-06,
+      "loss": 0.4899,
+      "step": 32005
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.251255184457543e-06,
+      "loss": 0.4049,
+      "step": 32006
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.250919349151177e-06,
+      "loss": 0.2397,
+      "step": 32007
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.25058351384481e-06,
+      "loss": 0.242,
+      "step": 32008
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.250247678538445e-06,
+      "loss": 0.3511,
+      "step": 32009
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.24991184323208e-06,
+      "loss": 0.3179,
+      "step": 32010
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.249576007925714e-06,
+      "loss": 0.4598,
+      "step": 32011
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.249240172619349e-06,
+      "loss": 0.3691,
+      "step": 32012
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.248904337312983e-06,
+      "loss": 0.2358,
+      "step": 32013
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.248568502006616e-06,
+      "loss": 0.3776,
+      "step": 32014
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.248232666700251e-06,
+      "loss": 0.3085,
+      "step": 32015
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.247896831393886e-06,
+      "loss": 0.2237,
+      "step": 32016
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.247560996087519e-06,
+      "loss": 0.5337,
+      "step": 32017
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.247225160781153e-06,
+      "loss": 0.2927,
+      "step": 32018
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.246889325474788e-06,
+      "loss": 0.1961,
+      "step": 32019
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.246553490168423e-06,
+      "loss": 0.2354,
+      "step": 32020
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.246217654862057e-06,
+      "loss": 0.3634,
+      "step": 32021
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.245881819555692e-06,
+      "loss": 0.192,
+      "step": 32022
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.245545984249325e-06,
+      "loss": 0.2089,
+      "step": 32023
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.24521014894296e-06,
+      "loss": 0.2307,
+      "step": 32024
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.244874313636594e-06,
+      "loss": 0.1752,
+      "step": 32025
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.244538478330227e-06,
+      "loss": 0.3121,
+      "step": 32026
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.244202643023862e-06,
+      "loss": 0.2076,
+      "step": 32027
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.243866807717496e-06,
+      "loss": 0.3846,
+      "step": 32028
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.24353097241113e-06,
+      "loss": 0.1571,
+      "step": 32029
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.243195137104764e-06,
+      "loss": 0.191,
+      "step": 32030
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.242859301798399e-06,
+      "loss": 0.3285,
+      "step": 32031
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.242523466492033e-06,
+      "loss": 0.2676,
+      "step": 32032
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.242187631185668e-06,
+      "loss": 0.2986,
+      "step": 32033
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.241851795879302e-06,
+      "loss": 0.3619,
+      "step": 32034
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.241515960572935e-06,
+      "loss": 0.266,
+      "step": 32035
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.24118012526657e-06,
+      "loss": 0.326,
+      "step": 32036
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.240844289960205e-06,
+      "loss": 0.1827,
+      "step": 32037
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.240508454653838e-06,
+      "loss": 0.296,
+      "step": 32038
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.240172619347472e-06,
+      "loss": 0.2954,
+      "step": 32039
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.239836784041107e-06,
+      "loss": 0.2297,
+      "step": 32040
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.239500948734742e-06,
+      "loss": 0.3803,
+      "step": 32041
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.239165113428376e-06,
+      "loss": 0.2829,
+      "step": 32042
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.238829278122011e-06,
+      "loss": 0.5293,
+      "step": 32043
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.238493442815644e-06,
+      "loss": 0.3221,
+      "step": 32044
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.238157607509279e-06,
+      "loss": 0.3577,
+      "step": 32045
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.237821772202913e-06,
+      "loss": 0.1836,
+      "step": 32046
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.237485936896546e-06,
+      "loss": 0.2383,
+      "step": 32047
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.23715010159018e-06,
+      "loss": 0.3458,
+      "step": 32048
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.236814266283815e-06,
+      "loss": 0.3467,
+      "step": 32049
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.236478430977448e-06,
+      "loss": 0.2934,
+      "step": 32050
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.236142595671083e-06,
+      "loss": 0.2421,
+      "step": 32051
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.235806760364718e-06,
+      "loss": 0.2606,
+      "step": 32052
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.235470925058352e-06,
+      "loss": 0.4458,
+      "step": 32053
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.235135089751987e-06,
+      "loss": 0.4432,
+      "step": 32054
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.234799254445622e-06,
+      "loss": 0.4437,
+      "step": 32055
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.234463419139255e-06,
+      "loss": 0.444,
+      "step": 32056
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.23412758383289e-06,
+      "loss": 0.2703,
+      "step": 32057
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.233791748526524e-06,
+      "loss": 0.2249,
+      "step": 32058
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 9.233455913220157e-06,
+      "loss": 0.3067,
+      "step": 32059
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.233120077913791e-06,
+      "loss": 0.1924,
+      "step": 32060
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.232784242607426e-06,
+      "loss": 0.3407,
+      "step": 32061
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.23244840730106e-06,
+      "loss": 0.3498,
+      "step": 32062
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.232112571994695e-06,
+      "loss": 0.2441,
+      "step": 32063
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.23177673668833e-06,
+      "loss": 0.1899,
+      "step": 32064
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.231440901381963e-06,
+      "loss": 0.3429,
+      "step": 32065
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.231105066075598e-06,
+      "loss": 0.2275,
+      "step": 32066
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.230769230769232e-06,
+      "loss": 0.2338,
+      "step": 32067
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.230433395462865e-06,
+      "loss": 0.2491,
+      "step": 32068
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.2300975601565e-06,
+      "loss": 0.3011,
+      "step": 32069
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.229761724850135e-06,
+      "loss": 0.2972,
+      "step": 32070
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.229425889543768e-06,
+      "loss": 0.2963,
+      "step": 32071
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.229090054237402e-06,
+      "loss": 0.331,
+      "step": 32072
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.228754218931037e-06,
+      "loss": 0.3497,
+      "step": 32073
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.228418383624671e-06,
+      "loss": 0.1896,
+      "step": 32074
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.228082548318306e-06,
+      "loss": 0.2264,
+      "step": 32075
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.22774671301194e-06,
+      "loss": 0.2574,
+      "step": 32076
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.227410877705574e-06,
+      "loss": 0.2208,
+      "step": 32077
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.227075042399208e-06,
+      "loss": 0.2062,
+      "step": 32078
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.226739207092841e-06,
+      "loss": 0.2907,
+      "step": 32079
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.226403371786476e-06,
+      "loss": 0.2543,
+      "step": 32080
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.22606753648011e-06,
+      "loss": 0.2363,
+      "step": 32081
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.225731701173745e-06,
+      "loss": 0.2707,
+      "step": 32082
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.22539586586738e-06,
+      "loss": 0.2515,
+      "step": 32083
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.225060030561015e-06,
+      "loss": 0.3955,
+      "step": 32084
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.224724195254647e-06,
+      "loss": 0.2334,
+      "step": 32085
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.224388359948282e-06,
+      "loss": 0.2489,
+      "step": 32086
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.224052524641917e-06,
+      "loss": 0.3373,
+      "step": 32087
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.22371668933555e-06,
+      "loss": 0.2987,
+      "step": 32088
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.223380854029184e-06,
+      "loss": 0.3284,
+      "step": 32089
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.223045018722819e-06,
+      "loss": 0.2165,
+      "step": 32090
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.222709183416452e-06,
+      "loss": 0.159,
+      "step": 32091
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.222373348110087e-06,
+      "loss": 0.2866,
+      "step": 32092
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.222037512803721e-06,
+      "loss": 0.262,
+      "step": 32093
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.221701677497356e-06,
+      "loss": 0.2232,
+      "step": 32094
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.22136584219099e-06,
+      "loss": 0.2944,
+      "step": 32095
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.221030006884625e-06,
+      "loss": 0.1934,
+      "step": 32096
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.220694171578258e-06,
+      "loss": 0.2771,
+      "step": 32097
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.220358336271893e-06,
+      "loss": 0.3507,
+      "step": 32098
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.220022500965527e-06,
+      "loss": 0.2452,
+      "step": 32099
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.21968666565916e-06,
+      "loss": 0.3408,
+      "step": 32100
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.219350830352795e-06,
+      "loss": 0.1799,
+      "step": 32101
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.21901499504643e-06,
+      "loss": 0.2585,
+      "step": 32102
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.218679159740064e-06,
+      "loss": 0.3341,
+      "step": 32103
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.218343324433699e-06,
+      "loss": 0.2886,
+      "step": 32104
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.218007489127334e-06,
+      "loss": 0.3868,
+      "step": 32105
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.217671653820967e-06,
+      "loss": 0.2034,
+      "step": 32106
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.217335818514601e-06,
+      "loss": 0.2139,
+      "step": 32107
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.216999983208236e-06,
+      "loss": 0.4312,
+      "step": 32108
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.216664147901869e-06,
+      "loss": 0.3649,
+      "step": 32109
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.216328312595504e-06,
+      "loss": 0.4808,
+      "step": 32110
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.215992477289138e-06,
+      "loss": 0.2244,
+      "step": 32111
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.215656641982773e-06,
+      "loss": 0.2552,
+      "step": 32112
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.215320806676407e-06,
+      "loss": 0.2455,
+      "step": 32113
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.21498497137004e-06,
+      "loss": 0.2734,
+      "step": 32114
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.214649136063675e-06,
+      "loss": 0.349,
+      "step": 32115
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.21431330075731e-06,
+      "loss": 0.2692,
+      "step": 32116
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.213977465450944e-06,
+      "loss": 0.2056,
+      "step": 32117
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.213641630144577e-06,
+      "loss": 0.3058,
+      "step": 32118
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.213305794838212e-06,
+      "loss": 0.3355,
+      "step": 32119
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.212969959531847e-06,
+      "loss": 0.4874,
+      "step": 32120
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.21263412422548e-06,
+      "loss": 0.4262,
+      "step": 32121
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.212298288919114e-06,
+      "loss": 0.2055,
+      "step": 32122
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.211962453612749e-06,
+      "loss": 0.3625,
+      "step": 32123
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.211626618306383e-06,
+      "loss": 0.3326,
+      "step": 32124
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.211290783000018e-06,
+      "loss": 0.2896,
+      "step": 32125
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.210954947693653e-06,
+      "loss": 0.4387,
+      "step": 32126
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.210619112387286e-06,
+      "loss": 0.3175,
+      "step": 32127
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.21028327708092e-06,
+      "loss": 0.3947,
+      "step": 32128
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.209947441774555e-06,
+      "loss": 0.3339,
+      "step": 32129
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.209611606468188e-06,
+      "loss": 0.2846,
+      "step": 32130
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.209275771161823e-06,
+      "loss": 0.189,
+      "step": 32131
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.208939935855457e-06,
+      "loss": 0.4303,
+      "step": 32132
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.208604100549092e-06,
+      "loss": 0.282,
+      "step": 32133
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.208268265242727e-06,
+      "loss": 0.2004,
+      "step": 32134
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.207932429936361e-06,
+      "loss": 0.2655,
+      "step": 32135
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.207596594629994e-06,
+      "loss": 0.2542,
+      "step": 32136
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.207260759323629e-06,
+      "loss": 0.3482,
+      "step": 32137
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.206924924017263e-06,
+      "loss": 0.2464,
+      "step": 32138
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.206589088710896e-06,
+      "loss": 0.2848,
+      "step": 32139
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.206253253404531e-06,
+      "loss": 0.2456,
+      "step": 32140
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.205917418098166e-06,
+      "loss": 0.2871,
+      "step": 32141
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.205581582791799e-06,
+      "loss": 0.1806,
+      "step": 32142
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.205245747485433e-06,
+      "loss": 0.24,
+      "step": 32143
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.204909912179068e-06,
+      "loss": 0.3267,
+      "step": 32144
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.204574076872703e-06,
+      "loss": 0.2219,
+      "step": 32145
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.204238241566337e-06,
+      "loss": 0.3424,
+      "step": 32146
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.203902406259972e-06,
+      "loss": 0.2208,
+      "step": 32147
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.203566570953605e-06,
+      "loss": 0.2717,
+      "step": 32148
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.20323073564724e-06,
+      "loss": 0.5071,
+      "step": 32149
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.202894900340874e-06,
+      "loss": 0.4572,
+      "step": 32150
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.202559065034507e-06,
+      "loss": 0.3129,
+      "step": 32151
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.202223229728142e-06,
+      "loss": 0.3239,
+      "step": 32152
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.201887394421776e-06,
+      "loss": 0.2431,
+      "step": 32153
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.201551559115411e-06,
+      "loss": 0.2116,
+      "step": 32154
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.201215723809046e-06,
+      "loss": 0.3657,
+      "step": 32155
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.20087988850268e-06,
+      "loss": 0.3396,
+      "step": 32156
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.200544053196313e-06,
+      "loss": 0.3171,
+      "step": 32157
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.200208217889948e-06,
+      "loss": 0.415,
+      "step": 32158
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.199872382583583e-06,
+      "loss": 0.5402,
+      "step": 32159
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.199536547277216e-06,
+      "loss": 0.2693,
+      "step": 32160
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.19920071197085e-06,
+      "loss": 0.3615,
+      "step": 32161
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.198864876664485e-06,
+      "loss": 0.3602,
+      "step": 32162
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.198529041358118e-06,
+      "loss": 0.4331,
+      "step": 32163
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.198193206051752e-06,
+      "loss": 0.1734,
+      "step": 32164
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.197857370745387e-06,
+      "loss": 0.2792,
+      "step": 32165
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.197521535439022e-06,
+      "loss": 0.2557,
+      "step": 32166
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.197185700132656e-06,
+      "loss": 0.1956,
+      "step": 32167
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.19684986482629e-06,
+      "loss": 0.3895,
+      "step": 32168
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.196514029519924e-06,
+      "loss": 0.4375,
+      "step": 32169
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.196178194213559e-06,
+      "loss": 0.2322,
+      "step": 32170
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.195842358907192e-06,
+      "loss": 0.2378,
+      "step": 32171
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.195506523600826e-06,
+      "loss": 0.4408,
+      "step": 32172
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.195170688294461e-06,
+      "loss": 0.1017,
+      "step": 32173
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.194834852988096e-06,
+      "loss": 0.4362,
+      "step": 32174
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.19449901768173e-06,
+      "loss": 0.3623,
+      "step": 32175
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.194163182375365e-06,
+      "loss": 0.4009,
+      "step": 32176
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.193827347068998e-06,
+      "loss": 0.3761,
+      "step": 32177
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.193491511762632e-06,
+      "loss": 0.4227,
+      "step": 32178
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.193155676456267e-06,
+      "loss": 0.3648,
+      "step": 32179
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.1928198411499e-06,
+      "loss": 0.1999,
+      "step": 32180
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.192484005843535e-06,
+      "loss": 0.5406,
+      "step": 32181
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.19214817053717e-06,
+      "loss": 0.3565,
+      "step": 32182
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.191812335230802e-06,
+      "loss": 0.4456,
+      "step": 32183
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.191476499924437e-06,
+      "loss": 0.4475,
+      "step": 32184
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.191140664618072e-06,
+      "loss": 0.1931,
+      "step": 32185
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.190804829311706e-06,
+      "loss": 0.2417,
+      "step": 32186
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.190468994005341e-06,
+      "loss": 0.3064,
+      "step": 32187
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.190133158698976e-06,
+      "loss": 0.3737,
+      "step": 32188
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.189797323392608e-06,
+      "loss": 0.3546,
+      "step": 32189
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.189461488086243e-06,
+      "loss": 0.1645,
+      "step": 32190
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.189125652779878e-06,
+      "loss": 0.3528,
+      "step": 32191
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.18878981747351e-06,
+      "loss": 0.3347,
+      "step": 32192
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.188453982167145e-06,
+      "loss": 0.3164,
+      "step": 32193
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.18811814686078e-06,
+      "loss": 0.1663,
+      "step": 32194
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.187782311554415e-06,
+      "loss": 0.2073,
+      "step": 32195
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.18744647624805e-06,
+      "loss": 0.3301,
+      "step": 32196
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.187110640941684e-06,
+      "loss": 0.1821,
+      "step": 32197
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.186774805635317e-06,
+      "loss": 0.2166,
+      "step": 32198
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.186438970328952e-06,
+      "loss": 0.3095,
+      "step": 32199
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.186103135022586e-06,
+      "loss": 0.35,
+      "step": 32200
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.185767299716219e-06,
+      "loss": 0.319,
+      "step": 32201
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.185431464409854e-06,
+      "loss": 0.4066,
+      "step": 32202
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.185095629103488e-06,
+      "loss": 0.261,
+      "step": 32203
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.184759793797121e-06,
+      "loss": 0.2122,
+      "step": 32204
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.184423958490756e-06,
+      "loss": 0.1908,
+      "step": 32205
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.18408812318439e-06,
+      "loss": 0.2882,
+      "step": 32206
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.183752287878025e-06,
+      "loss": 0.2192,
+      "step": 32207
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.18341645257166e-06,
+      "loss": 0.2993,
+      "step": 32208
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.183080617265295e-06,
+      "loss": 0.2717,
+      "step": 32209
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.182744781958928e-06,
+      "loss": 0.2622,
+      "step": 32210
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.182408946652562e-06,
+      "loss": 0.2966,
+      "step": 32211
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.182073111346197e-06,
+      "loss": 0.2531,
+      "step": 32212
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.18173727603983e-06,
+      "loss": 0.2002,
+      "step": 32213
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.181401440733464e-06,
+      "loss": 0.2003,
+      "step": 32214
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.181065605427099e-06,
+      "loss": 0.1736,
+      "step": 32215
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.180729770120734e-06,
+      "loss": 0.3184,
+      "step": 32216
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.180393934814368e-06,
+      "loss": 0.3291,
+      "step": 32217
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.180058099508003e-06,
+      "loss": 0.3159,
+      "step": 32218
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.179722264201636e-06,
+      "loss": 0.375,
+      "step": 32219
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.17938642889527e-06,
+      "loss": 0.284,
+      "step": 32220
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.179050593588905e-06,
+      "loss": 0.3022,
+      "step": 32221
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.178714758282538e-06,
+      "loss": 0.4252,
+      "step": 32222
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.178378922976173e-06,
+      "loss": 0.2099,
+      "step": 32223
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.178043087669808e-06,
+      "loss": 0.2675,
+      "step": 32224
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.177707252363442e-06,
+      "loss": 0.2615,
+      "step": 32225
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.177371417057075e-06,
+      "loss": 0.3592,
+      "step": 32226
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.17703558175071e-06,
+      "loss": 0.2518,
+      "step": 32227
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.176699746444344e-06,
+      "loss": 0.499,
+      "step": 32228
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.176363911137979e-06,
+      "loss": 0.3733,
+      "step": 32229
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.176028075831614e-06,
+      "loss": 0.3019,
+      "step": 32230
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.175692240525247e-06,
+      "loss": 0.3779,
+      "step": 32231
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.175356405218881e-06,
+      "loss": 0.4398,
+      "step": 32232
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.175020569912516e-06,
+      "loss": 0.2822,
+      "step": 32233
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.174684734606149e-06,
+      "loss": 0.2587,
+      "step": 32234
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.174348899299784e-06,
+      "loss": 0.3932,
+      "step": 32235
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.174013063993418e-06,
+      "loss": 0.35,
+      "step": 32236
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.173677228687053e-06,
+      "loss": 0.345,
+      "step": 32237
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.173341393380688e-06,
+      "loss": 0.4047,
+      "step": 32238
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.173005558074322e-06,
+      "loss": 0.4049,
+      "step": 32239
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.172669722767955e-06,
+      "loss": 0.2725,
+      "step": 32240
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.17233388746159e-06,
+      "loss": 0.2707,
+      "step": 32241
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.171998052155224e-06,
+      "loss": 0.2262,
+      "step": 32242
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.171662216848857e-06,
+      "loss": 0.4107,
+      "step": 32243
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.171326381542492e-06,
+      "loss": 0.3893,
+      "step": 32244
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.170990546236127e-06,
+      "loss": 0.3671,
+      "step": 32245
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.170654710929761e-06,
+      "loss": 0.2407,
+      "step": 32246
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.170318875623396e-06,
+      "loss": 0.3907,
+      "step": 32247
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.16998304031703e-06,
+      "loss": 0.219,
+      "step": 32248
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.169647205010664e-06,
+      "loss": 0.3559,
+      "step": 32249
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.169311369704298e-06,
+      "loss": 0.2265,
+      "step": 32250
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.168975534397933e-06,
+      "loss": 0.3224,
+      "step": 32251
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.168639699091566e-06,
+      "loss": 0.2532,
+      "step": 32252
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.1683038637852e-06,
+      "loss": 0.2155,
+      "step": 32253
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.167968028478835e-06,
+      "loss": 0.3439,
+      "step": 32254
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.167632193172468e-06,
+      "loss": 0.3654,
+      "step": 32255
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.167296357866103e-06,
+      "loss": 0.2511,
+      "step": 32256
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 9.166960522559737e-06,
+      "loss": 0.4665,
+      "step": 32257
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.166624687253372e-06,
+      "loss": 0.1907,
+      "step": 32258
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.166288851947007e-06,
+      "loss": 0.1868,
+      "step": 32259
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.16595301664064e-06,
+      "loss": 0.2015,
+      "step": 32260
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.165617181334274e-06,
+      "loss": 0.2576,
+      "step": 32261
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.165281346027909e-06,
+      "loss": 0.4208,
+      "step": 32262
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.164945510721542e-06,
+      "loss": 0.2069,
+      "step": 32263
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.164609675415177e-06,
+      "loss": 0.3472,
+      "step": 32264
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.164273840108811e-06,
+      "loss": 0.1853,
+      "step": 32265
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.163938004802446e-06,
+      "loss": 0.3711,
+      "step": 32266
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.16360216949608e-06,
+      "loss": 0.3485,
+      "step": 32267
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.163266334189715e-06,
+      "loss": 0.283,
+      "step": 32268
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.162930498883348e-06,
+      "loss": 0.2622,
+      "step": 32269
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.162594663576983e-06,
+      "loss": 0.3339,
+      "step": 32270
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.162258828270617e-06,
+      "loss": 0.2085,
+      "step": 32271
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.16192299296425e-06,
+      "loss": 0.2947,
+      "step": 32272
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.161587157657885e-06,
+      "loss": 0.1782,
+      "step": 32273
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.16125132235152e-06,
+      "loss": 0.177,
+      "step": 32274
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.160915487045153e-06,
+      "loss": 0.3536,
+      "step": 32275
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.160579651738787e-06,
+      "loss": 0.3223,
+      "step": 32276
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.160243816432422e-06,
+      "loss": 0.2983,
+      "step": 32277
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.159907981126057e-06,
+      "loss": 0.4595,
+      "step": 32278
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.159572145819691e-06,
+      "loss": 0.2232,
+      "step": 32279
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.159236310513326e-06,
+      "loss": 0.3203,
+      "step": 32280
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.158900475206959e-06,
+      "loss": 0.2664,
+      "step": 32281
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.158564639900593e-06,
+      "loss": 0.2492,
+      "step": 32282
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.158228804594228e-06,
+      "loss": 0.3147,
+      "step": 32283
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.157892969287861e-06,
+      "loss": 0.3157,
+      "step": 32284
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.157557133981496e-06,
+      "loss": 0.1685,
+      "step": 32285
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.15722129867513e-06,
+      "loss": 0.2469,
+      "step": 32286
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.156885463368765e-06,
+      "loss": 0.2511,
+      "step": 32287
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.1565496280624e-06,
+      "loss": 0.2179,
+      "step": 32288
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.156213792756034e-06,
+      "loss": 0.4012,
+      "step": 32289
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.155877957449667e-06,
+      "loss": 0.1794,
+      "step": 32290
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.155542122143302e-06,
+      "loss": 0.2763,
+      "step": 32291
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.155206286836936e-06,
+      "loss": 0.1925,
+      "step": 32292
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.15487045153057e-06,
+      "loss": 0.3925,
+      "step": 32293
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.154534616224204e-06,
+      "loss": 0.19,
+      "step": 32294
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.154198780917839e-06,
+      "loss": 0.1735,
+      "step": 32295
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.153862945611472e-06,
+      "loss": 0.2746,
+      "step": 32296
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.153527110305106e-06,
+      "loss": 0.2103,
+      "step": 32297
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.153191274998741e-06,
+      "loss": 0.4119,
+      "step": 32298
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.152855439692376e-06,
+      "loss": 0.2793,
+      "step": 32299
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.15251960438601e-06,
+      "loss": 0.2616,
+      "step": 32300
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.152183769079645e-06,
+      "loss": 0.4524,
+      "step": 32301
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.151847933773278e-06,
+      "loss": 0.187,
+      "step": 32302
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.151512098466913e-06,
+      "loss": 0.2466,
+      "step": 32303
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.151176263160547e-06,
+      "loss": 0.3099,
+      "step": 32304
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.15084042785418e-06,
+      "loss": 0.3606,
+      "step": 32305
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.150504592547815e-06,
+      "loss": 0.1758,
+      "step": 32306
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.15016875724145e-06,
+      "loss": 0.3609,
+      "step": 32307
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.149832921935084e-06,
+      "loss": 0.3989,
+      "step": 32308
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.149497086628719e-06,
+      "loss": 0.352,
+      "step": 32309
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.149161251322353e-06,
+      "loss": 0.3589,
+      "step": 32310
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.148825416015986e-06,
+      "loss": 0.3312,
+      "step": 32311
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.148489580709621e-06,
+      "loss": 0.3036,
+      "step": 32312
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.148153745403256e-06,
+      "loss": 0.3171,
+      "step": 32313
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.147817910096889e-06,
+      "loss": 0.2402,
+      "step": 32314
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.147482074790523e-06,
+      "loss": 0.2781,
+      "step": 32315
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.147146239484158e-06,
+      "loss": 0.2207,
+      "step": 32316
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.14681040417779e-06,
+      "loss": 0.4636,
+      "step": 32317
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.146474568871425e-06,
+      "loss": 0.2858,
+      "step": 32318
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.14613873356506e-06,
+      "loss": 0.3037,
+      "step": 32319
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.145802898258695e-06,
+      "loss": 0.304,
+      "step": 32320
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.14546706295233e-06,
+      "loss": 0.4675,
+      "step": 32321
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.145131227645964e-06,
+      "loss": 0.4589,
+      "step": 32322
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.144795392339597e-06,
+      "loss": 0.3209,
+      "step": 32323
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.144459557033232e-06,
+      "loss": 0.2728,
+      "step": 32324
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.144123721726866e-06,
+      "loss": 0.2499,
+      "step": 32325
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.1437878864205e-06,
+      "loss": 0.2495,
+      "step": 32326
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.143452051114134e-06,
+      "loss": 0.356,
+      "step": 32327
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.143116215807769e-06,
+      "loss": 0.5992,
+      "step": 32328
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.142780380501403e-06,
+      "loss": 0.4937,
+      "step": 32329
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.142444545195038e-06,
+      "loss": 0.3322,
+      "step": 32330
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.142108709888672e-06,
+      "loss": 0.2017,
+      "step": 32331
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.141772874582305e-06,
+      "loss": 0.3623,
+      "step": 32332
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.14143703927594e-06,
+      "loss": 0.2981,
+      "step": 32333
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.141101203969575e-06,
+      "loss": 0.2917,
+      "step": 32334
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.140765368663208e-06,
+      "loss": 0.2667,
+      "step": 32335
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.140429533356842e-06,
+      "loss": 0.1732,
+      "step": 32336
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.140093698050477e-06,
+      "loss": 0.4146,
+      "step": 32337
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.139757862744112e-06,
+      "loss": 0.2755,
+      "step": 32338
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.139422027437745e-06,
+      "loss": 0.2737,
+      "step": 32339
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.13908619213138e-06,
+      "loss": 0.2836,
+      "step": 32340
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.138750356825014e-06,
+      "loss": 0.2924,
+      "step": 32341
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.138414521518649e-06,
+      "loss": 0.3726,
+      "step": 32342
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.138078686212283e-06,
+      "loss": 0.1832,
+      "step": 32343
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.137742850905916e-06,
+      "loss": 0.2218,
+      "step": 32344
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.13740701559955e-06,
+      "loss": 0.4369,
+      "step": 32345
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.137071180293185e-06,
+      "loss": 0.2284,
+      "step": 32346
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.136735344986818e-06,
+      "loss": 0.2528,
+      "step": 32347
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.136399509680453e-06,
+      "loss": 0.1929,
+      "step": 32348
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.136063674374088e-06,
+      "loss": 0.1721,
+      "step": 32349
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.135727839067722e-06,
+      "loss": 0.2751,
+      "step": 32350
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.135392003761357e-06,
+      "loss": 0.3467,
+      "step": 32351
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.13505616845499e-06,
+      "loss": 0.2279,
+      "step": 32352
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.134720333148625e-06,
+      "loss": 0.3067,
+      "step": 32353
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.13438449784226e-06,
+      "loss": 0.4114,
+      "step": 32354
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.134048662535892e-06,
+      "loss": 0.3666,
+      "step": 32355
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.133712827229527e-06,
+      "loss": 0.348,
+      "step": 32356
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.133376991923161e-06,
+      "loss": 0.2346,
+      "step": 32357
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.133041156616796e-06,
+      "loss": 0.2709,
+      "step": 32358
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.13270532131043e-06,
+      "loss": 0.3353,
+      "step": 32359
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.132369486004065e-06,
+      "loss": 0.185,
+      "step": 32360
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.132033650697698e-06,
+      "loss": 0.3817,
+      "step": 32361
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.131697815391333e-06,
+      "loss": 0.4441,
+      "step": 32362
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.131361980084968e-06,
+      "loss": 0.5056,
+      "step": 32363
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.1310261447786e-06,
+      "loss": 0.3422,
+      "step": 32364
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.130690309472235e-06,
+      "loss": 0.2601,
+      "step": 32365
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.13035447416587e-06,
+      "loss": 0.2934,
+      "step": 32366
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.130018638859503e-06,
+      "loss": 0.2885,
+      "step": 32367
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.129682803553138e-06,
+      "loss": 0.2373,
+      "step": 32368
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.129346968246772e-06,
+      "loss": 0.37,
+      "step": 32369
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.129011132940407e-06,
+      "loss": 0.2613,
+      "step": 32370
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.128675297634041e-06,
+      "loss": 0.2089,
+      "step": 32371
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.128339462327676e-06,
+      "loss": 0.2296,
+      "step": 32372
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.128003627021309e-06,
+      "loss": 0.3001,
+      "step": 32373
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.127667791714944e-06,
+      "loss": 0.5118,
+      "step": 32374
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.127331956408578e-06,
+      "loss": 0.1425,
+      "step": 32375
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.126996121102211e-06,
+      "loss": 0.3851,
+      "step": 32376
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.126660285795846e-06,
+      "loss": 0.2914,
+      "step": 32377
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.12632445048948e-06,
+      "loss": 0.2635,
+      "step": 32378
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.125988615183115e-06,
+      "loss": 0.2141,
+      "step": 32379
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.12565277987675e-06,
+      "loss": 0.4563,
+      "step": 32380
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.125316944570385e-06,
+      "loss": 0.2305,
+      "step": 32381
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.124981109264017e-06,
+      "loss": 0.1819,
+      "step": 32382
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.124645273957652e-06,
+      "loss": 0.4313,
+      "step": 32383
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.124309438651287e-06,
+      "loss": 0.3417,
+      "step": 32384
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.12397360334492e-06,
+      "loss": 0.3092,
+      "step": 32385
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.123637768038554e-06,
+      "loss": 0.3235,
+      "step": 32386
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.123301932732189e-06,
+      "loss": 0.2918,
+      "step": 32387
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.122966097425822e-06,
+      "loss": 0.2787,
+      "step": 32388
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.122630262119457e-06,
+      "loss": 0.2205,
+      "step": 32389
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.122294426813091e-06,
+      "loss": 0.2899,
+      "step": 32390
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.121958591506726e-06,
+      "loss": 0.2158,
+      "step": 32391
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.12162275620036e-06,
+      "loss": 0.1932,
+      "step": 32392
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.121286920893995e-06,
+      "loss": 0.2197,
+      "step": 32393
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.120951085587628e-06,
+      "loss": 0.2846,
+      "step": 32394
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.120615250281263e-06,
+      "loss": 0.4147,
+      "step": 32395
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.120279414974897e-06,
+      "loss": 0.3567,
+      "step": 32396
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.11994357966853e-06,
+      "loss": 0.3125,
+      "step": 32397
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.119607744362165e-06,
+      "loss": 0.2106,
+      "step": 32398
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.1192719090558e-06,
+      "loss": 0.243,
+      "step": 32399
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.118936073749434e-06,
+      "loss": 0.2053,
+      "step": 32400
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.118600238443069e-06,
+      "loss": 0.2317,
+      "step": 32401
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.118264403136704e-06,
+      "loss": 0.2679,
+      "step": 32402
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.117928567830337e-06,
+      "loss": 0.284,
+      "step": 32403
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.117592732523971e-06,
+      "loss": 0.2445,
+      "step": 32404
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.117256897217606e-06,
+      "loss": 0.2428,
+      "step": 32405
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.116921061911239e-06,
+      "loss": 0.2838,
+      "step": 32406
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.116585226604874e-06,
+      "loss": 0.2541,
+      "step": 32407
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.116249391298508e-06,
+      "loss": 0.2263,
+      "step": 32408
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.115913555992141e-06,
+      "loss": 0.3456,
+      "step": 32409
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.115577720685776e-06,
+      "loss": 0.3812,
+      "step": 32410
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.11524188537941e-06,
+      "loss": 0.2727,
+      "step": 32411
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.114906050073045e-06,
+      "loss": 0.3308,
+      "step": 32412
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.11457021476668e-06,
+      "loss": 0.1528,
+      "step": 32413
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.114234379460314e-06,
+      "loss": 0.2662,
+      "step": 32414
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.113898544153947e-06,
+      "loss": 0.348,
+      "step": 32415
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.113562708847582e-06,
+      "loss": 0.464,
+      "step": 32416
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.113226873541217e-06,
+      "loss": 0.3038,
+      "step": 32417
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.11289103823485e-06,
+      "loss": 0.3715,
+      "step": 32418
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.112555202928484e-06,
+      "loss": 0.2243,
+      "step": 32419
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.112219367622119e-06,
+      "loss": 0.4365,
+      "step": 32420
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.111883532315753e-06,
+      "loss": 0.2406,
+      "step": 32421
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.111547697009388e-06,
+      "loss": 0.3613,
+      "step": 32422
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.111211861703023e-06,
+      "loss": 0.2099,
+      "step": 32423
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.110876026396656e-06,
+      "loss": 0.3471,
+      "step": 32424
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.11054019109029e-06,
+      "loss": 0.2768,
+      "step": 32425
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.110204355783925e-06,
+      "loss": 0.1592,
+      "step": 32426
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.109868520477558e-06,
+      "loss": 0.2194,
+      "step": 32427
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.109532685171193e-06,
+      "loss": 0.4378,
+      "step": 32428
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.109196849864827e-06,
+      "loss": 0.409,
+      "step": 32429
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.10886101455846e-06,
+      "loss": 0.2836,
+      "step": 32430
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.108525179252095e-06,
+      "loss": 0.1269,
+      "step": 32431
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.10818934394573e-06,
+      "loss": 0.3027,
+      "step": 32432
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.107853508639364e-06,
+      "loss": 0.2052,
+      "step": 32433
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.107517673332999e-06,
+      "loss": 0.1641,
+      "step": 32434
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.107181838026633e-06,
+      "loss": 0.285,
+      "step": 32435
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.106846002720266e-06,
+      "loss": 0.2505,
+      "step": 32436
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.106510167413901e-06,
+      "loss": 0.176,
+      "step": 32437
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.106174332107536e-06,
+      "loss": 0.1048,
+      "step": 32438
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.105838496801169e-06,
+      "loss": 0.2082,
+      "step": 32439
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.105502661494803e-06,
+      "loss": 0.4263,
+      "step": 32440
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.105166826188438e-06,
+      "loss": 0.1836,
+      "step": 32441
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.104830990882073e-06,
+      "loss": 0.2305,
+      "step": 32442
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.104495155575707e-06,
+      "loss": 0.3328,
+      "step": 32443
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.10415932026934e-06,
+      "loss": 0.2345,
+      "step": 32444
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.103823484962975e-06,
+      "loss": 0.3424,
+      "step": 32445
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.10348764965661e-06,
+      "loss": 0.1321,
+      "step": 32446
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.103151814350242e-06,
+      "loss": 0.2932,
+      "step": 32447
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.102815979043877e-06,
+      "loss": 0.4144,
+      "step": 32448
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.102480143737512e-06,
+      "loss": 0.453,
+      "step": 32449
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.102144308431146e-06,
+      "loss": 0.3681,
+      "step": 32450
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.101808473124781e-06,
+      "loss": 0.2414,
+      "step": 32451
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.101472637818414e-06,
+      "loss": 0.1643,
+      "step": 32452
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.101136802512049e-06,
+      "loss": 0.3497,
+      "step": 32453
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.100800967205683e-06,
+      "loss": 0.2168,
+      "step": 32454
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.100465131899318e-06,
+      "loss": 0.2297,
+      "step": 32455
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 9.100129296592951e-06,
+      "loss": 0.4211,
+      "step": 32456
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.099793461286586e-06,
+      "loss": 0.2336,
+      "step": 32457
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.09945762598022e-06,
+      "loss": 0.2522,
+      "step": 32458
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.099121790673853e-06,
+      "loss": 0.2073,
+      "step": 32459
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.098785955367488e-06,
+      "loss": 0.3428,
+      "step": 32460
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.098450120061122e-06,
+      "loss": 0.2639,
+      "step": 32461
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.098114284754757e-06,
+      "loss": 0.0703,
+      "step": 32462
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.097778449448392e-06,
+      "loss": 0.4447,
+      "step": 32463
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.097442614142026e-06,
+      "loss": 0.2819,
+      "step": 32464
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.09710677883566e-06,
+      "loss": 0.1875,
+      "step": 32465
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.096770943529294e-06,
+      "loss": 0.3424,
+      "step": 32466
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.096435108222929e-06,
+      "loss": 0.2422,
+      "step": 32467
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.096099272916562e-06,
+      "loss": 0.4242,
+      "step": 32468
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.095763437610196e-06,
+      "loss": 0.2732,
+      "step": 32469
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.095427602303831e-06,
+      "loss": 0.2986,
+      "step": 32470
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.095091766997466e-06,
+      "loss": 0.3077,
+      "step": 32471
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.0947559316911e-06,
+      "loss": 0.3907,
+      "step": 32472
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.094420096384735e-06,
+      "loss": 0.1708,
+      "step": 32473
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.094084261078368e-06,
+      "loss": 0.508,
+      "step": 32474
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.093748425772002e-06,
+      "loss": 0.298,
+      "step": 32475
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.093412590465637e-06,
+      "loss": 0.3903,
+      "step": 32476
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.09307675515927e-06,
+      "loss": 0.2414,
+      "step": 32477
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.092740919852905e-06,
+      "loss": 0.2722,
+      "step": 32478
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.09240508454654e-06,
+      "loss": 0.285,
+      "step": 32479
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.092069249240172e-06,
+      "loss": 0.2084,
+      "step": 32480
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.091733413933807e-06,
+      "loss": 0.2935,
+      "step": 32481
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.091397578627442e-06,
+      "loss": 0.2522,
+      "step": 32482
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.091061743321076e-06,
+      "loss": 0.3495,
+      "step": 32483
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.090725908014711e-06,
+      "loss": 0.2065,
+      "step": 32484
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.090390072708346e-06,
+      "loss": 0.1803,
+      "step": 32485
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.090054237401978e-06,
+      "loss": 0.3459,
+      "step": 32486
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.089718402095613e-06,
+      "loss": 0.2799,
+      "step": 32487
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.089382566789248e-06,
+      "loss": 0.3121,
+      "step": 32488
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.08904673148288e-06,
+      "loss": 0.4372,
+      "step": 32489
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.088710896176515e-06,
+      "loss": 0.2478,
+      "step": 32490
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.08837506087015e-06,
+      "loss": 0.3523,
+      "step": 32491
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.088039225563785e-06,
+      "loss": 0.1601,
+      "step": 32492
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.08770339025742e-06,
+      "loss": 0.5653,
+      "step": 32493
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.087367554951054e-06,
+      "loss": 0.2746,
+      "step": 32494
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.087031719644687e-06,
+      "loss": 0.2434,
+      "step": 32495
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.086695884338322e-06,
+      "loss": 0.1986,
+      "step": 32496
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.086360049031956e-06,
+      "loss": 0.2295,
+      "step": 32497
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.086024213725589e-06,
+      "loss": 0.2604,
+      "step": 32498
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.085688378419224e-06,
+      "loss": 0.2665,
+      "step": 32499
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.085352543112858e-06,
+      "loss": 0.2649,
+      "step": 32500
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.085016707806491e-06,
+      "loss": 0.3253,
+      "step": 32501
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.084680872500126e-06,
+      "loss": 0.205,
+      "step": 32502
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.08434503719376e-06,
+      "loss": 0.156,
+      "step": 32503
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.084009201887395e-06,
+      "loss": 0.3434,
+      "step": 32504
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.08367336658103e-06,
+      "loss": 0.1721,
+      "step": 32505
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.083337531274665e-06,
+      "loss": 0.2367,
+      "step": 32506
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.083001695968298e-06,
+      "loss": 0.2779,
+      "step": 32507
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.082665860661932e-06,
+      "loss": 0.3155,
+      "step": 32508
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.082330025355567e-06,
+      "loss": 0.4128,
+      "step": 32509
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.0819941900492e-06,
+      "loss": 0.2362,
+      "step": 32510
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.081658354742834e-06,
+      "loss": 0.162,
+      "step": 32511
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.081322519436469e-06,
+      "loss": 0.1886,
+      "step": 32512
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.080986684130104e-06,
+      "loss": 0.3822,
+      "step": 32513
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.080650848823738e-06,
+      "loss": 0.398,
+      "step": 32514
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.080315013517373e-06,
+      "loss": 0.2282,
+      "step": 32515
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.079979178211006e-06,
+      "loss": 0.2023,
+      "step": 32516
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.07964334290464e-06,
+      "loss": 0.1688,
+      "step": 32517
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.079307507598275e-06,
+      "loss": 0.3288,
+      "step": 32518
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.078971672291908e-06,
+      "loss": 0.1372,
+      "step": 32519
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.078635836985543e-06,
+      "loss": 0.2492,
+      "step": 32520
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.078300001679178e-06,
+      "loss": 0.2863,
+      "step": 32521
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.07796416637281e-06,
+      "loss": 0.4019,
+      "step": 32522
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.077628331066445e-06,
+      "loss": 0.309,
+      "step": 32523
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.07729249576008e-06,
+      "loss": 0.2448,
+      "step": 32524
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.076956660453714e-06,
+      "loss": 0.2999,
+      "step": 32525
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.076620825147349e-06,
+      "loss": 0.3815,
+      "step": 32526
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.076284989840984e-06,
+      "loss": 0.2989,
+      "step": 32527
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.075949154534617e-06,
+      "loss": 0.2521,
+      "step": 32528
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.075613319228251e-06,
+      "loss": 0.2892,
+      "step": 32529
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.075277483921884e-06,
+      "loss": 0.1873,
+      "step": 32530
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.074941648615519e-06,
+      "loss": 0.2443,
+      "step": 32531
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.074605813309154e-06,
+      "loss": 0.1663,
+      "step": 32532
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.074269978002788e-06,
+      "loss": 0.1764,
+      "step": 32533
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.073934142696423e-06,
+      "loss": 0.5129,
+      "step": 32534
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.073598307390058e-06,
+      "loss": 0.1532,
+      "step": 32535
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.07326247208369e-06,
+      "loss": 0.3053,
+      "step": 32536
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.072926636777325e-06,
+      "loss": 0.2914,
+      "step": 32537
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.07259080147096e-06,
+      "loss": 0.302,
+      "step": 32538
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.072254966164593e-06,
+      "loss": 0.2753,
+      "step": 32539
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.071919130858227e-06,
+      "loss": 0.3675,
+      "step": 32540
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.071583295551862e-06,
+      "loss": 0.3306,
+      "step": 32541
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.071247460245497e-06,
+      "loss": 0.2248,
+      "step": 32542
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.07091162493913e-06,
+      "loss": 0.2405,
+      "step": 32543
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.070575789632764e-06,
+      "loss": 0.2348,
+      "step": 32544
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.070239954326399e-06,
+      "loss": 0.4146,
+      "step": 32545
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.069904119020034e-06,
+      "loss": 0.227,
+      "step": 32546
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.069568283713668e-06,
+      "loss": 0.1038,
+      "step": 32547
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.069232448407301e-06,
+      "loss": 0.3479,
+      "step": 32548
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.068896613100936e-06,
+      "loss": 0.2736,
+      "step": 32549
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.06856077779457e-06,
+      "loss": 0.2723,
+      "step": 32550
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.068224942488203e-06,
+      "loss": 0.27,
+      "step": 32551
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.067889107181838e-06,
+      "loss": 0.2105,
+      "step": 32552
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.067553271875473e-06,
+      "loss": 0.2712,
+      "step": 32553
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.067217436569107e-06,
+      "loss": 0.2809,
+      "step": 32554
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.066881601262742e-06,
+      "loss": 0.2207,
+      "step": 32555
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.066545765956377e-06,
+      "loss": 0.1907,
+      "step": 32556
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.06620993065001e-06,
+      "loss": 0.2059,
+      "step": 32557
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.065874095343644e-06,
+      "loss": 0.2462,
+      "step": 32558
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.065538260037279e-06,
+      "loss": 0.4047,
+      "step": 32559
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.065202424730912e-06,
+      "loss": 0.4038,
+      "step": 32560
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.064866589424547e-06,
+      "loss": 0.4003,
+      "step": 32561
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.064530754118181e-06,
+      "loss": 0.1793,
+      "step": 32562
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.064194918811816e-06,
+      "loss": 0.3724,
+      "step": 32563
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.06385908350545e-06,
+      "loss": 0.2533,
+      "step": 32564
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.063523248199083e-06,
+      "loss": 0.3942,
+      "step": 32565
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.063187412892718e-06,
+      "loss": 0.2676,
+      "step": 32566
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.062851577586353e-06,
+      "loss": 0.295,
+      "step": 32567
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.062515742279987e-06,
+      "loss": 0.5474,
+      "step": 32568
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.06217990697362e-06,
+      "loss": 0.3216,
+      "step": 32569
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.061844071667255e-06,
+      "loss": 0.2714,
+      "step": 32570
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.06150823636089e-06,
+      "loss": 0.2664,
+      "step": 32571
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.061172401054523e-06,
+      "loss": 0.456,
+      "step": 32572
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.060836565748157e-06,
+      "loss": 0.2126,
+      "step": 32573
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.060500730441792e-06,
+      "loss": 0.2216,
+      "step": 32574
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.060164895135427e-06,
+      "loss": 0.3367,
+      "step": 32575
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.059829059829061e-06,
+      "loss": 0.1593,
+      "step": 32576
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.059493224522696e-06,
+      "loss": 0.4578,
+      "step": 32577
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.059157389216329e-06,
+      "loss": 0.3592,
+      "step": 32578
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.058821553909963e-06,
+      "loss": 0.4067,
+      "step": 32579
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.058485718603598e-06,
+      "loss": 0.4451,
+      "step": 32580
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.058149883297231e-06,
+      "loss": 0.1693,
+      "step": 32581
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.057814047990866e-06,
+      "loss": 0.2292,
+      "step": 32582
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.0574782126845e-06,
+      "loss": 0.3879,
+      "step": 32583
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.057142377378135e-06,
+      "loss": 0.347,
+      "step": 32584
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.05680654207177e-06,
+      "loss": 0.3925,
+      "step": 32585
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.056470706765404e-06,
+      "loss": 0.3118,
+      "step": 32586
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.056134871459037e-06,
+      "loss": 0.3267,
+      "step": 32587
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.055799036152672e-06,
+      "loss": 0.3106,
+      "step": 32588
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.055463200846306e-06,
+      "loss": 0.4577,
+      "step": 32589
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.05512736553994e-06,
+      "loss": 0.3234,
+      "step": 32590
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.054791530233574e-06,
+      "loss": 0.2023,
+      "step": 32591
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.054455694927209e-06,
+      "loss": 0.1004,
+      "step": 32592
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.054119859620842e-06,
+      "loss": 0.3545,
+      "step": 32593
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.053784024314476e-06,
+      "loss": 0.3012,
+      "step": 32594
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.053448189008111e-06,
+      "loss": 0.4177,
+      "step": 32595
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.053112353701746e-06,
+      "loss": 0.2322,
+      "step": 32596
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.05277651839538e-06,
+      "loss": 0.2417,
+      "step": 32597
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.052440683089015e-06,
+      "loss": 0.2103,
+      "step": 32598
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.052104847782648e-06,
+      "loss": 0.4155,
+      "step": 32599
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.051769012476283e-06,
+      "loss": 0.2858,
+      "step": 32600
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.051433177169917e-06,
+      "loss": 0.2667,
+      "step": 32601
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.05109734186355e-06,
+      "loss": 0.2254,
+      "step": 32602
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.050761506557185e-06,
+      "loss": 0.2291,
+      "step": 32603
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.05042567125082e-06,
+      "loss": 0.2269,
+      "step": 32604
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.050089835944454e-06,
+      "loss": 0.4075,
+      "step": 32605
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.049754000638089e-06,
+      "loss": 0.2706,
+      "step": 32606
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.049418165331723e-06,
+      "loss": 0.3038,
+      "step": 32607
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.049082330025356e-06,
+      "loss": 0.1679,
+      "step": 32608
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.048746494718991e-06,
+      "loss": 0.2655,
+      "step": 32609
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.048410659412626e-06,
+      "loss": 0.305,
+      "step": 32610
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.048074824106259e-06,
+      "loss": 0.2156,
+      "step": 32611
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.047738988799893e-06,
+      "loss": 0.1542,
+      "step": 32612
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.047403153493528e-06,
+      "loss": 0.2809,
+      "step": 32613
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.04706731818716e-06,
+      "loss": 0.2608,
+      "step": 32614
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.046731482880795e-06,
+      "loss": 0.2644,
+      "step": 32615
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.04639564757443e-06,
+      "loss": 0.2381,
+      "step": 32616
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.046059812268065e-06,
+      "loss": 0.2888,
+      "step": 32617
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.0457239769617e-06,
+      "loss": 0.3977,
+      "step": 32618
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.045388141655332e-06,
+      "loss": 0.1968,
+      "step": 32619
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.045052306348967e-06,
+      "loss": 0.3669,
+      "step": 32620
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.044716471042602e-06,
+      "loss": 0.2823,
+      "step": 32621
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.044380635736235e-06,
+      "loss": 0.5308,
+      "step": 32622
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.04404480042987e-06,
+      "loss": 0.1993,
+      "step": 32623
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.043708965123504e-06,
+      "loss": 0.4261,
+      "step": 32624
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.043373129817139e-06,
+      "loss": 0.3198,
+      "step": 32625
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.043037294510773e-06,
+      "loss": 0.3066,
+      "step": 32626
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.042701459204408e-06,
+      "loss": 0.2509,
+      "step": 32627
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.04236562389804e-06,
+      "loss": 0.3517,
+      "step": 32628
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.042029788591675e-06,
+      "loss": 0.2478,
+      "step": 32629
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.04169395328531e-06,
+      "loss": 0.3619,
+      "step": 32630
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.041358117978943e-06,
+      "loss": 0.2883,
+      "step": 32631
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.041022282672578e-06,
+      "loss": 0.4119,
+      "step": 32632
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.040686447366212e-06,
+      "loss": 0.2822,
+      "step": 32633
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.040350612059845e-06,
+      "loss": 0.2168,
+      "step": 32634
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.04001477675348e-06,
+      "loss": 0.2244,
+      "step": 32635
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.039678941447115e-06,
+      "loss": 0.2155,
+      "step": 32636
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.03934310614075e-06,
+      "loss": 0.162,
+      "step": 32637
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.039007270834384e-06,
+      "loss": 0.3481,
+      "step": 32638
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.038671435528019e-06,
+      "loss": 0.4149,
+      "step": 32639
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.038335600221651e-06,
+      "loss": 0.2527,
+      "step": 32640
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.037999764915286e-06,
+      "loss": 0.3266,
+      "step": 32641
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.03766392960892e-06,
+      "loss": 0.3178,
+      "step": 32642
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.037328094302554e-06,
+      "loss": 0.3016,
+      "step": 32643
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.036992258996188e-06,
+      "loss": 0.2645,
+      "step": 32644
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.036656423689823e-06,
+      "loss": 0.3104,
+      "step": 32645
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.036320588383458e-06,
+      "loss": 0.1284,
+      "step": 32646
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.035984753077092e-06,
+      "loss": 0.3714,
+      "step": 32647
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.035648917770727e-06,
+      "loss": 0.1618,
+      "step": 32648
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.03531308246436e-06,
+      "loss": 0.3097,
+      "step": 32649
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.034977247157995e-06,
+      "loss": 0.2324,
+      "step": 32650
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.03464141185163e-06,
+      "loss": 0.3571,
+      "step": 32651
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.034305576545262e-06,
+      "loss": 0.3084,
+      "step": 32652
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.033969741238897e-06,
+      "loss": 0.3531,
+      "step": 32653
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 9.033633905932531e-06,
+      "loss": 0.3134,
+      "step": 32654
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.033298070626166e-06,
+      "loss": 0.346,
+      "step": 32655
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.032962235319799e-06,
+      "loss": 0.3639,
+      "step": 32656
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.032626400013434e-06,
+      "loss": 0.3839,
+      "step": 32657
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.032290564707068e-06,
+      "loss": 0.2347,
+      "step": 32658
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.031954729400703e-06,
+      "loss": 0.4178,
+      "step": 32659
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.031618894094338e-06,
+      "loss": 0.2219,
+      "step": 32660
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.03128305878797e-06,
+      "loss": 0.4669,
+      "step": 32661
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.030947223481605e-06,
+      "loss": 0.2594,
+      "step": 32662
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.03061138817524e-06,
+      "loss": 0.2409,
+      "step": 32663
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.030275552868873e-06,
+      "loss": 0.1372,
+      "step": 32664
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.029939717562507e-06,
+      "loss": 0.2922,
+      "step": 32665
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.029603882256142e-06,
+      "loss": 0.235,
+      "step": 32666
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.029268046949777e-06,
+      "loss": 0.3139,
+      "step": 32667
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.028932211643411e-06,
+      "loss": 0.2582,
+      "step": 32668
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.028596376337046e-06,
+      "loss": 0.1778,
+      "step": 32669
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.028260541030679e-06,
+      "loss": 0.2776,
+      "step": 32670
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.027924705724314e-06,
+      "loss": 0.2808,
+      "step": 32671
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.027588870417948e-06,
+      "loss": 0.245,
+      "step": 32672
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.027253035111581e-06,
+      "loss": 0.3055,
+      "step": 32673
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.026917199805216e-06,
+      "loss": 0.5258,
+      "step": 32674
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.02658136449885e-06,
+      "loss": 0.3908,
+      "step": 32675
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.026245529192485e-06,
+      "loss": 0.2933,
+      "step": 32676
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.02590969388612e-06,
+      "loss": 0.3198,
+      "step": 32677
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.025573858579753e-06,
+      "loss": 0.2464,
+      "step": 32678
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.025238023273387e-06,
+      "loss": 0.2098,
+      "step": 32679
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.024902187967022e-06,
+      "loss": 0.3606,
+      "step": 32680
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.024566352660657e-06,
+      "loss": 0.4405,
+      "step": 32681
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.02423051735429e-06,
+      "loss": 0.3287,
+      "step": 32682
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.023894682047924e-06,
+      "loss": 0.4193,
+      "step": 32683
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.023558846741559e-06,
+      "loss": 0.2785,
+      "step": 32684
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.023223011435192e-06,
+      "loss": 0.1646,
+      "step": 32685
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.022887176128827e-06,
+      "loss": 0.4681,
+      "step": 32686
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.022551340822461e-06,
+      "loss": 0.2162,
+      "step": 32687
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.022215505516096e-06,
+      "loss": 0.3546,
+      "step": 32688
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.02187967020973e-06,
+      "loss": 0.2442,
+      "step": 32689
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.021543834903365e-06,
+      "loss": 0.2267,
+      "step": 32690
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.021207999596998e-06,
+      "loss": 0.0935,
+      "step": 32691
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.020872164290633e-06,
+      "loss": 0.1786,
+      "step": 32692
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.020536328984267e-06,
+      "loss": 0.2357,
+      "step": 32693
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.0202004936779e-06,
+      "loss": 0.2897,
+      "step": 32694
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.019864658371535e-06,
+      "loss": 0.2225,
+      "step": 32695
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.01952882306517e-06,
+      "loss": 0.3426,
+      "step": 32696
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.019192987758804e-06,
+      "loss": 0.2563,
+      "step": 32697
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.018857152452439e-06,
+      "loss": 0.2096,
+      "step": 32698
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.018521317146074e-06,
+      "loss": 0.2664,
+      "step": 32699
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.018185481839707e-06,
+      "loss": 0.2433,
+      "step": 32700
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.017849646533341e-06,
+      "loss": 0.2679,
+      "step": 32701
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.017513811226976e-06,
+      "loss": 0.2641,
+      "step": 32702
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.017177975920609e-06,
+      "loss": 0.1746,
+      "step": 32703
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.016842140614243e-06,
+      "loss": 0.2528,
+      "step": 32704
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.016506305307878e-06,
+      "loss": 0.4178,
+      "step": 32705
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.016170470001511e-06,
+      "loss": 0.2758,
+      "step": 32706
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.015834634695146e-06,
+      "loss": 0.3248,
+      "step": 32707
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.01549879938878e-06,
+      "loss": 0.1964,
+      "step": 32708
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.015162964082415e-06,
+      "loss": 0.3601,
+      "step": 32709
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.01482712877605e-06,
+      "loss": 0.1344,
+      "step": 32710
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.014491293469683e-06,
+      "loss": 0.4969,
+      "step": 32711
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.014155458163317e-06,
+      "loss": 0.1673,
+      "step": 32712
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.013819622856952e-06,
+      "loss": 0.2274,
+      "step": 32713
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.013483787550585e-06,
+      "loss": 0.3759,
+      "step": 32714
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.01314795224422e-06,
+      "loss": 0.192,
+      "step": 32715
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.012812116937854e-06,
+      "loss": 0.3996,
+      "step": 32716
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.012476281631489e-06,
+      "loss": 0.1882,
+      "step": 32717
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.012140446325123e-06,
+      "loss": 0.2198,
+      "step": 32718
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.011804611018758e-06,
+      "loss": 0.2871,
+      "step": 32719
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.011468775712391e-06,
+      "loss": 0.284,
+      "step": 32720
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.011132940406026e-06,
+      "loss": 0.3778,
+      "step": 32721
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.01079710509966e-06,
+      "loss": 0.3818,
+      "step": 32722
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.010461269793293e-06,
+      "loss": 0.446,
+      "step": 32723
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.010125434486928e-06,
+      "loss": 0.2186,
+      "step": 32724
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.009789599180563e-06,
+      "loss": 0.3266,
+      "step": 32725
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.009453763874196e-06,
+      "loss": 0.2509,
+      "step": 32726
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.00911792856783e-06,
+      "loss": 0.2152,
+      "step": 32727
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.008782093261465e-06,
+      "loss": 0.2531,
+      "step": 32728
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.0084462579551e-06,
+      "loss": 0.4189,
+      "step": 32729
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.008110422648734e-06,
+      "loss": 0.2067,
+      "step": 32730
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.007774587342369e-06,
+      "loss": 0.2081,
+      "step": 32731
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.007438752036002e-06,
+      "loss": 0.2473,
+      "step": 32732
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.007102916729636e-06,
+      "loss": 0.452,
+      "step": 32733
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.006767081423271e-06,
+      "loss": 0.4436,
+      "step": 32734
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.006431246116904e-06,
+      "loss": 0.2738,
+      "step": 32735
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.006095410810539e-06,
+      "loss": 0.5636,
+      "step": 32736
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.005759575504173e-06,
+      "loss": 0.7488,
+      "step": 32737
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.005423740197808e-06,
+      "loss": 0.398,
+      "step": 32738
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.005087904891443e-06,
+      "loss": 0.3183,
+      "step": 32739
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.004752069585077e-06,
+      "loss": 0.1743,
+      "step": 32740
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.00441623427871e-06,
+      "loss": 0.287,
+      "step": 32741
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.004080398972345e-06,
+      "loss": 0.2317,
+      "step": 32742
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.00374456366598e-06,
+      "loss": 0.3505,
+      "step": 32743
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.003408728359612e-06,
+      "loss": 0.3,
+      "step": 32744
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.003072893053247e-06,
+      "loss": 0.2138,
+      "step": 32745
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.002737057746882e-06,
+      "loss": 0.2823,
+      "step": 32746
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.002401222440515e-06,
+      "loss": 0.2367,
+      "step": 32747
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.00206538713415e-06,
+      "loss": 0.3466,
+      "step": 32748
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.001729551827784e-06,
+      "loss": 0.22,
+      "step": 32749
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.001393716521419e-06,
+      "loss": 0.2761,
+      "step": 32750
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.001057881215053e-06,
+      "loss": 0.2861,
+      "step": 32751
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.000722045908688e-06,
+      "loss": 0.3212,
+      "step": 32752
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.000386210602321e-06,
+      "loss": 0.3366,
+      "step": 32753
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 9.000050375295956e-06,
+      "loss": 0.4072,
+      "step": 32754
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.99971453998959e-06,
+      "loss": 0.3186,
+      "step": 32755
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.999378704683223e-06,
+      "loss": 0.3545,
+      "step": 32756
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.999042869376858e-06,
+      "loss": 0.3304,
+      "step": 32757
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.998707034070492e-06,
+      "loss": 0.2094,
+      "step": 32758
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.998371198764127e-06,
+      "loss": 0.2638,
+      "step": 32759
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.998035363457762e-06,
+      "loss": 0.2112,
+      "step": 32760
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.997699528151396e-06,
+      "loss": 0.3002,
+      "step": 32761
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.99736369284503e-06,
+      "loss": 0.2502,
+      "step": 32762
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.997027857538664e-06,
+      "loss": 0.1598,
+      "step": 32763
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.996692022232299e-06,
+      "loss": 0.2869,
+      "step": 32764
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.996356186925932e-06,
+      "loss": 0.4536,
+      "step": 32765
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.996020351619566e-06,
+      "loss": 0.2369,
+      "step": 32766
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.995684516313201e-06,
+      "loss": 0.2785,
+      "step": 32767
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.995348681006834e-06,
+      "loss": 0.3432,
+      "step": 32768
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.995012845700468e-06,
+      "loss": 0.3992,
+      "step": 32769
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.994677010394103e-06,
+      "loss": 0.2629,
+      "step": 32770
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.994341175087738e-06,
+      "loss": 0.3712,
+      "step": 32771
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.994005339781372e-06,
+      "loss": 0.2561,
+      "step": 32772
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.993669504475007e-06,
+      "loss": 0.3316,
+      "step": 32773
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.99333366916864e-06,
+      "loss": 0.4847,
+      "step": 32774
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.992997833862275e-06,
+      "loss": 0.2953,
+      "step": 32775
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.99266199855591e-06,
+      "loss": 0.2704,
+      "step": 32776
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.992326163249542e-06,
+      "loss": 0.392,
+      "step": 32777
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.991990327943177e-06,
+      "loss": 0.2998,
+      "step": 32778
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.991654492636812e-06,
+      "loss": 0.3658,
+      "step": 32779
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.991318657330446e-06,
+      "loss": 0.2691,
+      "step": 32780
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.990982822024081e-06,
+      "loss": 0.3169,
+      "step": 32781
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.990646986717716e-06,
+      "loss": 0.3837,
+      "step": 32782
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.990311151411348e-06,
+      "loss": 0.3131,
+      "step": 32783
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.989975316104983e-06,
+      "loss": 0.2121,
+      "step": 32784
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.989639480798618e-06,
+      "loss": 0.2707,
+      "step": 32785
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.98930364549225e-06,
+      "loss": 0.3341,
+      "step": 32786
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.988967810185885e-06,
+      "loss": 0.3224,
+      "step": 32787
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.98863197487952e-06,
+      "loss": 0.1532,
+      "step": 32788
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.988296139573155e-06,
+      "loss": 0.2949,
+      "step": 32789
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.98796030426679e-06,
+      "loss": 0.2669,
+      "step": 32790
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.987624468960422e-06,
+      "loss": 0.1603,
+      "step": 32791
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.987288633654057e-06,
+      "loss": 0.3504,
+      "step": 32792
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.986952798347692e-06,
+      "loss": 0.2585,
+      "step": 32793
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.986616963041326e-06,
+      "loss": 0.3478,
+      "step": 32794
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.986281127734959e-06,
+      "loss": 0.2863,
+      "step": 32795
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.985945292428594e-06,
+      "loss": 0.3394,
+      "step": 32796
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.985609457122228e-06,
+      "loss": 0.2659,
+      "step": 32797
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.985273621815861e-06,
+      "loss": 0.2727,
+      "step": 32798
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.984937786509496e-06,
+      "loss": 0.4483,
+      "step": 32799
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.98460195120313e-06,
+      "loss": 0.2347,
+      "step": 32800
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.984266115896765e-06,
+      "loss": 0.2404,
+      "step": 32801
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.9839302805904e-06,
+      "loss": 0.2614,
+      "step": 32802
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.983594445284033e-06,
+      "loss": 0.4538,
+      "step": 32803
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.983258609977668e-06,
+      "loss": 0.3543,
+      "step": 32804
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.982922774671302e-06,
+      "loss": 0.1803,
+      "step": 32805
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.982586939364935e-06,
+      "loss": 0.1845,
+      "step": 32806
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.98225110405857e-06,
+      "loss": 0.3243,
+      "step": 32807
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.981915268752204e-06,
+      "loss": 0.3315,
+      "step": 32808
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.981579433445839e-06,
+      "loss": 0.3835,
+      "step": 32809
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.981243598139474e-06,
+      "loss": 0.307,
+      "step": 32810
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.980907762833108e-06,
+      "loss": 0.2893,
+      "step": 32811
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.980571927526741e-06,
+      "loss": 0.3655,
+      "step": 32812
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.980236092220376e-06,
+      "loss": 0.1422,
+      "step": 32813
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.97990025691401e-06,
+      "loss": 0.2907,
+      "step": 32814
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.979564421607644e-06,
+      "loss": 0.382,
+      "step": 32815
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.979228586301278e-06,
+      "loss": 0.26,
+      "step": 32816
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.978892750994913e-06,
+      "loss": 0.3266,
+      "step": 32817
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.978556915688546e-06,
+      "loss": 0.2285,
+      "step": 32818
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.97822108038218e-06,
+      "loss": 0.3272,
+      "step": 32819
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.977885245075815e-06,
+      "loss": 0.2227,
+      "step": 32820
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.97754940976945e-06,
+      "loss": 0.3202,
+      "step": 32821
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.977213574463084e-06,
+      "loss": 0.3715,
+      "step": 32822
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.976877739156719e-06,
+      "loss": 0.196,
+      "step": 32823
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.976541903850352e-06,
+      "loss": 0.272,
+      "step": 32824
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.976206068543987e-06,
+      "loss": 0.2073,
+      "step": 32825
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.975870233237621e-06,
+      "loss": 0.2302,
+      "step": 32826
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.975534397931254e-06,
+      "loss": 0.1979,
+      "step": 32827
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.975198562624889e-06,
+      "loss": 0.1847,
+      "step": 32828
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.974862727318524e-06,
+      "loss": 0.2317,
+      "step": 32829
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.974526892012158e-06,
+      "loss": 0.2196,
+      "step": 32830
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.974191056705793e-06,
+      "loss": 0.1696,
+      "step": 32831
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.973855221399428e-06,
+      "loss": 0.2271,
+      "step": 32832
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.97351938609306e-06,
+      "loss": 0.3218,
+      "step": 32833
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.973183550786695e-06,
+      "loss": 0.1279,
+      "step": 32834
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.97284771548033e-06,
+      "loss": 0.2214,
+      "step": 32835
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.972511880173963e-06,
+      "loss": 0.2128,
+      "step": 32836
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.972176044867597e-06,
+      "loss": 0.3804,
+      "step": 32837
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.971840209561232e-06,
+      "loss": 0.239,
+      "step": 32838
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.971504374254865e-06,
+      "loss": 0.4165,
+      "step": 32839
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.9711685389485e-06,
+      "loss": 0.6048,
+      "step": 32840
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.970832703642134e-06,
+      "loss": 0.2277,
+      "step": 32841
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.970496868335769e-06,
+      "loss": 0.3552,
+      "step": 32842
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.970161033029404e-06,
+      "loss": 0.3561,
+      "step": 32843
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.969825197723038e-06,
+      "loss": 0.3663,
+      "step": 32844
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.969489362416671e-06,
+      "loss": 0.2926,
+      "step": 32845
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.969153527110306e-06,
+      "loss": 0.2456,
+      "step": 32846
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.96881769180394e-06,
+      "loss": 0.2932,
+      "step": 32847
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.968481856497573e-06,
+      "loss": 0.2092,
+      "step": 32848
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.968146021191208e-06,
+      "loss": 0.2088,
+      "step": 32849
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.967810185884843e-06,
+      "loss": 0.2985,
+      "step": 32850
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.967474350578477e-06,
+      "loss": 0.3237,
+      "step": 32851
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.967138515272112e-06,
+      "loss": 0.5921,
+      "step": 32852
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 8.966802679965747e-06,
+      "loss": 0.3156,
+      "step": 32853
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.96646684465938e-06,
+      "loss": 0.2995,
+      "step": 32854
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.966131009353014e-06,
+      "loss": 0.2503,
+      "step": 32855
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.965795174046649e-06,
+      "loss": 0.2669,
+      "step": 32856
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.965459338740282e-06,
+      "loss": 0.2026,
+      "step": 32857
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.965123503433917e-06,
+      "loss": 0.2359,
+      "step": 32858
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.964787668127551e-06,
+      "loss": 0.5814,
+      "step": 32859
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.964451832821184e-06,
+      "loss": 0.182,
+      "step": 32860
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.964115997514819e-06,
+      "loss": 0.3678,
+      "step": 32861
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.963780162208453e-06,
+      "loss": 0.2803,
+      "step": 32862
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.963444326902088e-06,
+      "loss": 0.1848,
+      "step": 32863
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.963108491595723e-06,
+      "loss": 0.247,
+      "step": 32864
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.962772656289357e-06,
+      "loss": 0.3487,
+      "step": 32865
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.96243682098299e-06,
+      "loss": 0.2709,
+      "step": 32866
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.962100985676625e-06,
+      "loss": 0.1534,
+      "step": 32867
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.96176515037026e-06,
+      "loss": 0.2709,
+      "step": 32868
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.961429315063893e-06,
+      "loss": 0.3916,
+      "step": 32869
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.961093479757527e-06,
+      "loss": 0.3646,
+      "step": 32870
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.960757644451162e-06,
+      "loss": 0.3569,
+      "step": 32871
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.960421809144796e-06,
+      "loss": 0.3036,
+      "step": 32872
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.960085973838431e-06,
+      "loss": 0.2595,
+      "step": 32873
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.959750138532066e-06,
+      "loss": 0.3718,
+      "step": 32874
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.959414303225699e-06,
+      "loss": 0.1707,
+      "step": 32875
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.959078467919333e-06,
+      "loss": 0.2343,
+      "step": 32876
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.958742632612968e-06,
+      "loss": 0.3809,
+      "step": 32877
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.958406797306601e-06,
+      "loss": 0.3156,
+      "step": 32878
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.958070962000236e-06,
+      "loss": 0.2847,
+      "step": 32879
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.95773512669387e-06,
+      "loss": 0.6309,
+      "step": 32880
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.957399291387503e-06,
+      "loss": 0.1664,
+      "step": 32881
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.957063456081138e-06,
+      "loss": 0.3762,
+      "step": 32882
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.956727620774773e-06,
+      "loss": 0.4553,
+      "step": 32883
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.956391785468407e-06,
+      "loss": 0.3486,
+      "step": 32884
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.956055950162042e-06,
+      "loss": 0.1941,
+      "step": 32885
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.955720114855676e-06,
+      "loss": 0.3404,
+      "step": 32886
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.95538427954931e-06,
+      "loss": 0.3814,
+      "step": 32887
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.955048444242944e-06,
+      "loss": 0.2941,
+      "step": 32888
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.954712608936579e-06,
+      "loss": 0.406,
+      "step": 32889
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.954376773630212e-06,
+      "loss": 0.3457,
+      "step": 32890
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.954040938323846e-06,
+      "loss": 0.3572,
+      "step": 32891
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.953705103017481e-06,
+      "loss": 0.368,
+      "step": 32892
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.953369267711116e-06,
+      "loss": 0.225,
+      "step": 32893
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.95303343240475e-06,
+      "loss": 0.2353,
+      "step": 32894
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.952697597098383e-06,
+      "loss": 0.2311,
+      "step": 32895
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.952361761792018e-06,
+      "loss": 0.2815,
+      "step": 32896
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.952025926485653e-06,
+      "loss": 0.3458,
+      "step": 32897
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.951690091179285e-06,
+      "loss": 0.2987,
+      "step": 32898
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.95135425587292e-06,
+      "loss": 0.1594,
+      "step": 32899
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.951018420566555e-06,
+      "loss": 0.3576,
+      "step": 32900
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.95068258526019e-06,
+      "loss": 0.4457,
+      "step": 32901
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.950346749953824e-06,
+      "loss": 0.2934,
+      "step": 32902
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.950010914647457e-06,
+      "loss": 0.2296,
+      "step": 32903
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.949675079341092e-06,
+      "loss": 0.1349,
+      "step": 32904
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.949339244034726e-06,
+      "loss": 0.1451,
+      "step": 32905
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.949003408728361e-06,
+      "loss": 0.2414,
+      "step": 32906
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.948667573421994e-06,
+      "loss": 0.2414,
+      "step": 32907
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.948331738115629e-06,
+      "loss": 0.2962,
+      "step": 32908
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.947995902809263e-06,
+      "loss": 0.4348,
+      "step": 32909
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.947660067502896e-06,
+      "loss": 0.27,
+      "step": 32910
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.94732423219653e-06,
+      "loss": 0.2018,
+      "step": 32911
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.946988396890165e-06,
+      "loss": 0.2962,
+      "step": 32912
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.9466525615838e-06,
+      "loss": 0.3635,
+      "step": 32913
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.946316726277435e-06,
+      "loss": 0.2147,
+      "step": 32914
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.94598089097107e-06,
+      "loss": 0.3481,
+      "step": 32915
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.945645055664702e-06,
+      "loss": 0.2463,
+      "step": 32916
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.945309220358337e-06,
+      "loss": 0.1968,
+      "step": 32917
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.944973385051972e-06,
+      "loss": 0.2673,
+      "step": 32918
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.944637549745605e-06,
+      "loss": 0.1522,
+      "step": 32919
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.94430171443924e-06,
+      "loss": 0.3603,
+      "step": 32920
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.943965879132874e-06,
+      "loss": 0.2749,
+      "step": 32921
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.943630043826509e-06,
+      "loss": 0.2619,
+      "step": 32922
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.943294208520143e-06,
+      "loss": 0.226,
+      "step": 32923
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.942958373213778e-06,
+      "loss": 0.3076,
+      "step": 32924
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.94262253790741e-06,
+      "loss": 0.5231,
+      "step": 32925
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.942286702601045e-06,
+      "loss": 0.4135,
+      "step": 32926
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.94195086729468e-06,
+      "loss": 0.3178,
+      "step": 32927
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.941615031988313e-06,
+      "loss": 0.138,
+      "step": 32928
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.941279196681948e-06,
+      "loss": 0.2401,
+      "step": 32929
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.940943361375582e-06,
+      "loss": 0.2398,
+      "step": 32930
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.940607526069215e-06,
+      "loss": 0.3558,
+      "step": 32931
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.94027169076285e-06,
+      "loss": 0.325,
+      "step": 32932
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.939935855456485e-06,
+      "loss": 0.1856,
+      "step": 32933
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.93960002015012e-06,
+      "loss": 0.2815,
+      "step": 32934
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.939264184843754e-06,
+      "loss": 0.3017,
+      "step": 32935
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.938928349537389e-06,
+      "loss": 0.2149,
+      "step": 32936
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.938592514231021e-06,
+      "loss": 0.3093,
+      "step": 32937
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.938256678924656e-06,
+      "loss": 0.4501,
+      "step": 32938
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.93792084361829e-06,
+      "loss": 0.3505,
+      "step": 32939
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.937585008311924e-06,
+      "loss": 0.3575,
+      "step": 32940
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.937249173005558e-06,
+      "loss": 0.4414,
+      "step": 32941
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.936913337699193e-06,
+      "loss": 0.3117,
+      "step": 32942
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.936577502392828e-06,
+      "loss": 0.3716,
+      "step": 32943
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.936241667086462e-06,
+      "loss": 0.2931,
+      "step": 32944
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.935905831780097e-06,
+      "loss": 0.2116,
+      "step": 32945
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.93556999647373e-06,
+      "loss": 0.297,
+      "step": 32946
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.935234161167365e-06,
+      "loss": 0.3918,
+      "step": 32947
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.934898325861e-06,
+      "loss": 0.2528,
+      "step": 32948
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.934562490554632e-06,
+      "loss": 0.2097,
+      "step": 32949
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.934226655248267e-06,
+      "loss": 0.2909,
+      "step": 32950
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.933890819941901e-06,
+      "loss": 0.2528,
+      "step": 32951
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.933554984635534e-06,
+      "loss": 0.525,
+      "step": 32952
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.933219149329169e-06,
+      "loss": 0.2439,
+      "step": 32953
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.932883314022804e-06,
+      "loss": 0.437,
+      "step": 32954
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.932547478716438e-06,
+      "loss": 0.2483,
+      "step": 32955
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.932211643410073e-06,
+      "loss": 0.3574,
+      "step": 32956
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.931875808103708e-06,
+      "loss": 0.3218,
+      "step": 32957
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.93153997279734e-06,
+      "loss": 0.2711,
+      "step": 32958
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.931204137490975e-06,
+      "loss": 0.2907,
+      "step": 32959
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.93086830218461e-06,
+      "loss": 0.2842,
+      "step": 32960
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.930532466878243e-06,
+      "loss": 0.2529,
+      "step": 32961
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.930196631571877e-06,
+      "loss": 0.2349,
+      "step": 32962
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.929860796265512e-06,
+      "loss": 0.3063,
+      "step": 32963
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.929524960959147e-06,
+      "loss": 0.2225,
+      "step": 32964
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.929189125652781e-06,
+      "loss": 0.5098,
+      "step": 32965
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.928853290346416e-06,
+      "loss": 0.2658,
+      "step": 32966
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.928517455040049e-06,
+      "loss": 0.3138,
+      "step": 32967
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.928181619733684e-06,
+      "loss": 0.3066,
+      "step": 32968
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.927845784427318e-06,
+      "loss": 0.2252,
+      "step": 32969
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.927509949120951e-06,
+      "loss": 0.3133,
+      "step": 32970
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.927174113814586e-06,
+      "loss": 0.3285,
+      "step": 32971
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.92683827850822e-06,
+      "loss": 0.2958,
+      "step": 32972
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.926502443201854e-06,
+      "loss": 0.1602,
+      "step": 32973
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.926166607895488e-06,
+      "loss": 0.2845,
+      "step": 32974
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.925830772589123e-06,
+      "loss": 0.3241,
+      "step": 32975
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.925494937282757e-06,
+      "loss": 0.3186,
+      "step": 32976
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.925159101976392e-06,
+      "loss": 0.3341,
+      "step": 32977
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.924823266670027e-06,
+      "loss": 0.329,
+      "step": 32978
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.92448743136366e-06,
+      "loss": 0.2105,
+      "step": 32979
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.924151596057294e-06,
+      "loss": 0.329,
+      "step": 32980
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.923815760750927e-06,
+      "loss": 0.2232,
+      "step": 32981
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.923479925444562e-06,
+      "loss": 0.2443,
+      "step": 32982
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.923144090138197e-06,
+      "loss": 0.2609,
+      "step": 32983
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.922808254831831e-06,
+      "loss": 0.4001,
+      "step": 32984
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.922472419525466e-06,
+      "loss": 0.3415,
+      "step": 32985
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.9221365842191e-06,
+      "loss": 0.3381,
+      "step": 32986
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.921800748912734e-06,
+      "loss": 0.137,
+      "step": 32987
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.921464913606368e-06,
+      "loss": 0.4192,
+      "step": 32988
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.921129078300003e-06,
+      "loss": 0.3239,
+      "step": 32989
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.920793242993636e-06,
+      "loss": 0.2311,
+      "step": 32990
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.92045740768727e-06,
+      "loss": 0.2882,
+      "step": 32991
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.920121572380905e-06,
+      "loss": 0.2732,
+      "step": 32992
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.91978573707454e-06,
+      "loss": 0.4217,
+      "step": 32993
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.919449901768173e-06,
+      "loss": 0.2685,
+      "step": 32994
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.919114066461807e-06,
+      "loss": 0.3765,
+      "step": 32995
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.918778231155442e-06,
+      "loss": 0.3296,
+      "step": 32996
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.918442395849077e-06,
+      "loss": 0.462,
+      "step": 32997
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.918106560542711e-06,
+      "loss": 0.1945,
+      "step": 32998
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.917770725236344e-06,
+      "loss": 0.2176,
+      "step": 32999
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.917434889929979e-06,
+      "loss": 0.4845,
+      "step": 33000
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.917099054623613e-06,
+      "loss": 0.2074,
+      "step": 33001
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.916763219317246e-06,
+      "loss": 0.1857,
+      "step": 33002
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.916427384010881e-06,
+      "loss": 0.3732,
+      "step": 33003
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.916091548704516e-06,
+      "loss": 0.1543,
+      "step": 33004
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.91575571339815e-06,
+      "loss": 0.2264,
+      "step": 33005
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.915419878091785e-06,
+      "loss": 0.2488,
+      "step": 33006
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.91508404278542e-06,
+      "loss": 0.1985,
+      "step": 33007
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.914748207479053e-06,
+      "loss": 0.2079,
+      "step": 33008
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.914412372172687e-06,
+      "loss": 0.5029,
+      "step": 33009
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.914076536866322e-06,
+      "loss": 0.2125,
+      "step": 33010
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.913740701559955e-06,
+      "loss": 0.2057,
+      "step": 33011
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.91340486625359e-06,
+      "loss": 0.4223,
+      "step": 33012
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.913069030947224e-06,
+      "loss": 0.2465,
+      "step": 33013
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.912733195640859e-06,
+      "loss": 0.2736,
+      "step": 33014
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.912397360334493e-06,
+      "loss": 0.3293,
+      "step": 33015
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.912061525028126e-06,
+      "loss": 0.193,
+      "step": 33016
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.911725689721761e-06,
+      "loss": 0.297,
+      "step": 33017
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.911389854415396e-06,
+      "loss": 0.3515,
+      "step": 33018
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.91105401910903e-06,
+      "loss": 0.2686,
+      "step": 33019
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.910718183802663e-06,
+      "loss": 0.207,
+      "step": 33020
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.910382348496298e-06,
+      "loss": 0.2421,
+      "step": 33021
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.910046513189933e-06,
+      "loss": 0.3784,
+      "step": 33022
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.909710677883566e-06,
+      "loss": 0.4976,
+      "step": 33023
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.9093748425772e-06,
+      "loss": 0.2234,
+      "step": 33024
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.909039007270835e-06,
+      "loss": 0.2597,
+      "step": 33025
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.90870317196447e-06,
+      "loss": 0.2857,
+      "step": 33026
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.908367336658104e-06,
+      "loss": 0.1961,
+      "step": 33027
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.908031501351739e-06,
+      "loss": 0.2375,
+      "step": 33028
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.907695666045372e-06,
+      "loss": 0.2883,
+      "step": 33029
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.907359830739006e-06,
+      "loss": 0.2604,
+      "step": 33030
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.907023995432641e-06,
+      "loss": 0.3271,
+      "step": 33031
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.906688160126274e-06,
+      "loss": 0.4119,
+      "step": 33032
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.906352324819909e-06,
+      "loss": 0.3636,
+      "step": 33033
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.906016489513543e-06,
+      "loss": 0.2375,
+      "step": 33034
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.905680654207178e-06,
+      "loss": 0.2315,
+      "step": 33035
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.905344818900813e-06,
+      "loss": 0.2192,
+      "step": 33036
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.905008983594447e-06,
+      "loss": 0.2124,
+      "step": 33037
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.90467314828808e-06,
+      "loss": 0.2064,
+      "step": 33038
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.904337312981715e-06,
+      "loss": 0.3376,
+      "step": 33039
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.90400147767535e-06,
+      "loss": 0.3898,
+      "step": 33040
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.903665642368982e-06,
+      "loss": 0.399,
+      "step": 33041
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.903329807062617e-06,
+      "loss": 0.2403,
+      "step": 33042
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.902993971756252e-06,
+      "loss": 0.2778,
+      "step": 33043
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.902658136449885e-06,
+      "loss": 0.2406,
+      "step": 33044
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.90232230114352e-06,
+      "loss": 0.3013,
+      "step": 33045
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.901986465837154e-06,
+      "loss": 0.1986,
+      "step": 33046
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.901650630530789e-06,
+      "loss": 0.3101,
+      "step": 33047
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.901314795224423e-06,
+      "loss": 0.2336,
+      "step": 33048
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.900978959918058e-06,
+      "loss": 0.3662,
+      "step": 33049
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.900643124611691e-06,
+      "loss": 0.2491,
+      "step": 33050
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 8.900307289305326e-06,
+      "loss": 0.3473,
+      "step": 33051
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.89997145399896e-06,
+      "loss": 0.2742,
+      "step": 33052
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.899635618692593e-06,
+      "loss": 0.2005,
+      "step": 33053
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.899299783386228e-06,
+      "loss": 0.1817,
+      "step": 33054
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.898963948079862e-06,
+      "loss": 0.3343,
+      "step": 33055
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.898628112773497e-06,
+      "loss": 0.301,
+      "step": 33056
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.898292277467132e-06,
+      "loss": 0.1971,
+      "step": 33057
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.897956442160766e-06,
+      "loss": 0.37,
+      "step": 33058
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.8976206068544e-06,
+      "loss": 0.3101,
+      "step": 33059
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.897284771548034e-06,
+      "loss": 0.2829,
+      "step": 33060
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.896948936241669e-06,
+      "loss": 0.238,
+      "step": 33061
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.896613100935302e-06,
+      "loss": 0.4175,
+      "step": 33062
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.896277265628936e-06,
+      "loss": 0.5066,
+      "step": 33063
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.895941430322571e-06,
+      "loss": 0.565,
+      "step": 33064
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.895605595016204e-06,
+      "loss": 0.2232,
+      "step": 33065
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.895269759709838e-06,
+      "loss": 0.2979,
+      "step": 33066
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.894933924403473e-06,
+      "loss": 0.3848,
+      "step": 33067
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.894598089097108e-06,
+      "loss": 0.1406,
+      "step": 33068
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.894262253790742e-06,
+      "loss": 0.2796,
+      "step": 33069
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.893926418484375e-06,
+      "loss": 0.2761,
+      "step": 33070
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.89359058317801e-06,
+      "loss": 0.3508,
+      "step": 33071
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.893254747871645e-06,
+      "loss": 0.3908,
+      "step": 33072
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.892918912565278e-06,
+      "loss": 0.2303,
+      "step": 33073
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.892583077258912e-06,
+      "loss": 0.4059,
+      "step": 33074
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.892247241952547e-06,
+      "loss": 0.271,
+      "step": 33075
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.891911406646182e-06,
+      "loss": 0.2488,
+      "step": 33076
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.891575571339816e-06,
+      "loss": 0.3401,
+      "step": 33077
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.891239736033451e-06,
+      "loss": 0.2896,
+      "step": 33078
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.890903900727084e-06,
+      "loss": 0.2435,
+      "step": 33079
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.890568065420718e-06,
+      "loss": 0.2797,
+      "step": 33080
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.890232230114353e-06,
+      "loss": 0.2382,
+      "step": 33081
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.889896394807986e-06,
+      "loss": 0.2168,
+      "step": 33082
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.88956055950162e-06,
+      "loss": 0.2154,
+      "step": 33083
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.889224724195255e-06,
+      "loss": 0.3425,
+      "step": 33084
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.888888888888888e-06,
+      "loss": 0.2637,
+      "step": 33085
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.888553053582523e-06,
+      "loss": 0.2767,
+      "step": 33086
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.888217218276158e-06,
+      "loss": 0.3152,
+      "step": 33087
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.887881382969792e-06,
+      "loss": 0.3508,
+      "step": 33088
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.887545547663427e-06,
+      "loss": 0.2826,
+      "step": 33089
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.887209712357062e-06,
+      "loss": 0.2918,
+      "step": 33090
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.886873877050694e-06,
+      "loss": 0.1618,
+      "step": 33091
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.886538041744329e-06,
+      "loss": 0.3117,
+      "step": 33092
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.886202206437964e-06,
+      "loss": 0.2564,
+      "step": 33093
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.885866371131597e-06,
+      "loss": 0.4519,
+      "step": 33094
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.885530535825231e-06,
+      "loss": 0.434,
+      "step": 33095
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.885194700518866e-06,
+      "loss": 0.2191,
+      "step": 33096
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.8848588652125e-06,
+      "loss": 0.2686,
+      "step": 33097
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.884523029906135e-06,
+      "loss": 0.3054,
+      "step": 33098
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.88418719459977e-06,
+      "loss": 0.1952,
+      "step": 33099
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.883851359293403e-06,
+      "loss": 0.1669,
+      "step": 33100
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.883515523987038e-06,
+      "loss": 0.3706,
+      "step": 33101
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.883179688680672e-06,
+      "loss": 0.3545,
+      "step": 33102
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.882843853374305e-06,
+      "loss": 0.317,
+      "step": 33103
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.88250801806794e-06,
+      "loss": 0.2792,
+      "step": 33104
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.882172182761574e-06,
+      "loss": 0.3195,
+      "step": 33105
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.881836347455209e-06,
+      "loss": 0.2007,
+      "step": 33106
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.881500512148842e-06,
+      "loss": 0.436,
+      "step": 33107
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.881164676842477e-06,
+      "loss": 0.3508,
+      "step": 33108
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.880828841536111e-06,
+      "loss": 0.3354,
+      "step": 33109
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.880493006229746e-06,
+      "loss": 0.2205,
+      "step": 33110
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.88015717092338e-06,
+      "loss": 0.4676,
+      "step": 33111
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.879821335617014e-06,
+      "loss": 0.2291,
+      "step": 33112
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.879485500310648e-06,
+      "loss": 0.25,
+      "step": 33113
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.879149665004283e-06,
+      "loss": 0.3589,
+      "step": 33114
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.878813829697916e-06,
+      "loss": 0.1359,
+      "step": 33115
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.87847799439155e-06,
+      "loss": 0.2458,
+      "step": 33116
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.878142159085185e-06,
+      "loss": 0.3223,
+      "step": 33117
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.87780632377882e-06,
+      "loss": 0.1259,
+      "step": 33118
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.877470488472454e-06,
+      "loss": 0.2878,
+      "step": 33119
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.877134653166089e-06,
+      "loss": 0.3209,
+      "step": 33120
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.876798817859722e-06,
+      "loss": 0.4766,
+      "step": 33121
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.876462982553357e-06,
+      "loss": 0.3936,
+      "step": 33122
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.876127147246991e-06,
+      "loss": 0.4138,
+      "step": 33123
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.875791311940624e-06,
+      "loss": 0.3928,
+      "step": 33124
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.875455476634259e-06,
+      "loss": 0.4556,
+      "step": 33125
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.875119641327894e-06,
+      "loss": 0.3897,
+      "step": 33126
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.874783806021528e-06,
+      "loss": 0.1926,
+      "step": 33127
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.874447970715163e-06,
+      "loss": 0.386,
+      "step": 33128
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.874112135408796e-06,
+      "loss": 0.3749,
+      "step": 33129
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.87377630010243e-06,
+      "loss": 0.3342,
+      "step": 33130
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.873440464796065e-06,
+      "loss": 0.2879,
+      "step": 33131
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.8731046294897e-06,
+      "loss": 0.3075,
+      "step": 33132
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.872768794183333e-06,
+      "loss": 0.3155,
+      "step": 33133
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.872432958876967e-06,
+      "loss": 0.1856,
+      "step": 33134
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.872097123570602e-06,
+      "loss": 0.2219,
+      "step": 33135
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.871761288264235e-06,
+      "loss": 0.1887,
+      "step": 33136
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.87142545295787e-06,
+      "loss": 0.2403,
+      "step": 33137
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.871089617651504e-06,
+      "loss": 0.193,
+      "step": 33138
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.870753782345139e-06,
+      "loss": 0.4646,
+      "step": 33139
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.870417947038774e-06,
+      "loss": 0.3282,
+      "step": 33140
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.870082111732408e-06,
+      "loss": 0.3969,
+      "step": 33141
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.869746276426041e-06,
+      "loss": 0.2891,
+      "step": 33142
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.869410441119676e-06,
+      "loss": 0.3959,
+      "step": 33143
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.86907460581331e-06,
+      "loss": 0.4106,
+      "step": 33144
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.868738770506943e-06,
+      "loss": 0.301,
+      "step": 33145
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.868402935200578e-06,
+      "loss": 0.2709,
+      "step": 33146
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.868067099894213e-06,
+      "loss": 0.2133,
+      "step": 33147
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.867731264587847e-06,
+      "loss": 0.3029,
+      "step": 33148
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.867395429281482e-06,
+      "loss": 0.3818,
+      "step": 33149
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.867059593975117e-06,
+      "loss": 0.4394,
+      "step": 33150
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.86672375866875e-06,
+      "loss": 0.2542,
+      "step": 33151
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.866387923362384e-06,
+      "loss": 0.2879,
+      "step": 33152
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.866052088056019e-06,
+      "loss": 0.2481,
+      "step": 33153
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.865716252749652e-06,
+      "loss": 0.3425,
+      "step": 33154
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.865380417443287e-06,
+      "loss": 0.4542,
+      "step": 33155
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.865044582136921e-06,
+      "loss": 0.3384,
+      "step": 33156
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.864708746830554e-06,
+      "loss": 0.2486,
+      "step": 33157
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.864372911524189e-06,
+      "loss": 0.4692,
+      "step": 33158
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.864037076217823e-06,
+      "loss": 0.3856,
+      "step": 33159
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.863701240911458e-06,
+      "loss": 0.3314,
+      "step": 33160
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.863365405605093e-06,
+      "loss": 0.1687,
+      "step": 33161
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.863029570298726e-06,
+      "loss": 0.3237,
+      "step": 33162
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.86269373499236e-06,
+      "loss": 0.2048,
+      "step": 33163
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.862357899685995e-06,
+      "loss": 0.242,
+      "step": 33164
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.862022064379628e-06,
+      "loss": 0.2487,
+      "step": 33165
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.861686229073263e-06,
+      "loss": 0.2313,
+      "step": 33166
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.861350393766897e-06,
+      "loss": 0.1875,
+      "step": 33167
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.861014558460532e-06,
+      "loss": 0.246,
+      "step": 33168
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.860678723154166e-06,
+      "loss": 0.177,
+      "step": 33169
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.860342887847801e-06,
+      "loss": 0.2404,
+      "step": 33170
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.860007052541434e-06,
+      "loss": 0.519,
+      "step": 33171
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.859671217235069e-06,
+      "loss": 0.2624,
+      "step": 33172
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.859335381928703e-06,
+      "loss": 0.2902,
+      "step": 33173
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.858999546622336e-06,
+      "loss": 0.3548,
+      "step": 33174
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.858663711315971e-06,
+      "loss": 0.2496,
+      "step": 33175
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.858327876009606e-06,
+      "loss": 0.3396,
+      "step": 33176
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.857992040703239e-06,
+      "loss": 0.2123,
+      "step": 33177
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.857656205396873e-06,
+      "loss": 0.2627,
+      "step": 33178
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.857320370090508e-06,
+      "loss": 0.3007,
+      "step": 33179
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.856984534784143e-06,
+      "loss": 0.3409,
+      "step": 33180
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.856648699477777e-06,
+      "loss": 0.2602,
+      "step": 33181
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.856312864171412e-06,
+      "loss": 0.2397,
+      "step": 33182
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.855977028865045e-06,
+      "loss": 0.3005,
+      "step": 33183
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.85564119355868e-06,
+      "loss": 0.1263,
+      "step": 33184
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.855305358252314e-06,
+      "loss": 0.2803,
+      "step": 33185
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.854969522945947e-06,
+      "loss": 0.2761,
+      "step": 33186
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.854633687639582e-06,
+      "loss": 0.2281,
+      "step": 33187
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.854297852333216e-06,
+      "loss": 0.1982,
+      "step": 33188
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.853962017026851e-06,
+      "loss": 0.2881,
+      "step": 33189
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.853626181720486e-06,
+      "loss": 0.2988,
+      "step": 33190
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.85329034641412e-06,
+      "loss": 0.3979,
+      "step": 33191
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.852954511107753e-06,
+      "loss": 0.1199,
+      "step": 33192
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.852618675801388e-06,
+      "loss": 0.3552,
+      "step": 33193
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.852282840495023e-06,
+      "loss": 0.2391,
+      "step": 33194
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.851947005188655e-06,
+      "loss": 0.3479,
+      "step": 33195
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.85161116988229e-06,
+      "loss": 0.324,
+      "step": 33196
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.851275334575925e-06,
+      "loss": 0.2501,
+      "step": 33197
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.850939499269558e-06,
+      "loss": 0.3299,
+      "step": 33198
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.850603663963192e-06,
+      "loss": 0.1712,
+      "step": 33199
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.850267828656827e-06,
+      "loss": 0.6795,
+      "step": 33200
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.849931993350462e-06,
+      "loss": 0.3645,
+      "step": 33201
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.849596158044096e-06,
+      "loss": 0.4902,
+      "step": 33202
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.849260322737731e-06,
+      "loss": 0.2772,
+      "step": 33203
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.848924487431364e-06,
+      "loss": 0.336,
+      "step": 33204
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.848588652124999e-06,
+      "loss": 0.3171,
+      "step": 33205
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.848252816818633e-06,
+      "loss": 0.4332,
+      "step": 33206
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.847916981512266e-06,
+      "loss": 0.3604,
+      "step": 33207
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.8475811462059e-06,
+      "loss": 0.3276,
+      "step": 33208
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.847245310899535e-06,
+      "loss": 0.4494,
+      "step": 33209
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.84690947559317e-06,
+      "loss": 0.3655,
+      "step": 33210
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.846573640286805e-06,
+      "loss": 0.3225,
+      "step": 33211
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.84623780498044e-06,
+      "loss": 0.3167,
+      "step": 33212
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.845901969674072e-06,
+      "loss": 0.3285,
+      "step": 33213
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.845566134367707e-06,
+      "loss": 0.3073,
+      "step": 33214
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.845230299061342e-06,
+      "loss": 0.3921,
+      "step": 33215
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.844894463754975e-06,
+      "loss": 0.4959,
+      "step": 33216
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.84455862844861e-06,
+      "loss": 0.2179,
+      "step": 33217
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.844222793142244e-06,
+      "loss": 0.3778,
+      "step": 33218
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.843886957835879e-06,
+      "loss": 0.2519,
+      "step": 33219
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.843551122529511e-06,
+      "loss": 0.4663,
+      "step": 33220
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.843215287223146e-06,
+      "loss": 0.3234,
+      "step": 33221
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.84287945191678e-06,
+      "loss": 0.3106,
+      "step": 33222
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.842543616610415e-06,
+      "loss": 0.3595,
+      "step": 33223
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.84220778130405e-06,
+      "loss": 0.1901,
+      "step": 33224
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.841871945997683e-06,
+      "loss": 0.2254,
+      "step": 33225
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.841536110691318e-06,
+      "loss": 0.2785,
+      "step": 33226
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.841200275384952e-06,
+      "loss": 0.32,
+      "step": 33227
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.840864440078585e-06,
+      "loss": 0.3247,
+      "step": 33228
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.84052860477222e-06,
+      "loss": 0.3791,
+      "step": 33229
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.840192769465855e-06,
+      "loss": 0.2557,
+      "step": 33230
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.83985693415949e-06,
+      "loss": 0.3055,
+      "step": 33231
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.839521098853124e-06,
+      "loss": 0.2622,
+      "step": 33232
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.839185263546759e-06,
+      "loss": 0.2266,
+      "step": 33233
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.838849428240391e-06,
+      "loss": 0.2969,
+      "step": 33234
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.838513592934026e-06,
+      "loss": 0.3597,
+      "step": 33235
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.83817775762766e-06,
+      "loss": 0.2478,
+      "step": 33236
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.837841922321294e-06,
+      "loss": 0.3597,
+      "step": 33237
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.837506087014928e-06,
+      "loss": 0.3385,
+      "step": 33238
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.837170251708563e-06,
+      "loss": 0.3236,
+      "step": 33239
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.836834416402198e-06,
+      "loss": 0.3871,
+      "step": 33240
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.836498581095832e-06,
+      "loss": 0.3057,
+      "step": 33241
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.836162745789465e-06,
+      "loss": 0.376,
+      "step": 33242
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.8358269104831e-06,
+      "loss": 0.437,
+      "step": 33243
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.835491075176735e-06,
+      "loss": 0.3359,
+      "step": 33244
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.83515523987037e-06,
+      "loss": 0.3147,
+      "step": 33245
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.834819404564002e-06,
+      "loss": 0.2243,
+      "step": 33246
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.834483569257637e-06,
+      "loss": 0.241,
+      "step": 33247
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.834147733951271e-06,
+      "loss": 0.2125,
+      "step": 33248
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.833811898644904e-06,
+      "loss": 0.3083,
+      "step": 33249
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 8.833476063338539e-06,
+      "loss": 0.2071,
+      "step": 33250
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.833140228032174e-06,
+      "loss": 0.1897,
+      "step": 33251
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.832804392725808e-06,
+      "loss": 0.2779,
+      "step": 33252
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.832468557419443e-06,
+      "loss": 0.3474,
+      "step": 33253
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.832132722113076e-06,
+      "loss": 0.2766,
+      "step": 33254
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.83179688680671e-06,
+      "loss": 0.207,
+      "step": 33255
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.831461051500345e-06,
+      "loss": 0.4061,
+      "step": 33256
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.831125216193978e-06,
+      "loss": 0.3949,
+      "step": 33257
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.830789380887613e-06,
+      "loss": 0.2235,
+      "step": 33258
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.830453545581247e-06,
+      "loss": 0.2231,
+      "step": 33259
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.830117710274882e-06,
+      "loss": 0.2853,
+      "step": 33260
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.829781874968517e-06,
+      "loss": 0.405,
+      "step": 33261
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.829446039662151e-06,
+      "loss": 0.2844,
+      "step": 33262
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.829110204355784e-06,
+      "loss": 0.3412,
+      "step": 33263
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.828774369049419e-06,
+      "loss": 0.2354,
+      "step": 33264
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.828438533743054e-06,
+      "loss": 0.246,
+      "step": 33265
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.828102698436687e-06,
+      "loss": 0.1362,
+      "step": 33266
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.827766863130321e-06,
+      "loss": 0.2654,
+      "step": 33267
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.827431027823956e-06,
+      "loss": 0.2516,
+      "step": 33268
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.827095192517589e-06,
+      "loss": 0.3547,
+      "step": 33269
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.826759357211224e-06,
+      "loss": 0.3964,
+      "step": 33270
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.826423521904858e-06,
+      "loss": 0.4608,
+      "step": 33271
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.826087686598493e-06,
+      "loss": 0.3947,
+      "step": 33272
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.825751851292127e-06,
+      "loss": 0.4214,
+      "step": 33273
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.825416015985762e-06,
+      "loss": 0.3546,
+      "step": 33274
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.825080180679395e-06,
+      "loss": 0.2733,
+      "step": 33275
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.82474434537303e-06,
+      "loss": 0.1517,
+      "step": 33276
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.824408510066664e-06,
+      "loss": 0.2376,
+      "step": 33277
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.824072674760297e-06,
+      "loss": 0.3911,
+      "step": 33278
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.823736839453932e-06,
+      "loss": 0.3079,
+      "step": 33279
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.823401004147567e-06,
+      "loss": 0.2497,
+      "step": 33280
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.823065168841201e-06,
+      "loss": 0.2757,
+      "step": 33281
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.822729333534836e-06,
+      "loss": 0.351,
+      "step": 33282
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.82239349822847e-06,
+      "loss": 0.2999,
+      "step": 33283
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.822057662922104e-06,
+      "loss": 0.3177,
+      "step": 33284
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.821721827615738e-06,
+      "loss": 0.422,
+      "step": 33285
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.821385992309373e-06,
+      "loss": 0.3267,
+      "step": 33286
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.821050157003006e-06,
+      "loss": 0.3945,
+      "step": 33287
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.82071432169664e-06,
+      "loss": 0.2766,
+      "step": 33288
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.820378486390275e-06,
+      "loss": 0.345,
+      "step": 33289
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.820042651083908e-06,
+      "loss": 0.2222,
+      "step": 33290
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.819706815777543e-06,
+      "loss": 0.195,
+      "step": 33291
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.819370980471177e-06,
+      "loss": 0.3736,
+      "step": 33292
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.819035145164812e-06,
+      "loss": 0.3466,
+      "step": 33293
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.818699309858447e-06,
+      "loss": 0.4699,
+      "step": 33294
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.818363474552081e-06,
+      "loss": 0.3462,
+      "step": 33295
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.818027639245714e-06,
+      "loss": 0.2256,
+      "step": 33296
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.817691803939349e-06,
+      "loss": 0.2675,
+      "step": 33297
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.817355968632983e-06,
+      "loss": 0.1623,
+      "step": 33298
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.817020133326616e-06,
+      "loss": 0.2019,
+      "step": 33299
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.816684298020251e-06,
+      "loss": 0.2398,
+      "step": 33300
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.816348462713886e-06,
+      "loss": 0.2503,
+      "step": 33301
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.81601262740752e-06,
+      "loss": 0.3481,
+      "step": 33302
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.815676792101155e-06,
+      "loss": 0.4107,
+      "step": 33303
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.81534095679479e-06,
+      "loss": 0.2547,
+      "step": 33304
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.815005121488423e-06,
+      "loss": 0.2911,
+      "step": 33305
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.814669286182057e-06,
+      "loss": 0.2335,
+      "step": 33306
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.814333450875692e-06,
+      "loss": 0.5576,
+      "step": 33307
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.813997615569325e-06,
+      "loss": 0.195,
+      "step": 33308
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.81366178026296e-06,
+      "loss": 0.4555,
+      "step": 33309
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.813325944956594e-06,
+      "loss": 0.4339,
+      "step": 33310
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.812990109650227e-06,
+      "loss": 0.2072,
+      "step": 33311
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.812654274343862e-06,
+      "loss": 0.2773,
+      "step": 33312
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.812318439037496e-06,
+      "loss": 0.3228,
+      "step": 33313
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.811982603731131e-06,
+      "loss": 0.2079,
+      "step": 33314
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.811646768424766e-06,
+      "loss": 0.2875,
+      "step": 33315
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.8113109331184e-06,
+      "loss": 0.1698,
+      "step": 33316
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.810975097812033e-06,
+      "loss": 0.1978,
+      "step": 33317
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.810639262505668e-06,
+      "loss": 0.3658,
+      "step": 33318
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.810303427199303e-06,
+      "loss": 0.4844,
+      "step": 33319
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.809967591892936e-06,
+      "loss": 0.3081,
+      "step": 33320
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.80963175658657e-06,
+      "loss": 0.2909,
+      "step": 33321
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.809295921280205e-06,
+      "loss": 0.4923,
+      "step": 33322
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.80896008597384e-06,
+      "loss": 0.1722,
+      "step": 33323
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.808624250667474e-06,
+      "loss": 0.3955,
+      "step": 33324
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.808288415361109e-06,
+      "loss": 0.129,
+      "step": 33325
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.807952580054742e-06,
+      "loss": 0.4591,
+      "step": 33326
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.807616744748376e-06,
+      "loss": 0.2498,
+      "step": 33327
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.807280909442011e-06,
+      "loss": 0.3644,
+      "step": 33328
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.806945074135644e-06,
+      "loss": 0.3824,
+      "step": 33329
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.806609238829279e-06,
+      "loss": 0.3936,
+      "step": 33330
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.806273403522913e-06,
+      "loss": 0.2981,
+      "step": 33331
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.805937568216548e-06,
+      "loss": 0.3689,
+      "step": 33332
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.805601732910181e-06,
+      "loss": 0.3217,
+      "step": 33333
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.805265897603816e-06,
+      "loss": 0.3232,
+      "step": 33334
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.80493006229745e-06,
+      "loss": 0.4145,
+      "step": 33335
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.804594226991085e-06,
+      "loss": 0.1768,
+      "step": 33336
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.80425839168472e-06,
+      "loss": 0.2357,
+      "step": 33337
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.803922556378352e-06,
+      "loss": 0.1677,
+      "step": 33338
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.803586721071987e-06,
+      "loss": 0.314,
+      "step": 33339
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.803250885765622e-06,
+      "loss": 0.3127,
+      "step": 33340
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.802915050459255e-06,
+      "loss": 0.325,
+      "step": 33341
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.80257921515289e-06,
+      "loss": 0.1388,
+      "step": 33342
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.802243379846524e-06,
+      "loss": 0.3172,
+      "step": 33343
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.801907544540159e-06,
+      "loss": 0.3251,
+      "step": 33344
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.801571709233793e-06,
+      "loss": 0.3366,
+      "step": 33345
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.801235873927426e-06,
+      "loss": 0.2826,
+      "step": 33346
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.800900038621061e-06,
+      "loss": 0.165,
+      "step": 33347
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.800564203314696e-06,
+      "loss": 0.3825,
+      "step": 33348
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.800228368008328e-06,
+      "loss": 0.1913,
+      "step": 33349
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.799892532701963e-06,
+      "loss": 0.1983,
+      "step": 33350
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.799556697395598e-06,
+      "loss": 0.3197,
+      "step": 33351
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.799220862089232e-06,
+      "loss": 0.4249,
+      "step": 33352
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.798885026782867e-06,
+      "loss": 0.1574,
+      "step": 33353
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.798549191476502e-06,
+      "loss": 0.3756,
+      "step": 33354
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.798213356170135e-06,
+      "loss": 0.4596,
+      "step": 33355
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.79787752086377e-06,
+      "loss": 0.3286,
+      "step": 33356
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.797541685557404e-06,
+      "loss": 0.2791,
+      "step": 33357
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.797205850251037e-06,
+      "loss": 0.2476,
+      "step": 33358
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.796870014944672e-06,
+      "loss": 0.3604,
+      "step": 33359
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.796534179638306e-06,
+      "loss": 0.1564,
+      "step": 33360
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.79619834433194e-06,
+      "loss": 0.2912,
+      "step": 33361
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.795862509025574e-06,
+      "loss": 0.3072,
+      "step": 33362
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.795526673719208e-06,
+      "loss": 0.3295,
+      "step": 33363
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.795190838412843e-06,
+      "loss": 0.4297,
+      "step": 33364
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.794855003106478e-06,
+      "loss": 0.1909,
+      "step": 33365
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.794519167800112e-06,
+      "loss": 0.2467,
+      "step": 33366
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.794183332493745e-06,
+      "loss": 0.5036,
+      "step": 33367
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.79384749718738e-06,
+      "loss": 0.2297,
+      "step": 33368
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.793511661881015e-06,
+      "loss": 0.3921,
+      "step": 33369
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.793175826574648e-06,
+      "loss": 0.2374,
+      "step": 33370
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.792839991268282e-06,
+      "loss": 0.2915,
+      "step": 33371
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.792504155961917e-06,
+      "loss": 0.2995,
+      "step": 33372
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.792168320655552e-06,
+      "loss": 0.2986,
+      "step": 33373
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.791832485349186e-06,
+      "loss": 0.2083,
+      "step": 33374
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.79149665004282e-06,
+      "loss": 0.33,
+      "step": 33375
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.791160814736454e-06,
+      "loss": 0.2173,
+      "step": 33376
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.790824979430088e-06,
+      "loss": 0.3919,
+      "step": 33377
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.790489144123723e-06,
+      "loss": 0.384,
+      "step": 33378
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.790153308817356e-06,
+      "loss": 0.3324,
+      "step": 33379
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.78981747351099e-06,
+      "loss": 0.1651,
+      "step": 33380
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.789481638204625e-06,
+      "loss": 0.2883,
+      "step": 33381
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.789145802898258e-06,
+      "loss": 0.2824,
+      "step": 33382
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.788809967591893e-06,
+      "loss": 0.2302,
+      "step": 33383
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.788474132285528e-06,
+      "loss": 0.4112,
+      "step": 33384
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.788138296979162e-06,
+      "loss": 0.262,
+      "step": 33385
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.787802461672797e-06,
+      "loss": 0.2257,
+      "step": 33386
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.787466626366432e-06,
+      "loss": 0.2463,
+      "step": 33387
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.787130791060064e-06,
+      "loss": 0.3659,
+      "step": 33388
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.786794955753699e-06,
+      "loss": 0.4422,
+      "step": 33389
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.786459120447334e-06,
+      "loss": 0.2977,
+      "step": 33390
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.786123285140967e-06,
+      "loss": 0.2416,
+      "step": 33391
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.785787449834601e-06,
+      "loss": 0.238,
+      "step": 33392
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.785451614528236e-06,
+      "loss": 0.1643,
+      "step": 33393
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.78511577922187e-06,
+      "loss": 0.2698,
+      "step": 33394
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.784779943915505e-06,
+      "loss": 0.4884,
+      "step": 33395
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.78444410860914e-06,
+      "loss": 0.516,
+      "step": 33396
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.784108273302773e-06,
+      "loss": 0.1449,
+      "step": 33397
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.783772437996408e-06,
+      "loss": 0.3465,
+      "step": 33398
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.783436602690042e-06,
+      "loss": 0.1969,
+      "step": 33399
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.783100767383675e-06,
+      "loss": 0.3822,
+      "step": 33400
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.78276493207731e-06,
+      "loss": 0.3823,
+      "step": 33401
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.782429096770944e-06,
+      "loss": 0.4703,
+      "step": 33402
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.782093261464577e-06,
+      "loss": 0.2293,
+      "step": 33403
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.781757426158212e-06,
+      "loss": 0.3399,
+      "step": 33404
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.781421590851847e-06,
+      "loss": 0.3136,
+      "step": 33405
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.781085755545481e-06,
+      "loss": 0.4085,
+      "step": 33406
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.780749920239116e-06,
+      "loss": 0.2785,
+      "step": 33407
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.78041408493275e-06,
+      "loss": 0.253,
+      "step": 33408
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.780078249626384e-06,
+      "loss": 0.4902,
+      "step": 33409
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.779742414320018e-06,
+      "loss": 0.3393,
+      "step": 33410
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.779406579013653e-06,
+      "loss": 0.452,
+      "step": 33411
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.779070743707286e-06,
+      "loss": 0.3286,
+      "step": 33412
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.77873490840092e-06,
+      "loss": 0.2519,
+      "step": 33413
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.778399073094555e-06,
+      "loss": 0.2819,
+      "step": 33414
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.77806323778819e-06,
+      "loss": 0.3331,
+      "step": 33415
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.777727402481824e-06,
+      "loss": 0.3413,
+      "step": 33416
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.777391567175459e-06,
+      "loss": 0.3941,
+      "step": 33417
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.777055731869092e-06,
+      "loss": 0.235,
+      "step": 33418
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.776719896562727e-06,
+      "loss": 0.2438,
+      "step": 33419
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.776384061256361e-06,
+      "loss": 0.3607,
+      "step": 33420
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.776048225949994e-06,
+      "loss": 0.3117,
+      "step": 33421
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.775712390643629e-06,
+      "loss": 0.3003,
+      "step": 33422
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.775376555337264e-06,
+      "loss": 0.2719,
+      "step": 33423
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.775040720030897e-06,
+      "loss": 0.221,
+      "step": 33424
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.774704884724531e-06,
+      "loss": 0.219,
+      "step": 33425
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.774369049418166e-06,
+      "loss": 0.4212,
+      "step": 33426
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.7740332141118e-06,
+      "loss": 0.3103,
+      "step": 33427
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.773697378805435e-06,
+      "loss": 0.1483,
+      "step": 33428
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.77336154349907e-06,
+      "loss": 0.4906,
+      "step": 33429
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.773025708192703e-06,
+      "loss": 0.3082,
+      "step": 33430
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.772689872886337e-06,
+      "loss": 0.238,
+      "step": 33431
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.77235403757997e-06,
+      "loss": 0.3483,
+      "step": 33432
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.772018202273605e-06,
+      "loss": 0.2795,
+      "step": 33433
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.77168236696724e-06,
+      "loss": 0.2706,
+      "step": 33434
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.771346531660874e-06,
+      "loss": 0.404,
+      "step": 33435
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.771010696354509e-06,
+      "loss": 0.3003,
+      "step": 33436
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.770674861048144e-06,
+      "loss": 0.4236,
+      "step": 33437
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.770339025741777e-06,
+      "loss": 0.3409,
+      "step": 33438
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.770003190435411e-06,
+      "loss": 0.281,
+      "step": 33439
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.769667355129046e-06,
+      "loss": 0.2338,
+      "step": 33440
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.769331519822679e-06,
+      "loss": 0.2313,
+      "step": 33441
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.768995684516313e-06,
+      "loss": 0.3437,
+      "step": 33442
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.768659849209948e-06,
+      "loss": 0.3314,
+      "step": 33443
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.768324013903583e-06,
+      "loss": 0.3632,
+      "step": 33444
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.767988178597217e-06,
+      "loss": 0.1535,
+      "step": 33445
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.76765234329085e-06,
+      "loss": 0.2821,
+      "step": 33446
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.767316507984485e-06,
+      "loss": 0.2292,
+      "step": 33447
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 8.76698067267812e-06,
+      "loss": 0.4439,
+      "step": 33448
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.766644837371754e-06,
+      "loss": 0.3062,
+      "step": 33449
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.766309002065387e-06,
+      "loss": 0.278,
+      "step": 33450
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.765973166759022e-06,
+      "loss": 0.2831,
+      "step": 33451
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.765637331452657e-06,
+      "loss": 0.3185,
+      "step": 33452
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.76530149614629e-06,
+      "loss": 0.2676,
+      "step": 33453
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.764965660839924e-06,
+      "loss": 0.4517,
+      "step": 33454
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.764629825533559e-06,
+      "loss": 0.299,
+      "step": 33455
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.764293990227193e-06,
+      "loss": 0.3056,
+      "step": 33456
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.763958154920828e-06,
+      "loss": 0.3876,
+      "step": 33457
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.763622319614463e-06,
+      "loss": 0.248,
+      "step": 33458
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.763286484308096e-06,
+      "loss": 0.3276,
+      "step": 33459
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.76295064900173e-06,
+      "loss": 0.2745,
+      "step": 33460
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.762614813695365e-06,
+      "loss": 0.323,
+      "step": 33461
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.762278978388998e-06,
+      "loss": 0.3426,
+      "step": 33462
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.761943143082633e-06,
+      "loss": 0.2846,
+      "step": 33463
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.761607307776267e-06,
+      "loss": 0.3435,
+      "step": 33464
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.761271472469902e-06,
+      "loss": 0.3277,
+      "step": 33465
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.760935637163536e-06,
+      "loss": 0.3188,
+      "step": 33466
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.760599801857171e-06,
+      "loss": 0.4075,
+      "step": 33467
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.760263966550804e-06,
+      "loss": 0.1537,
+      "step": 33468
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.759928131244439e-06,
+      "loss": 0.4787,
+      "step": 33469
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.759592295938073e-06,
+      "loss": 0.4576,
+      "step": 33470
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.759256460631706e-06,
+      "loss": 0.412,
+      "step": 33471
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.758920625325341e-06,
+      "loss": 0.3028,
+      "step": 33472
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.758584790018976e-06,
+      "loss": 0.4531,
+      "step": 33473
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.758248954712609e-06,
+      "loss": 0.212,
+      "step": 33474
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.757913119406243e-06,
+      "loss": 0.2615,
+      "step": 33475
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.757577284099878e-06,
+      "loss": 0.2937,
+      "step": 33476
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.757241448793513e-06,
+      "loss": 0.3614,
+      "step": 33477
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.756905613487147e-06,
+      "loss": 0.2606,
+      "step": 33478
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.756569778180782e-06,
+      "loss": 0.3852,
+      "step": 33479
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.756233942874415e-06,
+      "loss": 0.1701,
+      "step": 33480
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.75589810756805e-06,
+      "loss": 0.4356,
+      "step": 33481
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.755562272261684e-06,
+      "loss": 0.382,
+      "step": 33482
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.755226436955317e-06,
+      "loss": 0.2091,
+      "step": 33483
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.754890601648952e-06,
+      "loss": 0.2751,
+      "step": 33484
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.754554766342586e-06,
+      "loss": 0.2374,
+      "step": 33485
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.754218931036221e-06,
+      "loss": 0.3096,
+      "step": 33486
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.753883095729856e-06,
+      "loss": 0.1874,
+      "step": 33487
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.75354726042349e-06,
+      "loss": 0.3302,
+      "step": 33488
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.753211425117123e-06,
+      "loss": 0.2937,
+      "step": 33489
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.752875589810758e-06,
+      "loss": 0.2543,
+      "step": 33490
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.752539754504393e-06,
+      "loss": 0.4015,
+      "step": 33491
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.752203919198025e-06,
+      "loss": 0.2569,
+      "step": 33492
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.75186808389166e-06,
+      "loss": 0.3782,
+      "step": 33493
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.751532248585295e-06,
+      "loss": 0.476,
+      "step": 33494
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.751196413278928e-06,
+      "loss": 0.2956,
+      "step": 33495
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.750860577972562e-06,
+      "loss": 0.205,
+      "step": 33496
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.750524742666197e-06,
+      "loss": 0.3481,
+      "step": 33497
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.750188907359832e-06,
+      "loss": 0.3197,
+      "step": 33498
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.749853072053466e-06,
+      "loss": 0.4493,
+      "step": 33499
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.749517236747101e-06,
+      "loss": 0.404,
+      "step": 33500
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.749181401440734e-06,
+      "loss": 0.3317,
+      "step": 33501
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.748845566134369e-06,
+      "loss": 0.3716,
+      "step": 33502
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.748509730828003e-06,
+      "loss": 0.4375,
+      "step": 33503
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.748173895521636e-06,
+      "loss": 0.3918,
+      "step": 33504
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.74783806021527e-06,
+      "loss": 0.2668,
+      "step": 33505
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.747502224908905e-06,
+      "loss": 0.3317,
+      "step": 33506
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.74716638960254e-06,
+      "loss": 0.3866,
+      "step": 33507
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.746830554296175e-06,
+      "loss": 0.4723,
+      "step": 33508
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.74649471898981e-06,
+      "loss": 0.2786,
+      "step": 33509
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.746158883683442e-06,
+      "loss": 0.2644,
+      "step": 33510
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.745823048377077e-06,
+      "loss": 0.1441,
+      "step": 33511
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.745487213070712e-06,
+      "loss": 0.3346,
+      "step": 33512
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.745151377764345e-06,
+      "loss": 0.2609,
+      "step": 33513
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.74481554245798e-06,
+      "loss": 0.3366,
+      "step": 33514
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.744479707151614e-06,
+      "loss": 0.3832,
+      "step": 33515
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.744143871845247e-06,
+      "loss": 0.5543,
+      "step": 33516
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.743808036538881e-06,
+      "loss": 0.2565,
+      "step": 33517
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.743472201232516e-06,
+      "loss": 0.3278,
+      "step": 33518
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.74313636592615e-06,
+      "loss": 0.2761,
+      "step": 33519
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.742800530619785e-06,
+      "loss": 0.3335,
+      "step": 33520
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.74246469531342e-06,
+      "loss": 0.1527,
+      "step": 33521
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.742128860007053e-06,
+      "loss": 0.3124,
+      "step": 33522
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.741793024700688e-06,
+      "loss": 0.2222,
+      "step": 33523
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.74145718939432e-06,
+      "loss": 0.2644,
+      "step": 33524
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.741121354087955e-06,
+      "loss": 0.314,
+      "step": 33525
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.74078551878159e-06,
+      "loss": 0.3997,
+      "step": 33526
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.740449683475225e-06,
+      "loss": 0.2691,
+      "step": 33527
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.74011384816886e-06,
+      "loss": 0.2763,
+      "step": 33528
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.739778012862494e-06,
+      "loss": 0.3047,
+      "step": 33529
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.739442177556127e-06,
+      "loss": 0.1973,
+      "step": 33530
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.739106342249761e-06,
+      "loss": 0.431,
+      "step": 33531
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.738770506943396e-06,
+      "loss": 0.2905,
+      "step": 33532
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.738434671637029e-06,
+      "loss": 0.3296,
+      "step": 33533
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.738098836330664e-06,
+      "loss": 0.2566,
+      "step": 33534
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.737763001024298e-06,
+      "loss": 0.2353,
+      "step": 33535
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.737427165717931e-06,
+      "loss": 0.2953,
+      "step": 33536
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.737091330411566e-06,
+      "loss": 0.2654,
+      "step": 33537
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.7367554951052e-06,
+      "loss": 0.3339,
+      "step": 33538
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.736419659798835e-06,
+      "loss": 0.3324,
+      "step": 33539
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.73608382449247e-06,
+      "loss": 0.1832,
+      "step": 33540
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.735747989186105e-06,
+      "loss": 0.2062,
+      "step": 33541
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.735412153879738e-06,
+      "loss": 0.2083,
+      "step": 33542
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.735076318573372e-06,
+      "loss": 0.1464,
+      "step": 33543
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.734740483267007e-06,
+      "loss": 0.3672,
+      "step": 33544
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.73440464796064e-06,
+      "loss": 0.3467,
+      "step": 33545
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.734068812654274e-06,
+      "loss": 0.2555,
+      "step": 33546
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.733732977347909e-06,
+      "loss": 0.171,
+      "step": 33547
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.733397142041544e-06,
+      "loss": 0.3992,
+      "step": 33548
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.733061306735178e-06,
+      "loss": 0.2579,
+      "step": 33549
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.732725471428813e-06,
+      "loss": 0.4547,
+      "step": 33550
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.732389636122446e-06,
+      "loss": 0.4825,
+      "step": 33551
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.73205380081608e-06,
+      "loss": 0.3671,
+      "step": 33552
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.731717965509715e-06,
+      "loss": 0.1914,
+      "step": 33553
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.731382130203348e-06,
+      "loss": 0.2286,
+      "step": 33554
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.731046294896983e-06,
+      "loss": 0.5305,
+      "step": 33555
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.730710459590617e-06,
+      "loss": 0.2591,
+      "step": 33556
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.730374624284252e-06,
+      "loss": 0.2371,
+      "step": 33557
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.730038788977885e-06,
+      "loss": 0.2856,
+      "step": 33558
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.72970295367152e-06,
+      "loss": 0.1832,
+      "step": 33559
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.729367118365154e-06,
+      "loss": 0.399,
+      "step": 33560
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.729031283058789e-06,
+      "loss": 0.2655,
+      "step": 33561
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.728695447752424e-06,
+      "loss": 0.4831,
+      "step": 33562
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.728359612446057e-06,
+      "loss": 0.2822,
+      "step": 33563
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.728023777139691e-06,
+      "loss": 0.2451,
+      "step": 33564
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.727687941833326e-06,
+      "loss": 0.3058,
+      "step": 33565
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.727352106526959e-06,
+      "loss": 0.3951,
+      "step": 33566
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.727016271220594e-06,
+      "loss": 0.2914,
+      "step": 33567
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.726680435914228e-06,
+      "loss": 0.2517,
+      "step": 33568
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.726344600607863e-06,
+      "loss": 0.2898,
+      "step": 33569
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.726008765301497e-06,
+      "loss": 0.384,
+      "step": 33570
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.725672929995132e-06,
+      "loss": 0.1386,
+      "step": 33571
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.725337094688765e-06,
+      "loss": 0.2006,
+      "step": 33572
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.7250012593824e-06,
+      "loss": 0.2852,
+      "step": 33573
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.724665424076034e-06,
+      "loss": 0.2586,
+      "step": 33574
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.724329588769667e-06,
+      "loss": 0.4944,
+      "step": 33575
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.723993753463302e-06,
+      "loss": 0.4418,
+      "step": 33576
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.723657918156937e-06,
+      "loss": 0.2298,
+      "step": 33577
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.723322082850571e-06,
+      "loss": 0.2168,
+      "step": 33578
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.722986247544206e-06,
+      "loss": 0.3515,
+      "step": 33579
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.72265041223784e-06,
+      "loss": 0.2455,
+      "step": 33580
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.722314576931474e-06,
+      "loss": 0.2518,
+      "step": 33581
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.721978741625108e-06,
+      "loss": 0.3009,
+      "step": 33582
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.721642906318743e-06,
+      "loss": 0.427,
+      "step": 33583
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.721307071012376e-06,
+      "loss": 0.2936,
+      "step": 33584
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.72097123570601e-06,
+      "loss": 0.2732,
+      "step": 33585
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.720635400399645e-06,
+      "loss": 0.2478,
+      "step": 33586
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.720299565093278e-06,
+      "loss": 0.2177,
+      "step": 33587
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.719963729786913e-06,
+      "loss": 0.2251,
+      "step": 33588
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.719627894480547e-06,
+      "loss": 0.215,
+      "step": 33589
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.719292059174182e-06,
+      "loss": 0.3745,
+      "step": 33590
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.718956223867817e-06,
+      "loss": 0.1144,
+      "step": 33591
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.718620388561451e-06,
+      "loss": 0.1498,
+      "step": 33592
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.718284553255084e-06,
+      "loss": 0.3312,
+      "step": 33593
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.717948717948719e-06,
+      "loss": 0.329,
+      "step": 33594
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.717612882642353e-06,
+      "loss": 0.5147,
+      "step": 33595
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.717277047335986e-06,
+      "loss": 0.2675,
+      "step": 33596
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.716941212029621e-06,
+      "loss": 0.2006,
+      "step": 33597
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.716605376723256e-06,
+      "loss": 0.2591,
+      "step": 33598
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.71626954141689e-06,
+      "loss": 0.3272,
+      "step": 33599
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.715933706110525e-06,
+      "loss": 0.2329,
+      "step": 33600
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.71559787080416e-06,
+      "loss": 0.3407,
+      "step": 33601
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.715262035497793e-06,
+      "loss": 0.2029,
+      "step": 33602
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.714926200191427e-06,
+      "loss": 0.3415,
+      "step": 33603
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.714590364885062e-06,
+      "loss": 0.2382,
+      "step": 33604
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.714254529578695e-06,
+      "loss": 0.2274,
+      "step": 33605
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.71391869427233e-06,
+      "loss": 0.353,
+      "step": 33606
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.713582858965964e-06,
+      "loss": 0.3079,
+      "step": 33607
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.713247023659597e-06,
+      "loss": 0.4117,
+      "step": 33608
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.712911188353232e-06,
+      "loss": 0.3925,
+      "step": 33609
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.712575353046866e-06,
+      "loss": 0.2276,
+      "step": 33610
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.712239517740501e-06,
+      "loss": 0.2654,
+      "step": 33611
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.711903682434136e-06,
+      "loss": 0.2372,
+      "step": 33612
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.711567847127769e-06,
+      "loss": 0.219,
+      "step": 33613
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.711232011821403e-06,
+      "loss": 0.1539,
+      "step": 33614
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.710896176515038e-06,
+      "loss": 0.2712,
+      "step": 33615
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.710560341208671e-06,
+      "loss": 0.3544,
+      "step": 33616
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.710224505902306e-06,
+      "loss": 0.2097,
+      "step": 33617
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.70988867059594e-06,
+      "loss": 0.3998,
+      "step": 33618
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.709552835289575e-06,
+      "loss": 0.3766,
+      "step": 33619
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.70921699998321e-06,
+      "loss": 0.2979,
+      "step": 33620
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.708881164676844e-06,
+      "loss": 0.1767,
+      "step": 33621
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.708545329370477e-06,
+      "loss": 0.3155,
+      "step": 33622
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.708209494064112e-06,
+      "loss": 0.3165,
+      "step": 33623
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.707873658757746e-06,
+      "loss": 0.4239,
+      "step": 33624
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.70753782345138e-06,
+      "loss": 0.2909,
+      "step": 33625
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.707201988145014e-06,
+      "loss": 0.2652,
+      "step": 33626
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.706866152838649e-06,
+      "loss": 0.2886,
+      "step": 33627
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.706530317532282e-06,
+      "loss": 0.3737,
+      "step": 33628
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.706194482225916e-06,
+      "loss": 0.1961,
+      "step": 33629
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.705858646919551e-06,
+      "loss": 0.3187,
+      "step": 33630
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.705522811613186e-06,
+      "loss": 0.2708,
+      "step": 33631
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.70518697630682e-06,
+      "loss": 0.3995,
+      "step": 33632
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.704851141000455e-06,
+      "loss": 0.3416,
+      "step": 33633
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.704515305694088e-06,
+      "loss": 0.1494,
+      "step": 33634
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.704179470387722e-06,
+      "loss": 0.2704,
+      "step": 33635
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.703843635081357e-06,
+      "loss": 0.2015,
+      "step": 33636
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.70350779977499e-06,
+      "loss": 0.2823,
+      "step": 33637
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.703171964468625e-06,
+      "loss": 0.2922,
+      "step": 33638
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.70283612916226e-06,
+      "loss": 0.3141,
+      "step": 33639
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.702500293855894e-06,
+      "loss": 0.375,
+      "step": 33640
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.702164458549529e-06,
+      "loss": 0.3132,
+      "step": 33641
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.701828623243163e-06,
+      "loss": 0.3234,
+      "step": 33642
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.701492787936796e-06,
+      "loss": 0.1656,
+      "step": 33643
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.701156952630431e-06,
+      "loss": 0.3648,
+      "step": 33644
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.700821117324066e-06,
+      "loss": 0.3364,
+      "step": 33645
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.700485282017698e-06,
+      "loss": 0.3151,
+      "step": 33646
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 8.700149446711333e-06,
+      "loss": 0.2489,
+      "step": 33647
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.699813611404968e-06,
+      "loss": 0.2306,
+      "step": 33648
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.6994777760986e-06,
+      "loss": 0.2913,
+      "step": 33649
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.699141940792235e-06,
+      "loss": 0.2954,
+      "step": 33650
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.69880610548587e-06,
+      "loss": 0.2837,
+      "step": 33651
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.698470270179505e-06,
+      "loss": 0.4489,
+      "step": 33652
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.69813443487314e-06,
+      "loss": 0.2664,
+      "step": 33653
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.697798599566774e-06,
+      "loss": 0.2019,
+      "step": 33654
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.697462764260407e-06,
+      "loss": 0.2942,
+      "step": 33655
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.697126928954042e-06,
+      "loss": 0.3006,
+      "step": 33656
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.696791093647676e-06,
+      "loss": 0.2704,
+      "step": 33657
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.69645525834131e-06,
+      "loss": 0.2306,
+      "step": 33658
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.696119423034944e-06,
+      "loss": 0.2724,
+      "step": 33659
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.695783587728578e-06,
+      "loss": 0.4174,
+      "step": 33660
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.695447752422213e-06,
+      "loss": 0.2676,
+      "step": 33661
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.695111917115848e-06,
+      "loss": 0.2267,
+      "step": 33662
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.694776081809482e-06,
+      "loss": 0.4742,
+      "step": 33663
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.694440246503115e-06,
+      "loss": 0.3015,
+      "step": 33664
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.69410441119675e-06,
+      "loss": 0.1579,
+      "step": 33665
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.693768575890385e-06,
+      "loss": 0.2277,
+      "step": 33666
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.693432740584018e-06,
+      "loss": 0.3307,
+      "step": 33667
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.693096905277652e-06,
+      "loss": 0.343,
+      "step": 33668
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.692761069971287e-06,
+      "loss": 0.221,
+      "step": 33669
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.692425234664922e-06,
+      "loss": 0.1212,
+      "step": 33670
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.692089399358554e-06,
+      "loss": 0.3936,
+      "step": 33671
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.691753564052189e-06,
+      "loss": 0.335,
+      "step": 33672
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.691417728745824e-06,
+      "loss": 0.2473,
+      "step": 33673
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.691081893439458e-06,
+      "loss": 0.3046,
+      "step": 33674
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.690746058133093e-06,
+      "loss": 0.2054,
+      "step": 33675
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.690410222826726e-06,
+      "loss": 0.2432,
+      "step": 33676
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.69007438752036e-06,
+      "loss": 0.3356,
+      "step": 33677
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.689738552213995e-06,
+      "loss": 0.3715,
+      "step": 33678
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.689402716907628e-06,
+      "loss": 0.513,
+      "step": 33679
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.689066881601263e-06,
+      "loss": 0.3419,
+      "step": 33680
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.688731046294898e-06,
+      "loss": 0.331,
+      "step": 33681
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.688395210988532e-06,
+      "loss": 0.2839,
+      "step": 33682
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.688059375682167e-06,
+      "loss": 0.3387,
+      "step": 33683
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.687723540375802e-06,
+      "loss": 0.2661,
+      "step": 33684
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.687387705069434e-06,
+      "loss": 0.2525,
+      "step": 33685
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.687051869763069e-06,
+      "loss": 0.3057,
+      "step": 33686
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.686716034456704e-06,
+      "loss": 0.3689,
+      "step": 33687
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.686380199150337e-06,
+      "loss": 0.2965,
+      "step": 33688
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.686044363843971e-06,
+      "loss": 0.3874,
+      "step": 33689
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.685708528537606e-06,
+      "loss": 0.2181,
+      "step": 33690
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.68537269323124e-06,
+      "loss": 0.188,
+      "step": 33691
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.685036857924875e-06,
+      "loss": 0.2853,
+      "step": 33692
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.684701022618508e-06,
+      "loss": 0.3188,
+      "step": 33693
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.684365187312143e-06,
+      "loss": 0.1126,
+      "step": 33694
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.684029352005778e-06,
+      "loss": 0.232,
+      "step": 33695
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.683693516699412e-06,
+      "loss": 0.4154,
+      "step": 33696
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.683357681393045e-06,
+      "loss": 0.3264,
+      "step": 33697
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.68302184608668e-06,
+      "loss": 0.2987,
+      "step": 33698
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.682686010780314e-06,
+      "loss": 0.2419,
+      "step": 33699
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.682350175473947e-06,
+      "loss": 0.2961,
+      "step": 33700
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.682014340167582e-06,
+      "loss": 0.3058,
+      "step": 33701
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.681678504861217e-06,
+      "loss": 0.1645,
+      "step": 33702
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.681342669554851e-06,
+      "loss": 0.235,
+      "step": 33703
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.681006834248486e-06,
+      "loss": 0.4006,
+      "step": 33704
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.680670998942119e-06,
+      "loss": 0.2542,
+      "step": 33705
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.680335163635754e-06,
+      "loss": 0.3335,
+      "step": 33706
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.679999328329388e-06,
+      "loss": 0.2494,
+      "step": 33707
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.679663493023021e-06,
+      "loss": 0.4745,
+      "step": 33708
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.679327657716656e-06,
+      "loss": 0.2469,
+      "step": 33709
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.67899182241029e-06,
+      "loss": 0.1927,
+      "step": 33710
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.678655987103925e-06,
+      "loss": 0.3306,
+      "step": 33711
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.67832015179756e-06,
+      "loss": 0.3652,
+      "step": 33712
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.677984316491194e-06,
+      "loss": 0.2706,
+      "step": 33713
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.677648481184827e-06,
+      "loss": 0.3143,
+      "step": 33714
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.677312645878462e-06,
+      "loss": 0.2944,
+      "step": 33715
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.676976810572097e-06,
+      "loss": 0.283,
+      "step": 33716
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.67664097526573e-06,
+      "loss": 0.3916,
+      "step": 33717
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.676305139959364e-06,
+      "loss": 0.592,
+      "step": 33718
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.675969304652999e-06,
+      "loss": 0.456,
+      "step": 33719
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.675633469346632e-06,
+      "loss": 0.473,
+      "step": 33720
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.675297634040267e-06,
+      "loss": 0.2012,
+      "step": 33721
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.674961798733901e-06,
+      "loss": 0.3266,
+      "step": 33722
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.674625963427536e-06,
+      "loss": 0.3855,
+      "step": 33723
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.67429012812117e-06,
+      "loss": 0.1939,
+      "step": 33724
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.673954292814805e-06,
+      "loss": 0.2994,
+      "step": 33725
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.673618457508438e-06,
+      "loss": 0.2426,
+      "step": 33726
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.673282622202073e-06,
+      "loss": 0.3136,
+      "step": 33727
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.672946786895707e-06,
+      "loss": 0.2297,
+      "step": 33728
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.67261095158934e-06,
+      "loss": 0.3315,
+      "step": 33729
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.672275116282975e-06,
+      "loss": 0.4078,
+      "step": 33730
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.67193928097661e-06,
+      "loss": 0.3366,
+      "step": 33731
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.671603445670244e-06,
+      "loss": 0.1654,
+      "step": 33732
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.671267610363879e-06,
+      "loss": 0.3249,
+      "step": 33733
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.670931775057514e-06,
+      "loss": 0.2769,
+      "step": 33734
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.670595939751147e-06,
+      "loss": 0.2267,
+      "step": 33735
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.670260104444781e-06,
+      "loss": 0.1824,
+      "step": 33736
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.669924269138416e-06,
+      "loss": 0.2193,
+      "step": 33737
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.669588433832049e-06,
+      "loss": 0.2639,
+      "step": 33738
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.669252598525683e-06,
+      "loss": 0.1745,
+      "step": 33739
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.668916763219318e-06,
+      "loss": 0.3273,
+      "step": 33740
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.668580927912951e-06,
+      "loss": 0.2296,
+      "step": 33741
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.668245092606586e-06,
+      "loss": 0.1741,
+      "step": 33742
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.66790925730022e-06,
+      "loss": 0.3197,
+      "step": 33743
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.667573421993855e-06,
+      "loss": 0.3675,
+      "step": 33744
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.66723758668749e-06,
+      "loss": 0.4073,
+      "step": 33745
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.666901751381124e-06,
+      "loss": 0.3454,
+      "step": 33746
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.666565916074757e-06,
+      "loss": 0.3653,
+      "step": 33747
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.666230080768392e-06,
+      "loss": 0.1019,
+      "step": 33748
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.665894245462026e-06,
+      "loss": 0.3395,
+      "step": 33749
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.66555841015566e-06,
+      "loss": 0.3487,
+      "step": 33750
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.665222574849294e-06,
+      "loss": 0.3459,
+      "step": 33751
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.664886739542929e-06,
+      "loss": 0.2152,
+      "step": 33752
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.664550904236563e-06,
+      "loss": 0.2582,
+      "step": 33753
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.664215068930198e-06,
+      "loss": 0.2853,
+      "step": 33754
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.663879233623833e-06,
+      "loss": 0.2806,
+      "step": 33755
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.663543398317466e-06,
+      "loss": 0.1936,
+      "step": 33756
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.6632075630111e-06,
+      "loss": 0.2561,
+      "step": 33757
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.662871727704735e-06,
+      "loss": 0.4241,
+      "step": 33758
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.662535892398368e-06,
+      "loss": 0.228,
+      "step": 33759
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.662200057092003e-06,
+      "loss": 0.3391,
+      "step": 33760
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.661864221785637e-06,
+      "loss": 0.2846,
+      "step": 33761
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.66152838647927e-06,
+      "loss": 0.4349,
+      "step": 33762
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.661192551172905e-06,
+      "loss": 0.2894,
+      "step": 33763
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.66085671586654e-06,
+      "loss": 0.3233,
+      "step": 33764
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.660520880560174e-06,
+      "loss": 0.2981,
+      "step": 33765
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.660185045253809e-06,
+      "loss": 0.2318,
+      "step": 33766
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.659849209947443e-06,
+      "loss": 0.1885,
+      "step": 33767
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.659513374641076e-06,
+      "loss": 0.2402,
+      "step": 33768
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.659177539334711e-06,
+      "loss": 0.2563,
+      "step": 33769
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.658841704028346e-06,
+      "loss": 0.2524,
+      "step": 33770
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.658505868721979e-06,
+      "loss": 0.3624,
+      "step": 33771
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.658170033415613e-06,
+      "loss": 0.1811,
+      "step": 33772
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.657834198109248e-06,
+      "loss": 0.452,
+      "step": 33773
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.657498362802883e-06,
+      "loss": 0.4126,
+      "step": 33774
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.657162527496517e-06,
+      "loss": 0.3167,
+      "step": 33775
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.656826692190152e-06,
+      "loss": 0.3163,
+      "step": 33776
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.656490856883785e-06,
+      "loss": 0.2686,
+      "step": 33777
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.65615502157742e-06,
+      "loss": 0.4242,
+      "step": 33778
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.655819186271054e-06,
+      "loss": 0.284,
+      "step": 33779
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.655483350964687e-06,
+      "loss": 0.4749,
+      "step": 33780
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.655147515658322e-06,
+      "loss": 0.3508,
+      "step": 33781
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.654811680351956e-06,
+      "loss": 0.2648,
+      "step": 33782
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.654475845045591e-06,
+      "loss": 0.3461,
+      "step": 33783
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.654140009739224e-06,
+      "loss": 0.3974,
+      "step": 33784
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.653804174432859e-06,
+      "loss": 0.287,
+      "step": 33785
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.653468339126493e-06,
+      "loss": 0.2942,
+      "step": 33786
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.653132503820128e-06,
+      "loss": 0.3717,
+      "step": 33787
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.652796668513763e-06,
+      "loss": 0.2352,
+      "step": 33788
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.652460833207395e-06,
+      "loss": 0.2283,
+      "step": 33789
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.65212499790103e-06,
+      "loss": 0.2807,
+      "step": 33790
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.651789162594665e-06,
+      "loss": 0.3176,
+      "step": 33791
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.651453327288298e-06,
+      "loss": 0.329,
+      "step": 33792
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.651117491981932e-06,
+      "loss": 0.3195,
+      "step": 33793
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.650781656675567e-06,
+      "loss": 0.4906,
+      "step": 33794
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.650445821369202e-06,
+      "loss": 0.1989,
+      "step": 33795
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.650109986062836e-06,
+      "loss": 0.255,
+      "step": 33796
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.64977415075647e-06,
+      "loss": 0.2443,
+      "step": 33797
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.649438315450104e-06,
+      "loss": 0.247,
+      "step": 33798
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.649102480143739e-06,
+      "loss": 0.3505,
+      "step": 33799
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.648766644837371e-06,
+      "loss": 0.4666,
+      "step": 33800
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.648430809531006e-06,
+      "loss": 0.2828,
+      "step": 33801
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.64809497422464e-06,
+      "loss": 0.1712,
+      "step": 33802
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.647759138918275e-06,
+      "loss": 0.2748,
+      "step": 33803
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.64742330361191e-06,
+      "loss": 0.3312,
+      "step": 33804
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.647087468305545e-06,
+      "loss": 0.1801,
+      "step": 33805
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.646751632999178e-06,
+      "loss": 0.2866,
+      "step": 33806
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.646415797692812e-06,
+      "loss": 0.1828,
+      "step": 33807
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.646079962386447e-06,
+      "loss": 0.1906,
+      "step": 33808
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.64574412708008e-06,
+      "loss": 0.1687,
+      "step": 33809
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.645408291773715e-06,
+      "loss": 0.2763,
+      "step": 33810
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.64507245646735e-06,
+      "loss": 0.3117,
+      "step": 33811
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.644736621160982e-06,
+      "loss": 0.3477,
+      "step": 33812
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.644400785854617e-06,
+      "loss": 0.1969,
+      "step": 33813
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.644064950548251e-06,
+      "loss": 0.1578,
+      "step": 33814
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.643729115241886e-06,
+      "loss": 0.3604,
+      "step": 33815
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.64339327993552e-06,
+      "loss": 0.3314,
+      "step": 33816
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.643057444629155e-06,
+      "loss": 0.2924,
+      "step": 33817
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.642721609322788e-06,
+      "loss": 0.2871,
+      "step": 33818
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.642385774016423e-06,
+      "loss": 0.16,
+      "step": 33819
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.642049938710058e-06,
+      "loss": 0.3075,
+      "step": 33820
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.64171410340369e-06,
+      "loss": 0.2883,
+      "step": 33821
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.641378268097325e-06,
+      "loss": 0.3049,
+      "step": 33822
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.64104243279096e-06,
+      "loss": 0.3189,
+      "step": 33823
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.640706597484595e-06,
+      "loss": 0.3595,
+      "step": 33824
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.64037076217823e-06,
+      "loss": 0.3237,
+      "step": 33825
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.640034926871864e-06,
+      "loss": 0.3153,
+      "step": 33826
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.639699091565497e-06,
+      "loss": 0.4293,
+      "step": 33827
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.639363256259131e-06,
+      "loss": 0.381,
+      "step": 33828
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.639027420952766e-06,
+      "loss": 0.1309,
+      "step": 33829
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.638691585646399e-06,
+      "loss": 0.316,
+      "step": 33830
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.638355750340034e-06,
+      "loss": 0.2506,
+      "step": 33831
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.638019915033668e-06,
+      "loss": 0.2883,
+      "step": 33832
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.637684079727301e-06,
+      "loss": 0.2519,
+      "step": 33833
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.637348244420936e-06,
+      "loss": 0.2871,
+      "step": 33834
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.63701240911457e-06,
+      "loss": 0.298,
+      "step": 33835
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.636676573808205e-06,
+      "loss": 0.2603,
+      "step": 33836
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.63634073850184e-06,
+      "loss": 0.5293,
+      "step": 33837
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.636004903195475e-06,
+      "loss": 0.475,
+      "step": 33838
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.635669067889107e-06,
+      "loss": 0.4403,
+      "step": 33839
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.635333232582742e-06,
+      "loss": 0.2046,
+      "step": 33840
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.634997397276377e-06,
+      "loss": 0.3577,
+      "step": 33841
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.63466156197001e-06,
+      "loss": 0.3135,
+      "step": 33842
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.634325726663644e-06,
+      "loss": 0.3623,
+      "step": 33843
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.633989891357279e-06,
+      "loss": 0.427,
+      "step": 33844
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 8.633654056050914e-06,
+      "loss": 0.2484,
+      "step": 33845
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.633318220744548e-06,
+      "loss": 0.3615,
+      "step": 33846
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.632982385438183e-06,
+      "loss": 0.318,
+      "step": 33847
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.632646550131816e-06,
+      "loss": 0.4078,
+      "step": 33848
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.63231071482545e-06,
+      "loss": 0.3562,
+      "step": 33849
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.631974879519085e-06,
+      "loss": 0.263,
+      "step": 33850
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.631639044212718e-06,
+      "loss": 0.3655,
+      "step": 33851
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.631303208906353e-06,
+      "loss": 0.2063,
+      "step": 33852
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.630967373599987e-06,
+      "loss": 0.3996,
+      "step": 33853
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.63063153829362e-06,
+      "loss": 0.2495,
+      "step": 33854
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.630295702987255e-06,
+      "loss": 0.4293,
+      "step": 33855
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.62995986768089e-06,
+      "loss": 0.2128,
+      "step": 33856
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.629624032374524e-06,
+      "loss": 0.2617,
+      "step": 33857
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.629288197068159e-06,
+      "loss": 0.3029,
+      "step": 33858
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.628952361761794e-06,
+      "loss": 0.2308,
+      "step": 33859
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.628616526455427e-06,
+      "loss": 0.3915,
+      "step": 33860
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.628280691149061e-06,
+      "loss": 0.2461,
+      "step": 33861
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.627944855842696e-06,
+      "loss": 0.1756,
+      "step": 33862
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.627609020536329e-06,
+      "loss": 0.4035,
+      "step": 33863
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.627273185229964e-06,
+      "loss": 0.3773,
+      "step": 33864
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.626937349923598e-06,
+      "loss": 0.2439,
+      "step": 33865
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.626601514617233e-06,
+      "loss": 0.178,
+      "step": 33866
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.626265679310867e-06,
+      "loss": 0.2351,
+      "step": 33867
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.625929844004502e-06,
+      "loss": 0.2533,
+      "step": 33868
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.625594008698135e-06,
+      "loss": 0.2777,
+      "step": 33869
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.62525817339177e-06,
+      "loss": 0.2519,
+      "step": 33870
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.624922338085404e-06,
+      "loss": 0.2526,
+      "step": 33871
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.624586502779037e-06,
+      "loss": 0.5051,
+      "step": 33872
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.624250667472672e-06,
+      "loss": 0.182,
+      "step": 33873
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.623914832166307e-06,
+      "loss": 0.1944,
+      "step": 33874
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.62357899685994e-06,
+      "loss": 0.1179,
+      "step": 33875
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.623243161553574e-06,
+      "loss": 0.5415,
+      "step": 33876
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.622907326247209e-06,
+      "loss": 0.2275,
+      "step": 33877
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.622571490940843e-06,
+      "loss": 0.2384,
+      "step": 33878
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.622235655634478e-06,
+      "loss": 0.2698,
+      "step": 33879
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.621899820328113e-06,
+      "loss": 0.2692,
+      "step": 33880
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.621563985021746e-06,
+      "loss": 0.3742,
+      "step": 33881
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.62122814971538e-06,
+      "loss": 0.3077,
+      "step": 33882
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.620892314409013e-06,
+      "loss": 0.2319,
+      "step": 33883
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.620556479102648e-06,
+      "loss": 0.185,
+      "step": 33884
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.620220643796283e-06,
+      "loss": 0.3507,
+      "step": 33885
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.619884808489917e-06,
+      "loss": 0.2185,
+      "step": 33886
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.619548973183552e-06,
+      "loss": 0.2792,
+      "step": 33887
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.619213137877187e-06,
+      "loss": 0.3282,
+      "step": 33888
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.61887730257082e-06,
+      "loss": 0.4114,
+      "step": 33889
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.618541467264454e-06,
+      "loss": 0.1776,
+      "step": 33890
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.618205631958089e-06,
+      "loss": 0.2649,
+      "step": 33891
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.617869796651722e-06,
+      "loss": 0.2973,
+      "step": 33892
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.617533961345356e-06,
+      "loss": 0.3464,
+      "step": 33893
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.617198126038991e-06,
+      "loss": 0.311,
+      "step": 33894
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.616862290732626e-06,
+      "loss": 0.4077,
+      "step": 33895
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.61652645542626e-06,
+      "loss": 0.1375,
+      "step": 33896
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.616190620119893e-06,
+      "loss": 0.2809,
+      "step": 33897
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.615854784813528e-06,
+      "loss": 0.27,
+      "step": 33898
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.615518949507163e-06,
+      "loss": 0.2051,
+      "step": 33899
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.615183114200797e-06,
+      "loss": 0.3021,
+      "step": 33900
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.61484727889443e-06,
+      "loss": 0.3493,
+      "step": 33901
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.614511443588065e-06,
+      "loss": 0.2121,
+      "step": 33902
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.6141756082817e-06,
+      "loss": 0.3239,
+      "step": 33903
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.613839772975332e-06,
+      "loss": 0.2541,
+      "step": 33904
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.613503937668967e-06,
+      "loss": 0.3176,
+      "step": 33905
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.613168102362602e-06,
+      "loss": 0.3613,
+      "step": 33906
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.612832267056236e-06,
+      "loss": 0.4477,
+      "step": 33907
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.612496431749871e-06,
+      "loss": 0.2322,
+      "step": 33908
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.612160596443506e-06,
+      "loss": 0.2294,
+      "step": 33909
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.611824761137139e-06,
+      "loss": 0.2416,
+      "step": 33910
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.611488925830773e-06,
+      "loss": 0.2708,
+      "step": 33911
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.611153090524408e-06,
+      "loss": 0.3072,
+      "step": 33912
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.610817255218041e-06,
+      "loss": 0.4993,
+      "step": 33913
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.610481419911676e-06,
+      "loss": 0.2268,
+      "step": 33914
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.61014558460531e-06,
+      "loss": 0.357,
+      "step": 33915
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.609809749298945e-06,
+      "loss": 0.4424,
+      "step": 33916
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.60947391399258e-06,
+      "loss": 0.284,
+      "step": 33917
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.609138078686214e-06,
+      "loss": 0.2967,
+      "step": 33918
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.608802243379847e-06,
+      "loss": 0.3591,
+      "step": 33919
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.608466408073482e-06,
+      "loss": 0.226,
+      "step": 33920
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.608130572767116e-06,
+      "loss": 0.2867,
+      "step": 33921
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.60779473746075e-06,
+      "loss": 0.1852,
+      "step": 33922
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.607458902154384e-06,
+      "loss": 0.2778,
+      "step": 33923
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.607123066848019e-06,
+      "loss": 0.1561,
+      "step": 33924
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.606787231541652e-06,
+      "loss": 0.5134,
+      "step": 33925
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.606451396235286e-06,
+      "loss": 0.2733,
+      "step": 33926
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.606115560928921e-06,
+      "loss": 0.2589,
+      "step": 33927
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.605779725622556e-06,
+      "loss": 0.2523,
+      "step": 33928
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.60544389031619e-06,
+      "loss": 0.2009,
+      "step": 33929
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.605108055009825e-06,
+      "loss": 0.3842,
+      "step": 33930
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.604772219703458e-06,
+      "loss": 0.2434,
+      "step": 33931
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.604436384397092e-06,
+      "loss": 0.1973,
+      "step": 33932
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.604100549090727e-06,
+      "loss": 0.4057,
+      "step": 33933
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.60376471378436e-06,
+      "loss": 0.3022,
+      "step": 33934
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.603428878477995e-06,
+      "loss": 0.2858,
+      "step": 33935
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.60309304317163e-06,
+      "loss": 0.2517,
+      "step": 33936
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.602757207865264e-06,
+      "loss": 0.4179,
+      "step": 33937
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.602421372558899e-06,
+      "loss": 0.4208,
+      "step": 33938
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.602085537252533e-06,
+      "loss": 0.2963,
+      "step": 33939
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.601749701946166e-06,
+      "loss": 0.3229,
+      "step": 33940
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.601413866639801e-06,
+      "loss": 0.3502,
+      "step": 33941
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.601078031333436e-06,
+      "loss": 0.2816,
+      "step": 33942
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.600742196027068e-06,
+      "loss": 0.364,
+      "step": 33943
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.600406360720703e-06,
+      "loss": 0.2658,
+      "step": 33944
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.600070525414338e-06,
+      "loss": 0.3009,
+      "step": 33945
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.59973469010797e-06,
+      "loss": 0.1859,
+      "step": 33946
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.599398854801605e-06,
+      "loss": 0.299,
+      "step": 33947
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.59906301949524e-06,
+      "loss": 0.2979,
+      "step": 33948
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.598727184188875e-06,
+      "loss": 0.363,
+      "step": 33949
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.59839134888251e-06,
+      "loss": 0.2318,
+      "step": 33950
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.598055513576144e-06,
+      "loss": 0.2601,
+      "step": 33951
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.597719678269777e-06,
+      "loss": 0.3257,
+      "step": 33952
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.597383842963412e-06,
+      "loss": 0.3858,
+      "step": 33953
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.597048007657046e-06,
+      "loss": 0.2924,
+      "step": 33954
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.59671217235068e-06,
+      "loss": 0.2463,
+      "step": 33955
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.596376337044314e-06,
+      "loss": 0.1356,
+      "step": 33956
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.596040501737948e-06,
+      "loss": 0.3114,
+      "step": 33957
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.595704666431583e-06,
+      "loss": 0.2633,
+      "step": 33958
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.595368831125218e-06,
+      "loss": 0.2644,
+      "step": 33959
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.595032995818852e-06,
+      "loss": 0.4549,
+      "step": 33960
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.594697160512485e-06,
+      "loss": 0.4545,
+      "step": 33961
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.59436132520612e-06,
+      "loss": 0.2011,
+      "step": 33962
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.594025489899755e-06,
+      "loss": 0.2642,
+      "step": 33963
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.593689654593388e-06,
+      "loss": 0.1896,
+      "step": 33964
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.593353819287022e-06,
+      "loss": 0.2711,
+      "step": 33965
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.593017983980657e-06,
+      "loss": 0.3994,
+      "step": 33966
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.59268214867429e-06,
+      "loss": 0.3233,
+      "step": 33967
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.592346313367924e-06,
+      "loss": 0.5703,
+      "step": 33968
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.592010478061559e-06,
+      "loss": 0.4025,
+      "step": 33969
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.591674642755194e-06,
+      "loss": 0.3038,
+      "step": 33970
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.591338807448828e-06,
+      "loss": 0.3661,
+      "step": 33971
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.591002972142463e-06,
+      "loss": 0.4555,
+      "step": 33972
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.590667136836096e-06,
+      "loss": 0.3503,
+      "step": 33973
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.59033130152973e-06,
+      "loss": 0.27,
+      "step": 33974
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.589995466223364e-06,
+      "loss": 0.3241,
+      "step": 33975
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.589659630916998e-06,
+      "loss": 0.3506,
+      "step": 33976
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.589323795610633e-06,
+      "loss": 0.2528,
+      "step": 33977
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.588987960304268e-06,
+      "loss": 0.1797,
+      "step": 33978
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.588652124997902e-06,
+      "loss": 0.37,
+      "step": 33979
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.588316289691537e-06,
+      "loss": 0.2745,
+      "step": 33980
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.58798045438517e-06,
+      "loss": 0.3064,
+      "step": 33981
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.587644619078804e-06,
+      "loss": 0.2166,
+      "step": 33982
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.587308783772439e-06,
+      "loss": 0.3112,
+      "step": 33983
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.586972948466072e-06,
+      "loss": 0.3303,
+      "step": 33984
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.586637113159707e-06,
+      "loss": 0.3172,
+      "step": 33985
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.586301277853341e-06,
+      "loss": 0.2786,
+      "step": 33986
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.585965442546976e-06,
+      "loss": 0.3281,
+      "step": 33987
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.585629607240609e-06,
+      "loss": 0.3227,
+      "step": 33988
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.585293771934244e-06,
+      "loss": 0.3116,
+      "step": 33989
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.584957936627878e-06,
+      "loss": 0.2433,
+      "step": 33990
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.584622101321513e-06,
+      "loss": 0.2811,
+      "step": 33991
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.584286266015148e-06,
+      "loss": 0.198,
+      "step": 33992
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.58395043070878e-06,
+      "loss": 0.2784,
+      "step": 33993
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.583614595402415e-06,
+      "loss": 0.3411,
+      "step": 33994
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.58327876009605e-06,
+      "loss": 0.2327,
+      "step": 33995
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.582942924789683e-06,
+      "loss": 0.2942,
+      "step": 33996
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.582607089483317e-06,
+      "loss": 0.3261,
+      "step": 33997
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.582271254176952e-06,
+      "loss": 0.4251,
+      "step": 33998
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.581935418870587e-06,
+      "loss": 0.3082,
+      "step": 33999
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.581599583564221e-06,
+      "loss": 0.2253,
+      "step": 34000
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.581263748257856e-06,
+      "loss": 0.3475,
+      "step": 34001
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.580927912951489e-06,
+      "loss": 0.5305,
+      "step": 34002
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.580592077645124e-06,
+      "loss": 0.2914,
+      "step": 34003
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.580256242338758e-06,
+      "loss": 0.2355,
+      "step": 34004
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.579920407032391e-06,
+      "loss": 0.1863,
+      "step": 34005
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.579584571726026e-06,
+      "loss": 0.3534,
+      "step": 34006
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.57924873641966e-06,
+      "loss": 0.4706,
+      "step": 34007
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.578912901113295e-06,
+      "loss": 0.3405,
+      "step": 34008
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.57857706580693e-06,
+      "loss": 0.2463,
+      "step": 34009
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.578241230500563e-06,
+      "loss": 0.24,
+      "step": 34010
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.577905395194197e-06,
+      "loss": 0.2292,
+      "step": 34011
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.577569559887832e-06,
+      "loss": 0.2982,
+      "step": 34012
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.577233724581467e-06,
+      "loss": 0.3279,
+      "step": 34013
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.5768978892751e-06,
+      "loss": 0.4077,
+      "step": 34014
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.576562053968734e-06,
+      "loss": 0.4597,
+      "step": 34015
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.576226218662369e-06,
+      "loss": 0.2921,
+      "step": 34016
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.575890383356002e-06,
+      "loss": 0.2607,
+      "step": 34017
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.575554548049637e-06,
+      "loss": 0.3618,
+      "step": 34018
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.575218712743271e-06,
+      "loss": 0.3032,
+      "step": 34019
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.574882877436906e-06,
+      "loss": 0.2049,
+      "step": 34020
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.57454704213054e-06,
+      "loss": 0.319,
+      "step": 34021
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.574211206824175e-06,
+      "loss": 0.3096,
+      "step": 34022
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.573875371517808e-06,
+      "loss": 0.2603,
+      "step": 34023
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.573539536211443e-06,
+      "loss": 0.3166,
+      "step": 34024
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.573203700905077e-06,
+      "loss": 0.3376,
+      "step": 34025
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.57286786559871e-06,
+      "loss": 0.4839,
+      "step": 34026
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.572532030292345e-06,
+      "loss": 0.2648,
+      "step": 34027
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.57219619498598e-06,
+      "loss": 0.3395,
+      "step": 34028
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.571860359679614e-06,
+      "loss": 0.2001,
+      "step": 34029
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.571524524373249e-06,
+      "loss": 0.2021,
+      "step": 34030
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.571188689066884e-06,
+      "loss": 0.1838,
+      "step": 34031
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.570852853760517e-06,
+      "loss": 0.289,
+      "step": 34032
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.570517018454151e-06,
+      "loss": 0.1806,
+      "step": 34033
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.570181183147786e-06,
+      "loss": 0.3597,
+      "step": 34034
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.569845347841419e-06,
+      "loss": 0.2634,
+      "step": 34035
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.569509512535053e-06,
+      "loss": 0.267,
+      "step": 34036
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.569173677228688e-06,
+      "loss": 0.3359,
+      "step": 34037
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.568837841922321e-06,
+      "loss": 0.341,
+      "step": 34038
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.568502006615956e-06,
+      "loss": 0.2145,
+      "step": 34039
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.56816617130959e-06,
+      "loss": 0.4241,
+      "step": 34040
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.567830336003225e-06,
+      "loss": 0.2309,
+      "step": 34041
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.56749450069686e-06,
+      "loss": 0.344,
+      "step": 34042
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.567158665390494e-06,
+      "loss": 0.4657,
+      "step": 34043
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 8.566822830084127e-06,
+      "loss": 0.3689,
+      "step": 34044
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.566486994777762e-06,
+      "loss": 0.4023,
+      "step": 34045
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.566151159471396e-06,
+      "loss": 0.2427,
+      "step": 34046
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.56581532416503e-06,
+      "loss": 0.2854,
+      "step": 34047
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.565479488858664e-06,
+      "loss": 0.2018,
+      "step": 34048
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.565143653552299e-06,
+      "loss": 0.3247,
+      "step": 34049
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.564807818245933e-06,
+      "loss": 0.2535,
+      "step": 34050
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.564471982939568e-06,
+      "loss": 0.3628,
+      "step": 34051
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.564136147633203e-06,
+      "loss": 0.2595,
+      "step": 34052
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.563800312326836e-06,
+      "loss": 0.3973,
+      "step": 34053
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.56346447702047e-06,
+      "loss": 0.284,
+      "step": 34054
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.563128641714105e-06,
+      "loss": 0.3107,
+      "step": 34055
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.562792806407738e-06,
+      "loss": 0.3317,
+      "step": 34056
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.562456971101373e-06,
+      "loss": 0.5477,
+      "step": 34057
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.562121135795007e-06,
+      "loss": 0.5482,
+      "step": 34058
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.56178530048864e-06,
+      "loss": 0.2963,
+      "step": 34059
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.561449465182275e-06,
+      "loss": 0.2843,
+      "step": 34060
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.56111362987591e-06,
+      "loss": 0.2448,
+      "step": 34061
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.560777794569544e-06,
+      "loss": 0.3954,
+      "step": 34062
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.560441959263179e-06,
+      "loss": 0.2748,
+      "step": 34063
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.560106123956812e-06,
+      "loss": 0.302,
+      "step": 34064
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.559770288650446e-06,
+      "loss": 0.2031,
+      "step": 34065
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.559434453344081e-06,
+      "loss": 0.3227,
+      "step": 34066
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.559098618037714e-06,
+      "loss": 0.1952,
+      "step": 34067
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.558762782731349e-06,
+      "loss": 0.3066,
+      "step": 34068
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.558426947424983e-06,
+      "loss": 0.3542,
+      "step": 34069
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.558091112118618e-06,
+      "loss": 0.2088,
+      "step": 34070
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.557755276812253e-06,
+      "loss": 0.1827,
+      "step": 34071
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.557419441505887e-06,
+      "loss": 0.2774,
+      "step": 34072
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.55708360619952e-06,
+      "loss": 0.161,
+      "step": 34073
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.556747770893155e-06,
+      "loss": 0.3835,
+      "step": 34074
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.55641193558679e-06,
+      "loss": 0.3644,
+      "step": 34075
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.556076100280422e-06,
+      "loss": 0.2312,
+      "step": 34076
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.555740264974057e-06,
+      "loss": 0.2701,
+      "step": 34077
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.555404429667692e-06,
+      "loss": 0.2591,
+      "step": 34078
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.555068594361325e-06,
+      "loss": 0.2163,
+      "step": 34079
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.55473275905496e-06,
+      "loss": 0.4151,
+      "step": 34080
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.554396923748594e-06,
+      "loss": 0.3694,
+      "step": 34081
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.554061088442229e-06,
+      "loss": 0.4362,
+      "step": 34082
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.553725253135863e-06,
+      "loss": 0.3742,
+      "step": 34083
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.553389417829498e-06,
+      "loss": 0.2674,
+      "step": 34084
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.55305358252313e-06,
+      "loss": 0.3774,
+      "step": 34085
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.552717747216765e-06,
+      "loss": 0.3053,
+      "step": 34086
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.5523819119104e-06,
+      "loss": 0.3173,
+      "step": 34087
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.552046076604033e-06,
+      "loss": 0.3388,
+      "step": 34088
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.551710241297668e-06,
+      "loss": 0.2182,
+      "step": 34089
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.551374405991302e-06,
+      "loss": 0.309,
+      "step": 34090
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.551038570684937e-06,
+      "loss": 0.3944,
+      "step": 34091
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.550702735378572e-06,
+      "loss": 0.1812,
+      "step": 34092
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.550366900072206e-06,
+      "loss": 0.2618,
+      "step": 34093
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.55003106476584e-06,
+      "loss": 0.2306,
+      "step": 34094
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.549695229459474e-06,
+      "loss": 0.1951,
+      "step": 34095
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.549359394153109e-06,
+      "loss": 0.2023,
+      "step": 34096
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.549023558846741e-06,
+      "loss": 0.3206,
+      "step": 34097
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.548687723540376e-06,
+      "loss": 0.1308,
+      "step": 34098
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.54835188823401e-06,
+      "loss": 0.2628,
+      "step": 34099
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.548016052927644e-06,
+      "loss": 0.2081,
+      "step": 34100
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.547680217621278e-06,
+      "loss": 0.2548,
+      "step": 34101
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.547344382314913e-06,
+      "loss": 0.3451,
+      "step": 34102
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.547008547008548e-06,
+      "loss": 0.2878,
+      "step": 34103
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.546672711702182e-06,
+      "loss": 0.299,
+      "step": 34104
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.546336876395817e-06,
+      "loss": 0.4466,
+      "step": 34105
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.54600104108945e-06,
+      "loss": 0.3944,
+      "step": 34106
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.545665205783085e-06,
+      "loss": 0.1485,
+      "step": 34107
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.54532937047672e-06,
+      "loss": 0.3249,
+      "step": 34108
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.544993535170352e-06,
+      "loss": 0.3456,
+      "step": 34109
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.544657699863987e-06,
+      "loss": 0.3114,
+      "step": 34110
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.544321864557621e-06,
+      "loss": 0.3819,
+      "step": 34111
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.543986029251256e-06,
+      "loss": 0.2711,
+      "step": 34112
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.54365019394489e-06,
+      "loss": 0.3146,
+      "step": 34113
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.543314358638525e-06,
+      "loss": 0.3828,
+      "step": 34114
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.542978523332158e-06,
+      "loss": 0.1816,
+      "step": 34115
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.542642688025793e-06,
+      "loss": 0.2576,
+      "step": 34116
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.542306852719428e-06,
+      "loss": 0.2978,
+      "step": 34117
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.54197101741306e-06,
+      "loss": 0.3406,
+      "step": 34118
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.541635182106695e-06,
+      "loss": 0.332,
+      "step": 34119
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.54129934680033e-06,
+      "loss": 0.1246,
+      "step": 34120
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.540963511493965e-06,
+      "loss": 0.2954,
+      "step": 34121
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.5406276761876e-06,
+      "loss": 0.3522,
+      "step": 34122
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.540291840881232e-06,
+      "loss": 0.1753,
+      "step": 34123
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.539956005574867e-06,
+      "loss": 0.2347,
+      "step": 34124
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.539620170268501e-06,
+      "loss": 0.2756,
+      "step": 34125
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.539284334962136e-06,
+      "loss": 0.2509,
+      "step": 34126
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.538948499655769e-06,
+      "loss": 0.2067,
+      "step": 34127
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.538612664349404e-06,
+      "loss": 0.2337,
+      "step": 34128
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.538276829043038e-06,
+      "loss": 0.4599,
+      "step": 34129
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.537940993736671e-06,
+      "loss": 0.2098,
+      "step": 34130
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.537605158430306e-06,
+      "loss": 0.3161,
+      "step": 34131
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.53726932312394e-06,
+      "loss": 0.3538,
+      "step": 34132
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.536933487817575e-06,
+      "loss": 0.1943,
+      "step": 34133
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.53659765251121e-06,
+      "loss": 0.3202,
+      "step": 34134
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.536261817204845e-06,
+      "loss": 0.4014,
+      "step": 34135
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.535925981898477e-06,
+      "loss": 0.2967,
+      "step": 34136
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.535590146592112e-06,
+      "loss": 0.1686,
+      "step": 34137
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.535254311285747e-06,
+      "loss": 0.2648,
+      "step": 34138
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.53491847597938e-06,
+      "loss": 0.2056,
+      "step": 34139
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.534582640673014e-06,
+      "loss": 0.2952,
+      "step": 34140
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.534246805366649e-06,
+      "loss": 0.2188,
+      "step": 34141
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.533910970060284e-06,
+      "loss": 0.4053,
+      "step": 34142
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.533575134753918e-06,
+      "loss": 0.2385,
+      "step": 34143
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.533239299447553e-06,
+      "loss": 0.1937,
+      "step": 34144
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.532903464141186e-06,
+      "loss": 0.2323,
+      "step": 34145
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.53256762883482e-06,
+      "loss": 0.261,
+      "step": 34146
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.532231793528455e-06,
+      "loss": 0.3352,
+      "step": 34147
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.531895958222088e-06,
+      "loss": 0.1718,
+      "step": 34148
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.531560122915723e-06,
+      "loss": 0.3567,
+      "step": 34149
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.531224287609357e-06,
+      "loss": 0.3629,
+      "step": 34150
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.53088845230299e-06,
+      "loss": 0.264,
+      "step": 34151
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.530552616996625e-06,
+      "loss": 0.3639,
+      "step": 34152
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.53021678169026e-06,
+      "loss": 0.2141,
+      "step": 34153
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.529880946383894e-06,
+      "loss": 0.3068,
+      "step": 34154
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.529545111077529e-06,
+      "loss": 0.2373,
+      "step": 34155
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.529209275771162e-06,
+      "loss": 0.2429,
+      "step": 34156
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.528873440464797e-06,
+      "loss": 0.4924,
+      "step": 34157
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.528537605158431e-06,
+      "loss": 0.299,
+      "step": 34158
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.528201769852064e-06,
+      "loss": 0.3538,
+      "step": 34159
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.527865934545699e-06,
+      "loss": 0.2447,
+      "step": 34160
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.527530099239334e-06,
+      "loss": 0.1865,
+      "step": 34161
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.527194263932968e-06,
+      "loss": 0.2993,
+      "step": 34162
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.526858428626603e-06,
+      "loss": 0.1645,
+      "step": 34163
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.526522593320237e-06,
+      "loss": 0.29,
+      "step": 34164
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.52618675801387e-06,
+      "loss": 0.5387,
+      "step": 34165
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.525850922707505e-06,
+      "loss": 0.1947,
+      "step": 34166
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.52551508740114e-06,
+      "loss": 0.396,
+      "step": 34167
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.525179252094773e-06,
+      "loss": 0.2182,
+      "step": 34168
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.524843416788407e-06,
+      "loss": 0.3047,
+      "step": 34169
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.524507581482042e-06,
+      "loss": 0.4477,
+      "step": 34170
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.524171746175675e-06,
+      "loss": 0.2456,
+      "step": 34171
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.52383591086931e-06,
+      "loss": 0.2554,
+      "step": 34172
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.523500075562944e-06,
+      "loss": 0.3738,
+      "step": 34173
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.523164240256579e-06,
+      "loss": 0.3974,
+      "step": 34174
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.522828404950213e-06,
+      "loss": 0.4522,
+      "step": 34175
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.522492569643848e-06,
+      "loss": 0.417,
+      "step": 34176
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.522156734337481e-06,
+      "loss": 0.2598,
+      "step": 34177
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.521820899031116e-06,
+      "loss": 0.3423,
+      "step": 34178
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.52148506372475e-06,
+      "loss": 0.2789,
+      "step": 34179
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.521149228418383e-06,
+      "loss": 0.3347,
+      "step": 34180
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.520813393112018e-06,
+      "loss": 0.2406,
+      "step": 34181
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.520477557805653e-06,
+      "loss": 0.1602,
+      "step": 34182
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.520141722499287e-06,
+      "loss": 0.3672,
+      "step": 34183
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.519805887192922e-06,
+      "loss": 0.3291,
+      "step": 34184
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.519470051886557e-06,
+      "loss": 0.3319,
+      "step": 34185
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.51913421658019e-06,
+      "loss": 0.3775,
+      "step": 34186
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.518798381273824e-06,
+      "loss": 0.2064,
+      "step": 34187
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.518462545967459e-06,
+      "loss": 0.206,
+      "step": 34188
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.518126710661092e-06,
+      "loss": 0.2938,
+      "step": 34189
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.517790875354726e-06,
+      "loss": 0.2226,
+      "step": 34190
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.517455040048361e-06,
+      "loss": 0.2662,
+      "step": 34191
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.517119204741994e-06,
+      "loss": 0.1945,
+      "step": 34192
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.516783369435629e-06,
+      "loss": 0.1924,
+      "step": 34193
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.516447534129263e-06,
+      "loss": 0.1308,
+      "step": 34194
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.516111698822898e-06,
+      "loss": 0.2988,
+      "step": 34195
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.515775863516533e-06,
+      "loss": 0.3055,
+      "step": 34196
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.515440028210167e-06,
+      "loss": 0.3442,
+      "step": 34197
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.5151041929038e-06,
+      "loss": 0.29,
+      "step": 34198
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.514768357597435e-06,
+      "loss": 0.212,
+      "step": 34199
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.51443252229107e-06,
+      "loss": 0.2118,
+      "step": 34200
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.514096686984702e-06,
+      "loss": 0.2129,
+      "step": 34201
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.513760851678337e-06,
+      "loss": 0.2249,
+      "step": 34202
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.513425016371972e-06,
+      "loss": 0.3551,
+      "step": 34203
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.513089181065606e-06,
+      "loss": 0.1341,
+      "step": 34204
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.512753345759241e-06,
+      "loss": 0.3346,
+      "step": 34205
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.512417510452876e-06,
+      "loss": 0.3903,
+      "step": 34206
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.512081675146509e-06,
+      "loss": 0.0681,
+      "step": 34207
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.511745839840143e-06,
+      "loss": 0.2155,
+      "step": 34208
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.511410004533778e-06,
+      "loss": 0.243,
+      "step": 34209
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.511074169227411e-06,
+      "loss": 0.4918,
+      "step": 34210
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.510738333921046e-06,
+      "loss": 0.2335,
+      "step": 34211
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.51040249861468e-06,
+      "loss": 0.287,
+      "step": 34212
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.510066663308313e-06,
+      "loss": 0.2829,
+      "step": 34213
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.509730828001948e-06,
+      "loss": 0.3313,
+      "step": 34214
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.509394992695582e-06,
+      "loss": 0.3355,
+      "step": 34215
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.509059157389217e-06,
+      "loss": 0.2156,
+      "step": 34216
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.508723322082852e-06,
+      "loss": 0.2717,
+      "step": 34217
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.508387486776486e-06,
+      "loss": 0.4161,
+      "step": 34218
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.50805165147012e-06,
+      "loss": 0.3474,
+      "step": 34219
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.507715816163754e-06,
+      "loss": 0.3671,
+      "step": 34220
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.507379980857389e-06,
+      "loss": 0.4021,
+      "step": 34221
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.507044145551022e-06,
+      "loss": 0.4719,
+      "step": 34222
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.506708310244656e-06,
+      "loss": 0.3944,
+      "step": 34223
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.506372474938291e-06,
+      "loss": 0.2235,
+      "step": 34224
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.506036639631926e-06,
+      "loss": 0.2962,
+      "step": 34225
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.50570080432556e-06,
+      "loss": 0.2497,
+      "step": 34226
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.505364969019195e-06,
+      "loss": 0.3839,
+      "step": 34227
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.505029133712828e-06,
+      "loss": 0.3241,
+      "step": 34228
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.504693298406462e-06,
+      "loss": 0.3106,
+      "step": 34229
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.504357463100097e-06,
+      "loss": 0.2452,
+      "step": 34230
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.50402162779373e-06,
+      "loss": 0.2328,
+      "step": 34231
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.503685792487365e-06,
+      "loss": 0.3332,
+      "step": 34232
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.503349957181e-06,
+      "loss": 0.2236,
+      "step": 34233
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.503014121874634e-06,
+      "loss": 0.3187,
+      "step": 34234
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.502678286568267e-06,
+      "loss": 0.4085,
+      "step": 34235
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.502342451261902e-06,
+      "loss": 0.4114,
+      "step": 34236
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.502006615955536e-06,
+      "loss": 0.233,
+      "step": 34237
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.501670780649171e-06,
+      "loss": 0.2568,
+      "step": 34238
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.501334945342806e-06,
+      "loss": 0.4178,
+      "step": 34239
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.500999110036438e-06,
+      "loss": 0.2096,
+      "step": 34240
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.500663274730073e-06,
+      "loss": 0.3377,
+      "step": 34241
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 8.500327439423708e-06,
+      "loss": 0.3415,
+      "step": 34242
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.49999160411734e-06,
+      "loss": 0.372,
+      "step": 34243
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.499655768810975e-06,
+      "loss": 0.5024,
+      "step": 34244
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.49931993350461e-06,
+      "loss": 0.3072,
+      "step": 34245
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.498984098198245e-06,
+      "loss": 0.2507,
+      "step": 34246
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.49864826289188e-06,
+      "loss": 0.2966,
+      "step": 34247
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.498312427585512e-06,
+      "loss": 0.3635,
+      "step": 34248
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.497976592279147e-06,
+      "loss": 0.2795,
+      "step": 34249
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.497640756972782e-06,
+      "loss": 0.3044,
+      "step": 34250
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.497304921666415e-06,
+      "loss": 0.3575,
+      "step": 34251
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.496969086360049e-06,
+      "loss": 0.2479,
+      "step": 34252
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.496633251053684e-06,
+      "loss": 0.2977,
+      "step": 34253
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.496297415747318e-06,
+      "loss": 0.2891,
+      "step": 34254
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.495961580440953e-06,
+      "loss": 0.3055,
+      "step": 34255
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.495625745134588e-06,
+      "loss": 0.2096,
+      "step": 34256
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.49528990982822e-06,
+      "loss": 0.3397,
+      "step": 34257
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.494954074521855e-06,
+      "loss": 0.2584,
+      "step": 34258
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.49461823921549e-06,
+      "loss": 0.1915,
+      "step": 34259
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.494282403909123e-06,
+      "loss": 0.255,
+      "step": 34260
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.493946568602758e-06,
+      "loss": 0.2614,
+      "step": 34261
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.493610733296392e-06,
+      "loss": 0.4375,
+      "step": 34262
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.493274897990025e-06,
+      "loss": 0.2895,
+      "step": 34263
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.49293906268366e-06,
+      "loss": 0.2386,
+      "step": 34264
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.492603227377294e-06,
+      "loss": 0.3949,
+      "step": 34265
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.492267392070929e-06,
+      "loss": 0.4178,
+      "step": 34266
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.491931556764564e-06,
+      "loss": 0.2981,
+      "step": 34267
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.491595721458198e-06,
+      "loss": 0.3587,
+      "step": 34268
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.491259886151831e-06,
+      "loss": 0.2613,
+      "step": 34269
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.490924050845466e-06,
+      "loss": 0.4499,
+      "step": 34270
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.4905882155391e-06,
+      "loss": 0.3038,
+      "step": 34271
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.490252380232734e-06,
+      "loss": 0.1683,
+      "step": 34272
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.489916544926368e-06,
+      "loss": 0.142,
+      "step": 34273
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.489580709620003e-06,
+      "loss": 0.3711,
+      "step": 34274
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.489244874313638e-06,
+      "loss": 0.2827,
+      "step": 34275
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.488909039007272e-06,
+      "loss": 0.2954,
+      "step": 34276
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.488573203700907e-06,
+      "loss": 0.3423,
+      "step": 34277
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.48823736839454e-06,
+      "loss": 0.3881,
+      "step": 34278
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.487901533088174e-06,
+      "loss": 0.2894,
+      "step": 34279
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.487565697781809e-06,
+      "loss": 0.2644,
+      "step": 34280
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.487229862475442e-06,
+      "loss": 0.5369,
+      "step": 34281
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.486894027169077e-06,
+      "loss": 0.2939,
+      "step": 34282
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.486558191862711e-06,
+      "loss": 0.3414,
+      "step": 34283
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.486222356556344e-06,
+      "loss": 0.3068,
+      "step": 34284
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.485886521249979e-06,
+      "loss": 0.2701,
+      "step": 34285
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.485550685943614e-06,
+      "loss": 0.3011,
+      "step": 34286
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.485214850637248e-06,
+      "loss": 0.1569,
+      "step": 34287
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.484879015330883e-06,
+      "loss": 0.3771,
+      "step": 34288
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.484543180024518e-06,
+      "loss": 0.3295,
+      "step": 34289
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.48420734471815e-06,
+      "loss": 0.2167,
+      "step": 34290
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.483871509411785e-06,
+      "loss": 0.228,
+      "step": 34291
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.48353567410542e-06,
+      "loss": 0.192,
+      "step": 34292
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.483199838799053e-06,
+      "loss": 0.4193,
+      "step": 34293
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.482864003492687e-06,
+      "loss": 0.2025,
+      "step": 34294
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.482528168186322e-06,
+      "loss": 0.1746,
+      "step": 34295
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.482192332879957e-06,
+      "loss": 0.3763,
+      "step": 34296
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.481856497573591e-06,
+      "loss": 0.1765,
+      "step": 34297
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.481520662267226e-06,
+      "loss": 0.3548,
+      "step": 34298
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.481184826960859e-06,
+      "loss": 0.3395,
+      "step": 34299
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.480848991654494e-06,
+      "loss": 0.2392,
+      "step": 34300
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.480513156348128e-06,
+      "loss": 0.1348,
+      "step": 34301
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.480177321041761e-06,
+      "loss": 0.3965,
+      "step": 34302
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.479841485735396e-06,
+      "loss": 0.2238,
+      "step": 34303
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.47950565042903e-06,
+      "loss": 0.3136,
+      "step": 34304
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.479169815122663e-06,
+      "loss": 0.1849,
+      "step": 34305
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.478833979816298e-06,
+      "loss": 0.3073,
+      "step": 34306
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.478498144509933e-06,
+      "loss": 0.3068,
+      "step": 34307
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.478162309203567e-06,
+      "loss": 0.3894,
+      "step": 34308
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.477826473897202e-06,
+      "loss": 0.1794,
+      "step": 34309
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.477490638590837e-06,
+      "loss": 0.3688,
+      "step": 34310
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.47715480328447e-06,
+      "loss": 0.4077,
+      "step": 34311
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.476818967978104e-06,
+      "loss": 0.2688,
+      "step": 34312
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.476483132671739e-06,
+      "loss": 0.2748,
+      "step": 34313
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.476147297365372e-06,
+      "loss": 0.2207,
+      "step": 34314
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.475811462059007e-06,
+      "loss": 0.2938,
+      "step": 34315
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.475475626752641e-06,
+      "loss": 0.372,
+      "step": 34316
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.475139791446276e-06,
+      "loss": 0.2301,
+      "step": 34317
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.47480395613991e-06,
+      "loss": 0.3156,
+      "step": 34318
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.474468120833545e-06,
+      "loss": 0.359,
+      "step": 34319
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.474132285527178e-06,
+      "loss": 0.5121,
+      "step": 34320
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.473796450220813e-06,
+      "loss": 0.3239,
+      "step": 34321
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.473460614914447e-06,
+      "loss": 0.3164,
+      "step": 34322
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.47312477960808e-06,
+      "loss": 0.3257,
+      "step": 34323
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.472788944301715e-06,
+      "loss": 0.3354,
+      "step": 34324
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.47245310899535e-06,
+      "loss": 0.2266,
+      "step": 34325
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.472117273688983e-06,
+      "loss": 0.2525,
+      "step": 34326
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.471781438382617e-06,
+      "loss": 0.2657,
+      "step": 34327
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.471445603076252e-06,
+      "loss": 0.224,
+      "step": 34328
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.471109767769887e-06,
+      "loss": 0.423,
+      "step": 34329
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.470773932463521e-06,
+      "loss": 0.15,
+      "step": 34330
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.470438097157156e-06,
+      "loss": 0.3476,
+      "step": 34331
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.470102261850789e-06,
+      "loss": 0.3584,
+      "step": 34332
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.469766426544423e-06,
+      "loss": 0.3062,
+      "step": 34333
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.469430591238056e-06,
+      "loss": 0.3594,
+      "step": 34334
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.469094755931691e-06,
+      "loss": 0.4836,
+      "step": 34335
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.468758920625326e-06,
+      "loss": 0.2266,
+      "step": 34336
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.46842308531896e-06,
+      "loss": 0.2863,
+      "step": 34337
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.468087250012595e-06,
+      "loss": 0.3496,
+      "step": 34338
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.46775141470623e-06,
+      "loss": 0.1604,
+      "step": 34339
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.467415579399863e-06,
+      "loss": 0.2806,
+      "step": 34340
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.467079744093497e-06,
+      "loss": 0.3487,
+      "step": 34341
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.466743908787132e-06,
+      "loss": 0.3045,
+      "step": 34342
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.466408073480765e-06,
+      "loss": 0.2718,
+      "step": 34343
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.4660722381744e-06,
+      "loss": 0.1742,
+      "step": 34344
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.465736402868034e-06,
+      "loss": 0.2287,
+      "step": 34345
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.465400567561669e-06,
+      "loss": 0.4022,
+      "step": 34346
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.465064732255303e-06,
+      "loss": 0.144,
+      "step": 34347
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.464728896948936e-06,
+      "loss": 0.2789,
+      "step": 34348
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.464393061642571e-06,
+      "loss": 0.3292,
+      "step": 34349
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.464057226336206e-06,
+      "loss": 0.4428,
+      "step": 34350
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.46372139102984e-06,
+      "loss": 0.1744,
+      "step": 34351
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.463385555723473e-06,
+      "loss": 0.3408,
+      "step": 34352
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.463049720417108e-06,
+      "loss": 0.2759,
+      "step": 34353
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.462713885110743e-06,
+      "loss": 0.1773,
+      "step": 34354
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.462378049804375e-06,
+      "loss": 0.1674,
+      "step": 34355
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.46204221449801e-06,
+      "loss": 0.2899,
+      "step": 34356
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.461706379191645e-06,
+      "loss": 0.2093,
+      "step": 34357
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.46137054388528e-06,
+      "loss": 0.1788,
+      "step": 34358
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.461034708578914e-06,
+      "loss": 0.5331,
+      "step": 34359
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.460698873272549e-06,
+      "loss": 0.4497,
+      "step": 34360
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.460363037966182e-06,
+      "loss": 0.252,
+      "step": 34361
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.460027202659816e-06,
+      "loss": 0.3029,
+      "step": 34362
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.459691367353451e-06,
+      "loss": 0.3179,
+      "step": 34363
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.459355532047084e-06,
+      "loss": 0.409,
+      "step": 34364
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.459019696740719e-06,
+      "loss": 0.4316,
+      "step": 34365
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.458683861434353e-06,
+      "loss": 0.4475,
+      "step": 34366
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.458348026127988e-06,
+      "loss": 0.3106,
+      "step": 34367
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.458012190821623e-06,
+      "loss": 0.3425,
+      "step": 34368
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.457676355515257e-06,
+      "loss": 0.4084,
+      "step": 34369
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.45734052020889e-06,
+      "loss": 0.1715,
+      "step": 34370
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.457004684902525e-06,
+      "loss": 0.4566,
+      "step": 34371
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.45666884959616e-06,
+      "loss": 0.3533,
+      "step": 34372
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.456333014289792e-06,
+      "loss": 0.37,
+      "step": 34373
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.455997178983427e-06,
+      "loss": 0.3214,
+      "step": 34374
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.455661343677062e-06,
+      "loss": 0.255,
+      "step": 34375
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.455325508370695e-06,
+      "loss": 0.3863,
+      "step": 34376
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.45498967306433e-06,
+      "loss": 0.3739,
+      "step": 34377
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.454653837757964e-06,
+      "loss": 0.2908,
+      "step": 34378
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.454318002451599e-06,
+      "loss": 0.293,
+      "step": 34379
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.453982167145233e-06,
+      "loss": 0.2922,
+      "step": 34380
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.453646331838868e-06,
+      "loss": 0.2995,
+      "step": 34381
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.4533104965325e-06,
+      "loss": 0.2805,
+      "step": 34382
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.452974661226135e-06,
+      "loss": 0.2125,
+      "step": 34383
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.45263882591977e-06,
+      "loss": 0.2988,
+      "step": 34384
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.452302990613403e-06,
+      "loss": 0.3183,
+      "step": 34385
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.451967155307038e-06,
+      "loss": 0.396,
+      "step": 34386
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.451631320000672e-06,
+      "loss": 0.553,
+      "step": 34387
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.451295484694307e-06,
+      "loss": 0.2272,
+      "step": 34388
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.450959649387942e-06,
+      "loss": 0.2469,
+      "step": 34389
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.450623814081576e-06,
+      "loss": 0.2275,
+      "step": 34390
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.45028797877521e-06,
+      "loss": 0.3542,
+      "step": 34391
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.449952143468844e-06,
+      "loss": 0.3479,
+      "step": 34392
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.449616308162479e-06,
+      "loss": 0.3849,
+      "step": 34393
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.449280472856111e-06,
+      "loss": 0.288,
+      "step": 34394
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.448944637549746e-06,
+      "loss": 0.3794,
+      "step": 34395
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.44860880224338e-06,
+      "loss": 0.1699,
+      "step": 34396
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.448272966937014e-06,
+      "loss": 0.2119,
+      "step": 34397
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.447937131630648e-06,
+      "loss": 0.3048,
+      "step": 34398
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.447601296324283e-06,
+      "loss": 0.2768,
+      "step": 34399
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.447265461017918e-06,
+      "loss": 0.2586,
+      "step": 34400
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.446929625711552e-06,
+      "loss": 0.3032,
+      "step": 34401
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.446593790405187e-06,
+      "loss": 0.1911,
+      "step": 34402
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.44625795509882e-06,
+      "loss": 0.261,
+      "step": 34403
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.445922119792455e-06,
+      "loss": 0.5666,
+      "step": 34404
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.44558628448609e-06,
+      "loss": 0.3273,
+      "step": 34405
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.445250449179722e-06,
+      "loss": 0.1527,
+      "step": 34406
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.444914613873357e-06,
+      "loss": 0.2996,
+      "step": 34407
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.444578778566991e-06,
+      "loss": 0.4261,
+      "step": 34408
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.444242943260626e-06,
+      "loss": 0.3182,
+      "step": 34409
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.44390710795426e-06,
+      "loss": 0.3099,
+      "step": 34410
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.443571272647895e-06,
+      "loss": 0.3585,
+      "step": 34411
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.443235437341528e-06,
+      "loss": 0.3565,
+      "step": 34412
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.442899602035163e-06,
+      "loss": 0.1986,
+      "step": 34413
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.442563766728798e-06,
+      "loss": 0.4121,
+      "step": 34414
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.44222793142243e-06,
+      "loss": 0.2781,
+      "step": 34415
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.441892096116065e-06,
+      "loss": 0.3396,
+      "step": 34416
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.4415562608097e-06,
+      "loss": 0.3119,
+      "step": 34417
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.441220425503333e-06,
+      "loss": 0.3039,
+      "step": 34418
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.440884590196968e-06,
+      "loss": 0.2708,
+      "step": 34419
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.440548754890602e-06,
+      "loss": 0.212,
+      "step": 34420
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.440212919584237e-06,
+      "loss": 0.3143,
+      "step": 34421
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.439877084277871e-06,
+      "loss": 0.308,
+      "step": 34422
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.439541248971506e-06,
+      "loss": 0.2376,
+      "step": 34423
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.439205413665139e-06,
+      "loss": 0.3752,
+      "step": 34424
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.438869578358774e-06,
+      "loss": 0.2542,
+      "step": 34425
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.438533743052407e-06,
+      "loss": 0.2218,
+      "step": 34426
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.438197907746041e-06,
+      "loss": 0.3012,
+      "step": 34427
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.437862072439676e-06,
+      "loss": 0.2814,
+      "step": 34428
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.43752623713331e-06,
+      "loss": 0.2594,
+      "step": 34429
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.437190401826945e-06,
+      "loss": 0.2529,
+      "step": 34430
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.43685456652058e-06,
+      "loss": 0.3156,
+      "step": 34431
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.436518731214213e-06,
+      "loss": 0.2738,
+      "step": 34432
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.436182895907847e-06,
+      "loss": 0.3739,
+      "step": 34433
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.435847060601482e-06,
+      "loss": 0.327,
+      "step": 34434
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.435511225295115e-06,
+      "loss": 0.2748,
+      "step": 34435
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.43517538998875e-06,
+      "loss": 0.349,
+      "step": 34436
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.434839554682384e-06,
+      "loss": 0.3606,
+      "step": 34437
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.434503719376019e-06,
+      "loss": 0.1506,
+      "step": 34438
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.434167884069652e-06,
+      "loss": 0.2301,
+      "step": 34439
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.433832048763287e-06,
+      "loss": 0.252,
+      "step": 34440
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 8.433496213456921e-06,
+      "loss": 0.1517,
+      "step": 34441
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.433160378150556e-06,
+      "loss": 0.2075,
+      "step": 34442
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.43282454284419e-06,
+      "loss": 0.2793,
+      "step": 34443
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.432488707537824e-06,
+      "loss": 0.1592,
+      "step": 34444
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.432152872231458e-06,
+      "loss": 0.3205,
+      "step": 34445
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.431817036925093e-06,
+      "loss": 0.2221,
+      "step": 34446
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.431481201618726e-06,
+      "loss": 0.3133,
+      "step": 34447
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.43114536631236e-06,
+      "loss": 0.4114,
+      "step": 34448
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.430809531005995e-06,
+      "loss": 0.3093,
+      "step": 34449
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.43047369569963e-06,
+      "loss": 0.1513,
+      "step": 34450
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.430137860393264e-06,
+      "loss": 0.3812,
+      "step": 34451
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.429802025086899e-06,
+      "loss": 0.2305,
+      "step": 34452
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.429466189780532e-06,
+      "loss": 0.2887,
+      "step": 34453
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.429130354474167e-06,
+      "loss": 0.443,
+      "step": 34454
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.428794519167801e-06,
+      "loss": 0.3398,
+      "step": 34455
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.428458683861434e-06,
+      "loss": 0.1675,
+      "step": 34456
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.428122848555069e-06,
+      "loss": 0.251,
+      "step": 34457
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.427787013248704e-06,
+      "loss": 0.1841,
+      "step": 34458
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.427451177942338e-06,
+      "loss": 0.257,
+      "step": 34459
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.427115342635973e-06,
+      "loss": 0.2733,
+      "step": 34460
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.426779507329606e-06,
+      "loss": 0.2406,
+      "step": 34461
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.42644367202324e-06,
+      "loss": 0.3388,
+      "step": 34462
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.426107836716875e-06,
+      "loss": 0.3033,
+      "step": 34463
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.42577200141051e-06,
+      "loss": 0.3929,
+      "step": 34464
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.425436166104143e-06,
+      "loss": 0.1878,
+      "step": 34465
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.425100330797777e-06,
+      "loss": 0.2345,
+      "step": 34466
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.424764495491412e-06,
+      "loss": 0.4008,
+      "step": 34467
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.424428660185045e-06,
+      "loss": 0.4692,
+      "step": 34468
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.42409282487868e-06,
+      "loss": 0.2619,
+      "step": 34469
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.423756989572314e-06,
+      "loss": 0.2333,
+      "step": 34470
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.423421154265949e-06,
+      "loss": 0.3351,
+      "step": 34471
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.423085318959583e-06,
+      "loss": 0.2763,
+      "step": 34472
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.422749483653218e-06,
+      "loss": 0.2894,
+      "step": 34473
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.422413648346851e-06,
+      "loss": 0.4183,
+      "step": 34474
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.422077813040486e-06,
+      "loss": 0.1964,
+      "step": 34475
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.42174197773412e-06,
+      "loss": 0.2959,
+      "step": 34476
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.421406142427753e-06,
+      "loss": 0.3801,
+      "step": 34477
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.421070307121388e-06,
+      "loss": 0.2768,
+      "step": 34478
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.420734471815023e-06,
+      "loss": 0.2399,
+      "step": 34479
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.420398636508657e-06,
+      "loss": 0.3457,
+      "step": 34480
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.420062801202292e-06,
+      "loss": 0.2561,
+      "step": 34481
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.419726965895927e-06,
+      "loss": 0.2609,
+      "step": 34482
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.41939113058956e-06,
+      "loss": 0.1827,
+      "step": 34483
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.419055295283194e-06,
+      "loss": 0.1449,
+      "step": 34484
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.418719459976829e-06,
+      "loss": 0.397,
+      "step": 34485
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.418383624670462e-06,
+      "loss": 0.1955,
+      "step": 34486
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.418047789364096e-06,
+      "loss": 0.2398,
+      "step": 34487
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.417711954057731e-06,
+      "loss": 0.1816,
+      "step": 34488
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.417376118751364e-06,
+      "loss": 0.2269,
+      "step": 34489
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.417040283444999e-06,
+      "loss": 0.175,
+      "step": 34490
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.416704448138633e-06,
+      "loss": 0.3985,
+      "step": 34491
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.416368612832268e-06,
+      "loss": 0.1944,
+      "step": 34492
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.416032777525903e-06,
+      "loss": 0.2478,
+      "step": 34493
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.415696942219537e-06,
+      "loss": 0.3437,
+      "step": 34494
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.41536110691317e-06,
+      "loss": 0.1995,
+      "step": 34495
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.415025271606805e-06,
+      "loss": 0.3463,
+      "step": 34496
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.41468943630044e-06,
+      "loss": 0.2748,
+      "step": 34497
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.414353600994072e-06,
+      "loss": 0.2558,
+      "step": 34498
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.414017765687707e-06,
+      "loss": 0.1462,
+      "step": 34499
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.413681930381342e-06,
+      "loss": 0.2608,
+      "step": 34500
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.413346095074976e-06,
+      "loss": 0.4483,
+      "step": 34501
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.413010259768611e-06,
+      "loss": 0.122,
+      "step": 34502
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.412674424462246e-06,
+      "loss": 0.159,
+      "step": 34503
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.412338589155879e-06,
+      "loss": 0.2702,
+      "step": 34504
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.412002753849513e-06,
+      "loss": 0.2513,
+      "step": 34505
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.411666918543148e-06,
+      "loss": 0.1536,
+      "step": 34506
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.411331083236781e-06,
+      "loss": 0.2006,
+      "step": 34507
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.410995247930416e-06,
+      "loss": 0.3681,
+      "step": 34508
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.41065941262405e-06,
+      "loss": 0.4683,
+      "step": 34509
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.410323577317683e-06,
+      "loss": 0.3547,
+      "step": 34510
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.409987742011318e-06,
+      "loss": 0.2803,
+      "step": 34511
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.409651906704952e-06,
+      "loss": 0.2261,
+      "step": 34512
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.409316071398587e-06,
+      "loss": 0.2315,
+      "step": 34513
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.408980236092222e-06,
+      "loss": 0.4537,
+      "step": 34514
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.408644400785855e-06,
+      "loss": 0.3022,
+      "step": 34515
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.40830856547949e-06,
+      "loss": 0.2395,
+      "step": 34516
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.407972730173124e-06,
+      "loss": 0.2858,
+      "step": 34517
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.407636894866757e-06,
+      "loss": 0.3208,
+      "step": 34518
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.407301059560392e-06,
+      "loss": 0.4089,
+      "step": 34519
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.406965224254026e-06,
+      "loss": 0.3741,
+      "step": 34520
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.406629388947661e-06,
+      "loss": 0.3093,
+      "step": 34521
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.406293553641296e-06,
+      "loss": 0.3438,
+      "step": 34522
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.40595771833493e-06,
+      "loss": 0.2114,
+      "step": 34523
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.405621883028563e-06,
+      "loss": 0.1852,
+      "step": 34524
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.405286047722198e-06,
+      "loss": 0.3746,
+      "step": 34525
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.404950212415832e-06,
+      "loss": 0.2368,
+      "step": 34526
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.404614377109465e-06,
+      "loss": 0.2831,
+      "step": 34527
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.4042785418031e-06,
+      "loss": 0.2887,
+      "step": 34528
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.403942706496735e-06,
+      "loss": 0.2388,
+      "step": 34529
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.403606871190368e-06,
+      "loss": 0.3593,
+      "step": 34530
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.403271035884002e-06,
+      "loss": 0.1711,
+      "step": 34531
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.402935200577637e-06,
+      "loss": 0.2837,
+      "step": 34532
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.402599365271272e-06,
+      "loss": 0.2581,
+      "step": 34533
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.402263529964906e-06,
+      "loss": 0.2642,
+      "step": 34534
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.401927694658541e-06,
+      "loss": 0.3132,
+      "step": 34535
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.401591859352174e-06,
+      "loss": 0.341,
+      "step": 34536
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.401256024045808e-06,
+      "loss": 0.3649,
+      "step": 34537
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.400920188739443e-06,
+      "loss": 0.2387,
+      "step": 34538
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.400584353433076e-06,
+      "loss": 0.2691,
+      "step": 34539
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.40024851812671e-06,
+      "loss": 0.2026,
+      "step": 34540
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.399912682820345e-06,
+      "loss": 0.2479,
+      "step": 34541
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.39957684751398e-06,
+      "loss": 0.2455,
+      "step": 34542
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.399241012207615e-06,
+      "loss": 0.5781,
+      "step": 34543
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.39890517690125e-06,
+      "loss": 0.4016,
+      "step": 34544
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.398569341594882e-06,
+      "loss": 0.2494,
+      "step": 34545
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.398233506288517e-06,
+      "loss": 0.3065,
+      "step": 34546
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.397897670982152e-06,
+      "loss": 0.3264,
+      "step": 34547
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.397561835675785e-06,
+      "loss": 0.2525,
+      "step": 34548
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.397226000369419e-06,
+      "loss": 0.4106,
+      "step": 34549
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.396890165063054e-06,
+      "loss": 0.1563,
+      "step": 34550
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.396554329756688e-06,
+      "loss": 0.2916,
+      "step": 34551
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.396218494450321e-06,
+      "loss": 0.3991,
+      "step": 34552
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.395882659143956e-06,
+      "loss": 0.4111,
+      "step": 34553
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.39554682383759e-06,
+      "loss": 0.377,
+      "step": 34554
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.395210988531225e-06,
+      "loss": 0.3833,
+      "step": 34555
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.39487515322486e-06,
+      "loss": 0.2896,
+      "step": 34556
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.394539317918493e-06,
+      "loss": 0.3287,
+      "step": 34557
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.394203482612128e-06,
+      "loss": 0.3615,
+      "step": 34558
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.393867647305762e-06,
+      "loss": 0.2999,
+      "step": 34559
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.393531811999395e-06,
+      "loss": 0.4631,
+      "step": 34560
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.39319597669303e-06,
+      "loss": 0.2629,
+      "step": 34561
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.392860141386664e-06,
+      "loss": 0.2324,
+      "step": 34562
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.392524306080299e-06,
+      "loss": 0.2205,
+      "step": 34563
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.392188470773934e-06,
+      "loss": 0.219,
+      "step": 34564
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.391852635467568e-06,
+      "loss": 0.3255,
+      "step": 34565
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.391516800161201e-06,
+      "loss": 0.1761,
+      "step": 34566
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.391180964854836e-06,
+      "loss": 0.3369,
+      "step": 34567
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.39084512954847e-06,
+      "loss": 0.2317,
+      "step": 34568
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.390509294242104e-06,
+      "loss": 0.132,
+      "step": 34569
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.390173458935738e-06,
+      "loss": 0.2398,
+      "step": 34570
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.389837623629373e-06,
+      "loss": 0.2075,
+      "step": 34571
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.389501788323008e-06,
+      "loss": 0.231,
+      "step": 34572
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.389165953016642e-06,
+      "loss": 0.279,
+      "step": 34573
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.388830117710275e-06,
+      "loss": 0.2513,
+      "step": 34574
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.38849428240391e-06,
+      "loss": 0.3035,
+      "step": 34575
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.388158447097544e-06,
+      "loss": 0.2693,
+      "step": 34576
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.387822611791179e-06,
+      "loss": 0.351,
+      "step": 34577
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.387486776484812e-06,
+      "loss": 0.3656,
+      "step": 34578
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.387150941178447e-06,
+      "loss": 0.2487,
+      "step": 34579
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.386815105872081e-06,
+      "loss": 0.2522,
+      "step": 34580
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.386479270565714e-06,
+      "loss": 0.4865,
+      "step": 34581
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.386143435259349e-06,
+      "loss": 0.333,
+      "step": 34582
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.385807599952984e-06,
+      "loss": 0.458,
+      "step": 34583
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.385471764646618e-06,
+      "loss": 0.2781,
+      "step": 34584
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.385135929340253e-06,
+      "loss": 0.4262,
+      "step": 34585
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.384800094033888e-06,
+      "loss": 0.32,
+      "step": 34586
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.38446425872752e-06,
+      "loss": 0.1655,
+      "step": 34587
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.384128423421155e-06,
+      "loss": 0.2294,
+      "step": 34588
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.38379258811479e-06,
+      "loss": 0.3502,
+      "step": 34589
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.383456752808423e-06,
+      "loss": 0.509,
+      "step": 34590
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.383120917502057e-06,
+      "loss": 0.2033,
+      "step": 34591
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.382785082195692e-06,
+      "loss": 0.3495,
+      "step": 34592
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.382449246889327e-06,
+      "loss": 0.2722,
+      "step": 34593
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.382113411582961e-06,
+      "loss": 0.3494,
+      "step": 34594
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.381777576276596e-06,
+      "loss": 0.3092,
+      "step": 34595
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.381441740970229e-06,
+      "loss": 0.2402,
+      "step": 34596
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.381105905663864e-06,
+      "loss": 0.3758,
+      "step": 34597
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.380770070357498e-06,
+      "loss": 0.1884,
+      "step": 34598
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.380434235051131e-06,
+      "loss": 0.2427,
+      "step": 34599
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.380098399744766e-06,
+      "loss": 0.1936,
+      "step": 34600
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.3797625644384e-06,
+      "loss": 0.2114,
+      "step": 34601
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.379426729132033e-06,
+      "loss": 0.1919,
+      "step": 34602
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.379090893825668e-06,
+      "loss": 0.3494,
+      "step": 34603
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.378755058519303e-06,
+      "loss": 0.308,
+      "step": 34604
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.378419223212937e-06,
+      "loss": 0.5122,
+      "step": 34605
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.378083387906572e-06,
+      "loss": 0.2957,
+      "step": 34606
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.377747552600205e-06,
+      "loss": 0.3013,
+      "step": 34607
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.37741171729384e-06,
+      "loss": 0.1499,
+      "step": 34608
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.377075881987474e-06,
+      "loss": 0.2025,
+      "step": 34609
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.376740046681107e-06,
+      "loss": 0.3321,
+      "step": 34610
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.376404211374742e-06,
+      "loss": 0.1971,
+      "step": 34611
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.376068376068377e-06,
+      "loss": 0.2561,
+      "step": 34612
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.375732540762011e-06,
+      "loss": 0.3536,
+      "step": 34613
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.375396705455646e-06,
+      "loss": 0.4087,
+      "step": 34614
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.37506087014928e-06,
+      "loss": 0.2265,
+      "step": 34615
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.374725034842913e-06,
+      "loss": 0.253,
+      "step": 34616
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.374389199536548e-06,
+      "loss": 0.3337,
+      "step": 34617
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.374053364230183e-06,
+      "loss": 0.3438,
+      "step": 34618
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.373717528923816e-06,
+      "loss": 0.2324,
+      "step": 34619
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.37338169361745e-06,
+      "loss": 0.3164,
+      "step": 34620
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.373045858311085e-06,
+      "loss": 0.453,
+      "step": 34621
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.372710023004718e-06,
+      "loss": 0.226,
+      "step": 34622
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.372374187698353e-06,
+      "loss": 0.3184,
+      "step": 34623
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.372038352391987e-06,
+      "loss": 0.4276,
+      "step": 34624
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.371702517085622e-06,
+      "loss": 0.4233,
+      "step": 34625
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.371366681779257e-06,
+      "loss": 0.2688,
+      "step": 34626
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.371030846472891e-06,
+      "loss": 0.3562,
+      "step": 34627
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.370695011166524e-06,
+      "loss": 0.403,
+      "step": 34628
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.370359175860159e-06,
+      "loss": 0.3617,
+      "step": 34629
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.370023340553793e-06,
+      "loss": 0.3549,
+      "step": 34630
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.369687505247426e-06,
+      "loss": 0.2196,
+      "step": 34631
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.369351669941061e-06,
+      "loss": 0.4922,
+      "step": 34632
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.369015834634696e-06,
+      "loss": 0.144,
+      "step": 34633
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.36867999932833e-06,
+      "loss": 0.2715,
+      "step": 34634
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.368344164021965e-06,
+      "loss": 0.1721,
+      "step": 34635
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.3680083287156e-06,
+      "loss": 0.2757,
+      "step": 34636
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.367672493409233e-06,
+      "loss": 0.2841,
+      "step": 34637
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.367336658102867e-06,
+      "loss": 0.2889,
+      "step": 34638
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 8.367000822796502e-06,
+      "loss": 0.2412,
+      "step": 34639
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.366664987490135e-06,
+      "loss": 0.3569,
+      "step": 34640
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.36632915218377e-06,
+      "loss": 0.3179,
+      "step": 34641
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.365993316877404e-06,
+      "loss": 0.3659,
+      "step": 34642
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.365657481571037e-06,
+      "loss": 0.3033,
+      "step": 34643
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.365321646264672e-06,
+      "loss": 0.4648,
+      "step": 34644
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.364985810958306e-06,
+      "loss": 0.2396,
+      "step": 34645
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.364649975651941e-06,
+      "loss": 0.238,
+      "step": 34646
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.364314140345576e-06,
+      "loss": 0.3901,
+      "step": 34647
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.36397830503921e-06,
+      "loss": 0.1855,
+      "step": 34648
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.363642469732843e-06,
+      "loss": 0.2726,
+      "step": 34649
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.363306634426478e-06,
+      "loss": 0.3092,
+      "step": 34650
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.362970799120113e-06,
+      "loss": 0.3019,
+      "step": 34651
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.362634963813745e-06,
+      "loss": 0.2652,
+      "step": 34652
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.36229912850738e-06,
+      "loss": 0.2492,
+      "step": 34653
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.361963293201015e-06,
+      "loss": 0.2661,
+      "step": 34654
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.36162745789465e-06,
+      "loss": 0.3632,
+      "step": 34655
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.361291622588284e-06,
+      "loss": 0.266,
+      "step": 34656
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.360955787281919e-06,
+      "loss": 0.2732,
+      "step": 34657
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.360619951975552e-06,
+      "loss": 0.2012,
+      "step": 34658
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.360284116669186e-06,
+      "loss": 0.3147,
+      "step": 34659
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.359948281362821e-06,
+      "loss": 0.3459,
+      "step": 34660
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.359612446056454e-06,
+      "loss": 0.3771,
+      "step": 34661
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.359276610750089e-06,
+      "loss": 0.132,
+      "step": 34662
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.358940775443723e-06,
+      "loss": 0.2417,
+      "step": 34663
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.358604940137358e-06,
+      "loss": 0.2588,
+      "step": 34664
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.35826910483099e-06,
+      "loss": 0.1831,
+      "step": 34665
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.357933269524625e-06,
+      "loss": 0.3674,
+      "step": 34666
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.35759743421826e-06,
+      "loss": 0.168,
+      "step": 34667
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.357261598911895e-06,
+      "loss": 0.3455,
+      "step": 34668
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.35692576360553e-06,
+      "loss": 0.1615,
+      "step": 34669
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.356589928299162e-06,
+      "loss": 0.3915,
+      "step": 34670
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.356254092992797e-06,
+      "loss": 0.2344,
+      "step": 34671
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.355918257686432e-06,
+      "loss": 0.3682,
+      "step": 34672
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.355582422380065e-06,
+      "loss": 0.1802,
+      "step": 34673
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.3552465870737e-06,
+      "loss": 0.1497,
+      "step": 34674
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.354910751767334e-06,
+      "loss": 0.3744,
+      "step": 34675
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.354574916460969e-06,
+      "loss": 0.3709,
+      "step": 34676
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.354239081154603e-06,
+      "loss": 0.3106,
+      "step": 34677
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.353903245848238e-06,
+      "loss": 0.2275,
+      "step": 34678
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.35356741054187e-06,
+      "loss": 0.2866,
+      "step": 34679
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.353231575235505e-06,
+      "loss": 0.4022,
+      "step": 34680
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.35289573992914e-06,
+      "loss": 0.1705,
+      "step": 34681
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.352559904622773e-06,
+      "loss": 0.2813,
+      "step": 34682
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.352224069316408e-06,
+      "loss": 0.2696,
+      "step": 34683
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.351888234010042e-06,
+      "loss": 0.5804,
+      "step": 34684
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.351552398703677e-06,
+      "loss": 0.4065,
+      "step": 34685
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.351216563397312e-06,
+      "loss": 0.4019,
+      "step": 34686
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.350880728090945e-06,
+      "loss": 0.349,
+      "step": 34687
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.35054489278458e-06,
+      "loss": 0.2716,
+      "step": 34688
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.350209057478214e-06,
+      "loss": 0.1769,
+      "step": 34689
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.349873222171849e-06,
+      "loss": 0.3655,
+      "step": 34690
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.349537386865481e-06,
+      "loss": 0.4132,
+      "step": 34691
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.349201551559116e-06,
+      "loss": 0.3172,
+      "step": 34692
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.34886571625275e-06,
+      "loss": 0.2635,
+      "step": 34693
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.348529880946384e-06,
+      "loss": 0.1996,
+      "step": 34694
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.348194045640018e-06,
+      "loss": 0.1678,
+      "step": 34695
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.347858210333653e-06,
+      "loss": 0.2735,
+      "step": 34696
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.347522375027288e-06,
+      "loss": 0.3291,
+      "step": 34697
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.347186539720922e-06,
+      "loss": 0.2726,
+      "step": 34698
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.346850704414555e-06,
+      "loss": 0.281,
+      "step": 34699
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.34651486910819e-06,
+      "loss": 0.2587,
+      "step": 34700
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.346179033801825e-06,
+      "loss": 0.3381,
+      "step": 34701
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.345843198495458e-06,
+      "loss": 0.3316,
+      "step": 34702
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.345507363189092e-06,
+      "loss": 0.2929,
+      "step": 34703
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.345171527882727e-06,
+      "loss": 0.2854,
+      "step": 34704
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.344835692576361e-06,
+      "loss": 0.1249,
+      "step": 34705
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.344499857269996e-06,
+      "loss": 0.2082,
+      "step": 34706
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.34416402196363e-06,
+      "loss": 0.1343,
+      "step": 34707
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.343828186657264e-06,
+      "loss": 0.1984,
+      "step": 34708
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.343492351350898e-06,
+      "loss": 0.2398,
+      "step": 34709
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.343156516044533e-06,
+      "loss": 0.3831,
+      "step": 34710
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.342820680738166e-06,
+      "loss": 0.2062,
+      "step": 34711
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.3424848454318e-06,
+      "loss": 0.2002,
+      "step": 34712
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.342149010125435e-06,
+      "loss": 0.2806,
+      "step": 34713
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.341813174819068e-06,
+      "loss": 0.2601,
+      "step": 34714
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.341477339512703e-06,
+      "loss": 0.3096,
+      "step": 34715
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.341141504206337e-06,
+      "loss": 0.3118,
+      "step": 34716
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.340805668899972e-06,
+      "loss": 0.1935,
+      "step": 34717
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.340469833593607e-06,
+      "loss": 0.3768,
+      "step": 34718
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.340133998287241e-06,
+      "loss": 0.5785,
+      "step": 34719
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.339798162980874e-06,
+      "loss": 0.4172,
+      "step": 34720
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.339462327674509e-06,
+      "loss": 0.2578,
+      "step": 34721
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.339126492368144e-06,
+      "loss": 0.2706,
+      "step": 34722
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.338790657061777e-06,
+      "loss": 0.2949,
+      "step": 34723
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.338454821755411e-06,
+      "loss": 0.3193,
+      "step": 34724
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.338118986449046e-06,
+      "loss": 0.2977,
+      "step": 34725
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.33778315114268e-06,
+      "loss": 0.3243,
+      "step": 34726
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.337447315836315e-06,
+      "loss": 0.1859,
+      "step": 34727
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.33711148052995e-06,
+      "loss": 0.1749,
+      "step": 34728
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.336775645223583e-06,
+      "loss": 0.2939,
+      "step": 34729
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.336439809917217e-06,
+      "loss": 0.2577,
+      "step": 34730
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.336103974610852e-06,
+      "loss": 0.202,
+      "step": 34731
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.335768139304485e-06,
+      "loss": 0.3567,
+      "step": 34732
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.33543230399812e-06,
+      "loss": 0.5014,
+      "step": 34733
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.335096468691754e-06,
+      "loss": 0.502,
+      "step": 34734
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.334760633385387e-06,
+      "loss": 0.3106,
+      "step": 34735
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.334424798079022e-06,
+      "loss": 0.2839,
+      "step": 34736
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.334088962772657e-06,
+      "loss": 0.2249,
+      "step": 34737
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.333753127466291e-06,
+      "loss": 0.1809,
+      "step": 34738
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.333417292159926e-06,
+      "loss": 0.1481,
+      "step": 34739
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.33308145685356e-06,
+      "loss": 0.312,
+      "step": 34740
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.332745621547194e-06,
+      "loss": 0.3401,
+      "step": 34741
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.332409786240828e-06,
+      "loss": 0.336,
+      "step": 34742
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.332073950934463e-06,
+      "loss": 0.2367,
+      "step": 34743
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.331738115628096e-06,
+      "loss": 0.3755,
+      "step": 34744
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.33140228032173e-06,
+      "loss": 0.3052,
+      "step": 34745
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.331066445015365e-06,
+      "loss": 0.2999,
+      "step": 34746
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.330730609709e-06,
+      "loss": 0.2539,
+      "step": 34747
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.330394774402634e-06,
+      "loss": 0.4249,
+      "step": 34748
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.330058939096269e-06,
+      "loss": 0.2048,
+      "step": 34749
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.329723103789902e-06,
+      "loss": 0.1964,
+      "step": 34750
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.329387268483537e-06,
+      "loss": 0.4817,
+      "step": 34751
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.329051433177171e-06,
+      "loss": 0.3113,
+      "step": 34752
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.328715597870804e-06,
+      "loss": 0.4809,
+      "step": 34753
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.328379762564439e-06,
+      "loss": 0.1429,
+      "step": 34754
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.328043927258073e-06,
+      "loss": 0.2895,
+      "step": 34755
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.327708091951706e-06,
+      "loss": 0.3371,
+      "step": 34756
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.327372256645341e-06,
+      "loss": 0.2838,
+      "step": 34757
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.327036421338976e-06,
+      "loss": 0.3948,
+      "step": 34758
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.32670058603261e-06,
+      "loss": 0.2443,
+      "step": 34759
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.326364750726245e-06,
+      "loss": 0.3176,
+      "step": 34760
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.32602891541988e-06,
+      "loss": 0.2587,
+      "step": 34761
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.325693080113513e-06,
+      "loss": 0.2524,
+      "step": 34762
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.325357244807147e-06,
+      "loss": 0.3275,
+      "step": 34763
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.325021409500782e-06,
+      "loss": 0.1553,
+      "step": 34764
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.324685574194415e-06,
+      "loss": 0.4081,
+      "step": 34765
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.32434973888805e-06,
+      "loss": 0.3568,
+      "step": 34766
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.324013903581684e-06,
+      "loss": 0.3099,
+      "step": 34767
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.323678068275319e-06,
+      "loss": 0.3077,
+      "step": 34768
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.323342232968953e-06,
+      "loss": 0.3196,
+      "step": 34769
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.323006397662588e-06,
+      "loss": 0.4336,
+      "step": 34770
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.322670562356221e-06,
+      "loss": 0.263,
+      "step": 34771
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.322334727049856e-06,
+      "loss": 0.2565,
+      "step": 34772
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.32199889174349e-06,
+      "loss": 0.3119,
+      "step": 34773
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.321663056437123e-06,
+      "loss": 0.3253,
+      "step": 34774
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.321327221130758e-06,
+      "loss": 0.2957,
+      "step": 34775
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.320991385824393e-06,
+      "loss": 0.5218,
+      "step": 34776
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.320655550518027e-06,
+      "loss": 0.1915,
+      "step": 34777
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.32031971521166e-06,
+      "loss": 0.2384,
+      "step": 34778
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.319983879905295e-06,
+      "loss": 0.3116,
+      "step": 34779
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.31964804459893e-06,
+      "loss": 0.2472,
+      "step": 34780
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.319312209292564e-06,
+      "loss": 0.3254,
+      "step": 34781
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.318976373986199e-06,
+      "loss": 0.5152,
+      "step": 34782
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.318640538679832e-06,
+      "loss": 0.3316,
+      "step": 34783
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.318304703373466e-06,
+      "loss": 0.1977,
+      "step": 34784
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.3179688680671e-06,
+      "loss": 0.3722,
+      "step": 34785
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.317633032760734e-06,
+      "loss": 0.4756,
+      "step": 34786
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.317297197454369e-06,
+      "loss": 0.4683,
+      "step": 34787
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.316961362148003e-06,
+      "loss": 0.3063,
+      "step": 34788
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.316625526841638e-06,
+      "loss": 0.3732,
+      "step": 34789
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.316289691535273e-06,
+      "loss": 0.228,
+      "step": 34790
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.315953856228906e-06,
+      "loss": 0.2209,
+      "step": 34791
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.31561802092254e-06,
+      "loss": 0.2009,
+      "step": 34792
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.315282185616175e-06,
+      "loss": 0.2788,
+      "step": 34793
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.314946350309808e-06,
+      "loss": 0.2211,
+      "step": 34794
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.314610515003442e-06,
+      "loss": 0.3506,
+      "step": 34795
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.314274679697077e-06,
+      "loss": 0.2918,
+      "step": 34796
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.313938844390712e-06,
+      "loss": 0.4055,
+      "step": 34797
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.313603009084346e-06,
+      "loss": 0.2215,
+      "step": 34798
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.313267173777981e-06,
+      "loss": 0.2667,
+      "step": 34799
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.312931338471614e-06,
+      "loss": 0.2298,
+      "step": 34800
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.312595503165249e-06,
+      "loss": 0.2449,
+      "step": 34801
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.312259667858883e-06,
+      "loss": 0.4302,
+      "step": 34802
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.311923832552516e-06,
+      "loss": 0.4159,
+      "step": 34803
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.311587997246151e-06,
+      "loss": 0.2717,
+      "step": 34804
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.311252161939786e-06,
+      "loss": 0.3248,
+      "step": 34805
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.310916326633418e-06,
+      "loss": 0.298,
+      "step": 34806
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.310580491327053e-06,
+      "loss": 0.2124,
+      "step": 34807
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.310244656020688e-06,
+      "loss": 0.2251,
+      "step": 34808
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.309908820714322e-06,
+      "loss": 0.4316,
+      "step": 34809
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.309572985407957e-06,
+      "loss": 0.3255,
+      "step": 34810
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.309237150101592e-06,
+      "loss": 0.4549,
+      "step": 34811
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.308901314795225e-06,
+      "loss": 0.3272,
+      "step": 34812
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.30856547948886e-06,
+      "loss": 0.2406,
+      "step": 34813
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.308229644182494e-06,
+      "loss": 0.4202,
+      "step": 34814
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.307893808876127e-06,
+      "loss": 0.2775,
+      "step": 34815
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.307557973569762e-06,
+      "loss": 0.2678,
+      "step": 34816
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.307222138263396e-06,
+      "loss": 0.425,
+      "step": 34817
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.306886302957031e-06,
+      "loss": 0.2014,
+      "step": 34818
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.306550467650666e-06,
+      "loss": 0.153,
+      "step": 34819
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.3062146323443e-06,
+      "loss": 0.2787,
+      "step": 34820
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.305878797037933e-06,
+      "loss": 0.4061,
+      "step": 34821
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.305542961731568e-06,
+      "loss": 0.2362,
+      "step": 34822
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.305207126425202e-06,
+      "loss": 0.1319,
+      "step": 34823
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.304871291118835e-06,
+      "loss": 0.3106,
+      "step": 34824
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.30453545581247e-06,
+      "loss": 0.2011,
+      "step": 34825
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.304199620506105e-06,
+      "loss": 0.3484,
+      "step": 34826
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.303863785199738e-06,
+      "loss": 0.2705,
+      "step": 34827
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.303527949893372e-06,
+      "loss": 0.3296,
+      "step": 34828
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.303192114587007e-06,
+      "loss": 0.3829,
+      "step": 34829
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.302856279280642e-06,
+      "loss": 0.2224,
+      "step": 34830
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.302520443974276e-06,
+      "loss": 0.2193,
+      "step": 34831
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.302184608667911e-06,
+      "loss": 0.2596,
+      "step": 34832
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.301848773361544e-06,
+      "loss": 0.1521,
+      "step": 34833
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.301512938055178e-06,
+      "loss": 0.2622,
+      "step": 34834
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.301177102748813e-06,
+      "loss": 0.2809,
+      "step": 34835
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.300841267442446e-06,
+      "loss": 0.3707,
+      "step": 34836
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.30050543213608e-06,
+      "loss": 0.3984,
+      "step": 34837
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 8.300169596829715e-06,
+      "loss": 0.2052,
+      "step": 34838
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.29983376152335e-06,
+      "loss": 0.2546,
+      "step": 34839
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.299497926216985e-06,
+      "loss": 0.3499,
+      "step": 34840
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.29916209091062e-06,
+      "loss": 0.4086,
+      "step": 34841
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.298826255604252e-06,
+      "loss": 0.2706,
+      "step": 34842
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.298490420297887e-06,
+      "loss": 0.33,
+      "step": 34843
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.298154584991522e-06,
+      "loss": 0.2809,
+      "step": 34844
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.297818749685154e-06,
+      "loss": 0.1341,
+      "step": 34845
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.297482914378789e-06,
+      "loss": 0.2401,
+      "step": 34846
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.297147079072424e-06,
+      "loss": 0.2996,
+      "step": 34847
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.296811243766057e-06,
+      "loss": 0.251,
+      "step": 34848
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.296475408459691e-06,
+      "loss": 0.31,
+      "step": 34849
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.296139573153326e-06,
+      "loss": 0.2805,
+      "step": 34850
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.29580373784696e-06,
+      "loss": 0.3396,
+      "step": 34851
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.295467902540595e-06,
+      "loss": 0.4498,
+      "step": 34852
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.29513206723423e-06,
+      "loss": 0.3849,
+      "step": 34853
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.294796231927863e-06,
+      "loss": 0.2588,
+      "step": 34854
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.294460396621498e-06,
+      "loss": 0.3897,
+      "step": 34855
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.294124561315132e-06,
+      "loss": 0.2502,
+      "step": 34856
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.293788726008765e-06,
+      "loss": 0.3957,
+      "step": 34857
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.2934528907024e-06,
+      "loss": 0.2657,
+      "step": 34858
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.293117055396034e-06,
+      "loss": 0.2109,
+      "step": 34859
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.292781220089669e-06,
+      "loss": 0.3469,
+      "step": 34860
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.292445384783304e-06,
+      "loss": 0.3059,
+      "step": 34861
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.292109549476938e-06,
+      "loss": 0.2606,
+      "step": 34862
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.291773714170571e-06,
+      "loss": 0.4474,
+      "step": 34863
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.291437878864206e-06,
+      "loss": 0.3991,
+      "step": 34864
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.29110204355784e-06,
+      "loss": 0.2918,
+      "step": 34865
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.290766208251474e-06,
+      "loss": 0.2734,
+      "step": 34866
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.290430372945108e-06,
+      "loss": 0.2946,
+      "step": 34867
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.290094537638743e-06,
+      "loss": 0.2215,
+      "step": 34868
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.289758702332376e-06,
+      "loss": 0.309,
+      "step": 34869
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.28942286702601e-06,
+      "loss": 0.2325,
+      "step": 34870
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.289087031719645e-06,
+      "loss": 0.2022,
+      "step": 34871
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.28875119641328e-06,
+      "loss": 0.1889,
+      "step": 34872
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.288415361106914e-06,
+      "loss": 0.2493,
+      "step": 34873
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.288079525800549e-06,
+      "loss": 0.1822,
+      "step": 34874
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.287743690494182e-06,
+      "loss": 0.3471,
+      "step": 34875
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.287407855187817e-06,
+      "loss": 0.3048,
+      "step": 34876
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.28707201988145e-06,
+      "loss": 0.3286,
+      "step": 34877
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.286736184575084e-06,
+      "loss": 0.2467,
+      "step": 34878
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.286400349268719e-06,
+      "loss": 0.3706,
+      "step": 34879
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.286064513962354e-06,
+      "loss": 0.3424,
+      "step": 34880
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.285728678655988e-06,
+      "loss": 0.2554,
+      "step": 34881
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.285392843349623e-06,
+      "loss": 0.2244,
+      "step": 34882
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.285057008043256e-06,
+      "loss": 0.1984,
+      "step": 34883
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.28472117273689e-06,
+      "loss": 0.3232,
+      "step": 34884
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.284385337430525e-06,
+      "loss": 0.4882,
+      "step": 34885
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.284049502124158e-06,
+      "loss": 0.3866,
+      "step": 34886
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.283713666817793e-06,
+      "loss": 0.3512,
+      "step": 34887
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.283377831511427e-06,
+      "loss": 0.4699,
+      "step": 34888
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.283041996205062e-06,
+      "loss": 0.1822,
+      "step": 34889
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.282706160898695e-06,
+      "loss": 0.3613,
+      "step": 34890
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.28237032559233e-06,
+      "loss": 0.4156,
+      "step": 34891
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.282034490285964e-06,
+      "loss": 0.2751,
+      "step": 34892
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.281698654979599e-06,
+      "loss": 0.3631,
+      "step": 34893
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.281362819673234e-06,
+      "loss": 0.3676,
+      "step": 34894
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.281026984366867e-06,
+      "loss": 0.3119,
+      "step": 34895
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.280691149060501e-06,
+      "loss": 0.3033,
+      "step": 34896
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.280355313754136e-06,
+      "loss": 0.4878,
+      "step": 34897
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.280019478447769e-06,
+      "loss": 0.2356,
+      "step": 34898
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.279683643141403e-06,
+      "loss": 0.2217,
+      "step": 34899
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.279347807835038e-06,
+      "loss": 0.1551,
+      "step": 34900
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.279011972528673e-06,
+      "loss": 0.3581,
+      "step": 34901
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.278676137222307e-06,
+      "loss": 0.1884,
+      "step": 34902
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.278340301915942e-06,
+      "loss": 0.221,
+      "step": 34903
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.278004466609575e-06,
+      "loss": 0.2695,
+      "step": 34904
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.27766863130321e-06,
+      "loss": 0.3746,
+      "step": 34905
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.277332795996844e-06,
+      "loss": 0.3616,
+      "step": 34906
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.276996960690477e-06,
+      "loss": 0.3245,
+      "step": 34907
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.276661125384112e-06,
+      "loss": 0.2375,
+      "step": 34908
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.276325290077747e-06,
+      "loss": 0.201,
+      "step": 34909
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.275989454771381e-06,
+      "loss": 0.1906,
+      "step": 34910
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.275653619465016e-06,
+      "loss": 0.146,
+      "step": 34911
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.27531778415865e-06,
+      "loss": 0.472,
+      "step": 34912
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.274981948852283e-06,
+      "loss": 0.3402,
+      "step": 34913
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.274646113545918e-06,
+      "loss": 0.2895,
+      "step": 34914
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.274310278239553e-06,
+      "loss": 0.4663,
+      "step": 34915
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.273974442933186e-06,
+      "loss": 0.1889,
+      "step": 34916
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.27363860762682e-06,
+      "loss": 0.1924,
+      "step": 34917
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.273302772320455e-06,
+      "loss": 0.4477,
+      "step": 34918
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.272966937014088e-06,
+      "loss": 0.2601,
+      "step": 34919
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.272631101707723e-06,
+      "loss": 0.2137,
+      "step": 34920
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.272295266401357e-06,
+      "loss": 0.2322,
+      "step": 34921
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.271959431094992e-06,
+      "loss": 0.2353,
+      "step": 34922
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.271623595788626e-06,
+      "loss": 0.4305,
+      "step": 34923
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.271287760482261e-06,
+      "loss": 0.4726,
+      "step": 34924
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.270951925175894e-06,
+      "loss": 0.2331,
+      "step": 34925
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.270616089869529e-06,
+      "loss": 0.2413,
+      "step": 34926
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.270280254563163e-06,
+      "loss": 0.2512,
+      "step": 34927
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.269944419256796e-06,
+      "loss": 0.2995,
+      "step": 34928
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.269608583950431e-06,
+      "loss": 0.2881,
+      "step": 34929
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.269272748644066e-06,
+      "loss": 0.2564,
+      "step": 34930
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.2689369133377e-06,
+      "loss": 0.3726,
+      "step": 34931
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.268601078031335e-06,
+      "loss": 0.2957,
+      "step": 34932
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.26826524272497e-06,
+      "loss": 0.3523,
+      "step": 34933
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.267929407418603e-06,
+      "loss": 0.2427,
+      "step": 34934
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.267593572112237e-06,
+      "loss": 0.1643,
+      "step": 34935
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.267257736805872e-06,
+      "loss": 0.3218,
+      "step": 34936
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.266921901499505e-06,
+      "loss": 0.3212,
+      "step": 34937
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.26658606619314e-06,
+      "loss": 0.1721,
+      "step": 34938
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.266250230886774e-06,
+      "loss": 0.3263,
+      "step": 34939
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.265914395580407e-06,
+      "loss": 0.3531,
+      "step": 34940
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.265578560274042e-06,
+      "loss": 0.2815,
+      "step": 34941
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.265242724967676e-06,
+      "loss": 0.2398,
+      "step": 34942
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.264906889661311e-06,
+      "loss": 0.3301,
+      "step": 34943
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.264571054354946e-06,
+      "loss": 0.4475,
+      "step": 34944
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.26423521904858e-06,
+      "loss": 0.4174,
+      "step": 34945
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.263899383742213e-06,
+      "loss": 0.3326,
+      "step": 34946
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.263563548435848e-06,
+      "loss": 0.3064,
+      "step": 34947
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.263227713129483e-06,
+      "loss": 0.1483,
+      "step": 34948
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.262891877823115e-06,
+      "loss": 0.3397,
+      "step": 34949
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.26255604251675e-06,
+      "loss": 0.3201,
+      "step": 34950
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.262220207210385e-06,
+      "loss": 0.2631,
+      "step": 34951
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.26188437190402e-06,
+      "loss": 0.2307,
+      "step": 34952
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.261548536597654e-06,
+      "loss": 0.3074,
+      "step": 34953
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.261212701291289e-06,
+      "loss": 0.1645,
+      "step": 34954
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.260876865984922e-06,
+      "loss": 0.5962,
+      "step": 34955
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.260541030678556e-06,
+      "loss": 0.2854,
+      "step": 34956
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.260205195372191e-06,
+      "loss": 0.4499,
+      "step": 34957
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.259869360065824e-06,
+      "loss": 0.2579,
+      "step": 34958
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.259533524759459e-06,
+      "loss": 0.27,
+      "step": 34959
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.259197689453093e-06,
+      "loss": 0.3515,
+      "step": 34960
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.258861854146726e-06,
+      "loss": 0.2638,
+      "step": 34961
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.25852601884036e-06,
+      "loss": 0.1742,
+      "step": 34962
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.258190183533995e-06,
+      "loss": 0.2414,
+      "step": 34963
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.25785434822763e-06,
+      "loss": 0.2632,
+      "step": 34964
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.257518512921265e-06,
+      "loss": 0.3375,
+      "step": 34965
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.257182677614898e-06,
+      "loss": 0.3824,
+      "step": 34966
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.256846842308532e-06,
+      "loss": 0.2104,
+      "step": 34967
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.256511007002167e-06,
+      "loss": 0.3426,
+      "step": 34968
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.2561751716958e-06,
+      "loss": 0.2339,
+      "step": 34969
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.255839336389435e-06,
+      "loss": 0.3759,
+      "step": 34970
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.25550350108307e-06,
+      "loss": 0.4784,
+      "step": 34971
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.255167665776704e-06,
+      "loss": 0.1988,
+      "step": 34972
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.254831830470339e-06,
+      "loss": 0.2868,
+      "step": 34973
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.254495995163973e-06,
+      "loss": 0.3283,
+      "step": 34974
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.254160159857606e-06,
+      "loss": 0.3358,
+      "step": 34975
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.25382432455124e-06,
+      "loss": 0.1643,
+      "step": 34976
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.253488489244875e-06,
+      "loss": 0.2381,
+      "step": 34977
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.253152653938508e-06,
+      "loss": 0.2499,
+      "step": 34978
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.252816818632143e-06,
+      "loss": 0.5266,
+      "step": 34979
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.252480983325778e-06,
+      "loss": 0.3021,
+      "step": 34980
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.25214514801941e-06,
+      "loss": 0.3376,
+      "step": 34981
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.251809312713045e-06,
+      "loss": 0.2895,
+      "step": 34982
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.25147347740668e-06,
+      "loss": 0.1895,
+      "step": 34983
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.251137642100315e-06,
+      "loss": 0.255,
+      "step": 34984
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.25080180679395e-06,
+      "loss": 0.3725,
+      "step": 34985
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.250465971487584e-06,
+      "loss": 0.2778,
+      "step": 34986
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.250130136181217e-06,
+      "loss": 0.3791,
+      "step": 34987
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.249794300874851e-06,
+      "loss": 0.2843,
+      "step": 34988
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.249458465568486e-06,
+      "loss": 0.4078,
+      "step": 34989
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.249122630262119e-06,
+      "loss": 0.2306,
+      "step": 34990
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.248786794955754e-06,
+      "loss": 0.2227,
+      "step": 34991
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.248450959649388e-06,
+      "loss": 0.3024,
+      "step": 34992
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.248115124343023e-06,
+      "loss": 0.2168,
+      "step": 34993
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.247779289036658e-06,
+      "loss": 0.1579,
+      "step": 34994
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.247443453730292e-06,
+      "loss": 0.2724,
+      "step": 34995
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.247107618423925e-06,
+      "loss": 0.1806,
+      "step": 34996
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.24677178311756e-06,
+      "loss": 0.3147,
+      "step": 34997
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.246435947811195e-06,
+      "loss": 0.2881,
+      "step": 34998
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.246100112504828e-06,
+      "loss": 0.2068,
+      "step": 34999
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.245764277198462e-06,
+      "loss": 0.3327,
+      "step": 35000
+    },
+    {
+      "epoch": 1.76,
+      "eval_accuracy": 0.8839544810055907,
+      "eval_f1_microaverage": 0.8839544810055907,
+      "eval_loss": 0.30590540170669556,
+      "eval_precision_macroaverage": 0.7615655494978313,
+      "eval_precision_microaverage": 0.8839544810055907,
+      "eval_recall_macroaverage": 0.6744712539089281,
+      "eval_recall_microaverage": 0.8839544810055907,
+      "eval_runtime": 1172.903,
+      "eval_samples_per_second": 232.106,
+      "eval_steps_per_second": 3.627,
+      "step": 35000
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.245428441892097e-06,
+      "loss": 0.2933,
+      "step": 35001
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.245092606585731e-06,
+      "loss": 0.187,
+      "step": 35002
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.244756771279364e-06,
+      "loss": 0.3879,
+      "step": 35003
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.244420935972999e-06,
+      "loss": 0.2828,
+      "step": 35004
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.244085100666634e-06,
+      "loss": 0.2425,
+      "step": 35005
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.243749265360268e-06,
+      "loss": 0.4088,
+      "step": 35006
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.243413430053903e-06,
+      "loss": 0.3079,
+      "step": 35007
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.243077594747536e-06,
+      "loss": 0.2537,
+      "step": 35008
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.24274175944117e-06,
+      "loss": 0.2537,
+      "step": 35009
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.242405924134805e-06,
+      "loss": 0.3191,
+      "step": 35010
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.242070088828438e-06,
+      "loss": 0.2153,
+      "step": 35011
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.241734253522073e-06,
+      "loss": 0.2427,
+      "step": 35012
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.241398418215707e-06,
+      "loss": 0.2806,
+      "step": 35013
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.241062582909342e-06,
+      "loss": 0.2903,
+      "step": 35014
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.240726747602977e-06,
+      "loss": 0.2661,
+      "step": 35015
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.240390912296611e-06,
+      "loss": 0.1681,
+      "step": 35016
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.240055076990244e-06,
+      "loss": 0.4106,
+      "step": 35017
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.239719241683879e-06,
+      "loss": 0.3093,
+      "step": 35018
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.239383406377514e-06,
+      "loss": 0.3603,
+      "step": 35019
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.239047571071147e-06,
+      "loss": 0.3823,
+      "step": 35020
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.238711735764781e-06,
+      "loss": 0.3391,
+      "step": 35021
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.238375900458416e-06,
+      "loss": 0.2713,
+      "step": 35022
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.23804006515205e-06,
+      "loss": 0.2516,
+      "step": 35023
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.237704229845685e-06,
+      "loss": 0.2563,
+      "step": 35024
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.237368394539318e-06,
+      "loss": 0.3405,
+      "step": 35025
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.237032559232953e-06,
+      "loss": 0.2919,
+      "step": 35026
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.236696723926587e-06,
+      "loss": 0.2886,
+      "step": 35027
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.236360888620222e-06,
+      "loss": 0.4024,
+      "step": 35028
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.236025053313855e-06,
+      "loss": 0.2405,
+      "step": 35029
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.23568921800749e-06,
+      "loss": 0.4366,
+      "step": 35030
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.235353382701124e-06,
+      "loss": 0.1794,
+      "step": 35031
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.235017547394757e-06,
+      "loss": 0.2719,
+      "step": 35032
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.234681712088392e-06,
+      "loss": 0.211,
+      "step": 35033
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.234345876782027e-06,
+      "loss": 0.2992,
+      "step": 35034
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.234010041475661e-06,
+      "loss": 0.3317,
+      "step": 35035
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.233674206169296e-06,
+      "loss": 0.3055,
+      "step": 35036
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 8.23333837086293e-06,
+      "loss": 0.2389,
+      "step": 35037
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.233002535556564e-06,
+      "loss": 0.3389,
+      "step": 35038
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.232666700250198e-06,
+      "loss": 0.1995,
+      "step": 35039
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.232330864943833e-06,
+      "loss": 0.3496,
+      "step": 35040
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.231995029637466e-06,
+      "loss": 0.2971,
+      "step": 35041
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.2316591943311e-06,
+      "loss": 0.2192,
+      "step": 35042
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.231323359024735e-06,
+      "loss": 0.206,
+      "step": 35043
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.23098752371837e-06,
+      "loss": 0.2509,
+      "step": 35044
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.230651688412004e-06,
+      "loss": 0.2911,
+      "step": 35045
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.230315853105639e-06,
+      "loss": 0.2232,
+      "step": 35046
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.229980017799272e-06,
+      "loss": 0.3429,
+      "step": 35047
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.229644182492907e-06,
+      "loss": 0.4522,
+      "step": 35048
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.229308347186541e-06,
+      "loss": 0.3583,
+      "step": 35049
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.228972511880174e-06,
+      "loss": 0.2801,
+      "step": 35050
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.228636676573809e-06,
+      "loss": 0.349,
+      "step": 35051
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.228300841267443e-06,
+      "loss": 0.3123,
+      "step": 35052
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.227965005961076e-06,
+      "loss": 0.34,
+      "step": 35053
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.227629170654711e-06,
+      "loss": 0.3283,
+      "step": 35054
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.227293335348346e-06,
+      "loss": 0.3711,
+      "step": 35055
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.22695750004198e-06,
+      "loss": 0.2587,
+      "step": 35056
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.226621664735615e-06,
+      "loss": 0.3669,
+      "step": 35057
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.226285829429248e-06,
+      "loss": 0.172,
+      "step": 35058
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.225949994122883e-06,
+      "loss": 0.3032,
+      "step": 35059
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.225614158816517e-06,
+      "loss": 0.339,
+      "step": 35060
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.22527832351015e-06,
+      "loss": 0.254,
+      "step": 35061
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.224942488203785e-06,
+      "loss": 0.2568,
+      "step": 35062
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.22460665289742e-06,
+      "loss": 0.398,
+      "step": 35063
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.224270817591054e-06,
+      "loss": 0.329,
+      "step": 35064
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.223934982284689e-06,
+      "loss": 0.248,
+      "step": 35065
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.223599146978323e-06,
+      "loss": 0.2734,
+      "step": 35066
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.223263311671956e-06,
+      "loss": 0.3564,
+      "step": 35067
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.222927476365591e-06,
+      "loss": 0.207,
+      "step": 35068
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.222591641059226e-06,
+      "loss": 0.2552,
+      "step": 35069
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.222255805752859e-06,
+      "loss": 0.297,
+      "step": 35070
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.221919970446493e-06,
+      "loss": 0.3576,
+      "step": 35071
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.221584135140128e-06,
+      "loss": 0.3541,
+      "step": 35072
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.221248299833761e-06,
+      "loss": 0.3942,
+      "step": 35073
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.220912464527396e-06,
+      "loss": 0.2231,
+      "step": 35074
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.22057662922103e-06,
+      "loss": 0.1854,
+      "step": 35075
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.220240793914665e-06,
+      "loss": 0.4165,
+      "step": 35076
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.2199049586083e-06,
+      "loss": 0.2938,
+      "step": 35077
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.219569123301934e-06,
+      "loss": 0.2444,
+      "step": 35078
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.219233287995567e-06,
+      "loss": 0.3296,
+      "step": 35079
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.218897452689202e-06,
+      "loss": 0.3849,
+      "step": 35080
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.218561617382836e-06,
+      "loss": 0.264,
+      "step": 35081
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.21822578207647e-06,
+      "loss": 0.3078,
+      "step": 35082
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.217889946770104e-06,
+      "loss": 0.5477,
+      "step": 35083
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.217554111463739e-06,
+      "loss": 0.3188,
+      "step": 35084
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.217218276157373e-06,
+      "loss": 0.2654,
+      "step": 35085
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.216882440851008e-06,
+      "loss": 0.1934,
+      "step": 35086
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.216546605544643e-06,
+      "loss": 0.3624,
+      "step": 35087
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.216210770238276e-06,
+      "loss": 0.3785,
+      "step": 35088
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.21587493493191e-06,
+      "loss": 0.2565,
+      "step": 35089
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.215539099625545e-06,
+      "loss": 0.3053,
+      "step": 35090
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.215203264319178e-06,
+      "loss": 0.2714,
+      "step": 35091
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.214867429012812e-06,
+      "loss": 0.4741,
+      "step": 35092
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.214531593706447e-06,
+      "loss": 0.2703,
+      "step": 35093
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.21419575840008e-06,
+      "loss": 0.1683,
+      "step": 35094
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.213859923093715e-06,
+      "loss": 0.3495,
+      "step": 35095
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.21352408778735e-06,
+      "loss": 0.1987,
+      "step": 35096
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.213188252480984e-06,
+      "loss": 0.2826,
+      "step": 35097
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.212852417174619e-06,
+      "loss": 0.2617,
+      "step": 35098
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.212516581868253e-06,
+      "loss": 0.3817,
+      "step": 35099
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.212180746561886e-06,
+      "loss": 0.3834,
+      "step": 35100
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.211844911255521e-06,
+      "loss": 0.3789,
+      "step": 35101
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.211509075949156e-06,
+      "loss": 0.2765,
+      "step": 35102
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.211173240642788e-06,
+      "loss": 0.2968,
+      "step": 35103
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.210837405336423e-06,
+      "loss": 0.2869,
+      "step": 35104
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.210501570030058e-06,
+      "loss": 0.3175,
+      "step": 35105
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.210165734723692e-06,
+      "loss": 0.3257,
+      "step": 35106
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.209829899417327e-06,
+      "loss": 0.2456,
+      "step": 35107
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.209494064110962e-06,
+      "loss": 0.3652,
+      "step": 35108
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.209158228804595e-06,
+      "loss": 0.4483,
+      "step": 35109
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.20882239349823e-06,
+      "loss": 0.3006,
+      "step": 35110
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.208486558191864e-06,
+      "loss": 0.3076,
+      "step": 35111
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.208150722885497e-06,
+      "loss": 0.3729,
+      "step": 35112
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.207814887579132e-06,
+      "loss": 0.2027,
+      "step": 35113
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.207479052272766e-06,
+      "loss": 0.2062,
+      "step": 35114
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.207143216966401e-06,
+      "loss": 0.3104,
+      "step": 35115
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.206807381660034e-06,
+      "loss": 0.3225,
+      "step": 35116
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.206471546353668e-06,
+      "loss": 0.2983,
+      "step": 35117
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.206135711047303e-06,
+      "loss": 0.2179,
+      "step": 35118
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.205799875740938e-06,
+      "loss": 0.3197,
+      "step": 35119
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.205464040434572e-06,
+      "loss": 0.2064,
+      "step": 35120
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.205128205128205e-06,
+      "loss": 0.1665,
+      "step": 35121
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.20479236982184e-06,
+      "loss": 0.2912,
+      "step": 35122
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.204456534515475e-06,
+      "loss": 0.5026,
+      "step": 35123
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.204120699209108e-06,
+      "loss": 0.3166,
+      "step": 35124
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.203784863902742e-06,
+      "loss": 0.2403,
+      "step": 35125
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.203449028596377e-06,
+      "loss": 0.2057,
+      "step": 35126
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.203113193290012e-06,
+      "loss": 0.4359,
+      "step": 35127
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.202777357983646e-06,
+      "loss": 0.1861,
+      "step": 35128
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.202441522677281e-06,
+      "loss": 0.2813,
+      "step": 35129
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.202105687370914e-06,
+      "loss": 0.3321,
+      "step": 35130
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.201769852064548e-06,
+      "loss": 0.5061,
+      "step": 35131
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.201434016758183e-06,
+      "loss": 0.2675,
+      "step": 35132
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.201098181451816e-06,
+      "loss": 0.1683,
+      "step": 35133
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.20076234614545e-06,
+      "loss": 0.3514,
+      "step": 35134
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.200426510839085e-06,
+      "loss": 0.3239,
+      "step": 35135
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.20009067553272e-06,
+      "loss": 0.2737,
+      "step": 35136
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.199754840226355e-06,
+      "loss": 0.2673,
+      "step": 35137
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.199419004919988e-06,
+      "loss": 0.2549,
+      "step": 35138
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.199083169613622e-06,
+      "loss": 0.4011,
+      "step": 35139
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.198747334307257e-06,
+      "loss": 0.5477,
+      "step": 35140
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.198411499000892e-06,
+      "loss": 0.165,
+      "step": 35141
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.198075663694524e-06,
+      "loss": 0.2686,
+      "step": 35142
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.197739828388159e-06,
+      "loss": 0.2003,
+      "step": 35143
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.197403993081794e-06,
+      "loss": 0.2764,
+      "step": 35144
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.197068157775427e-06,
+      "loss": 0.4079,
+      "step": 35145
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.196732322469061e-06,
+      "loss": 0.284,
+      "step": 35146
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.196396487162696e-06,
+      "loss": 0.2305,
+      "step": 35147
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.19606065185633e-06,
+      "loss": 0.4261,
+      "step": 35148
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.195724816549965e-06,
+      "loss": 0.167,
+      "step": 35149
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.195388981243598e-06,
+      "loss": 0.3354,
+      "step": 35150
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.195053145937233e-06,
+      "loss": 0.2429,
+      "step": 35151
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.194717310630868e-06,
+      "loss": 0.162,
+      "step": 35152
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.1943814753245e-06,
+      "loss": 0.2482,
+      "step": 35153
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.194045640018135e-06,
+      "loss": 0.2509,
+      "step": 35154
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.19370980471177e-06,
+      "loss": 0.3299,
+      "step": 35155
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.193373969405404e-06,
+      "loss": 0.4682,
+      "step": 35156
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.193038134099039e-06,
+      "loss": 0.2056,
+      "step": 35157
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.192702298792674e-06,
+      "loss": 0.4139,
+      "step": 35158
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.192366463486307e-06,
+      "loss": 0.268,
+      "step": 35159
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.192030628179941e-06,
+      "loss": 0.3145,
+      "step": 35160
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.191694792873576e-06,
+      "loss": 0.2483,
+      "step": 35161
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.191358957567209e-06,
+      "loss": 0.1753,
+      "step": 35162
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.191023122260844e-06,
+      "loss": 0.2033,
+      "step": 35163
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.190687286954478e-06,
+      "loss": 0.2118,
+      "step": 35164
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.190351451648111e-06,
+      "loss": 0.3269,
+      "step": 35165
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.190015616341746e-06,
+      "loss": 0.3432,
+      "step": 35166
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.18967978103538e-06,
+      "loss": 0.2435,
+      "step": 35167
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.189343945729015e-06,
+      "loss": 0.2848,
+      "step": 35168
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.18900811042265e-06,
+      "loss": 0.2334,
+      "step": 35169
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.188672275116284e-06,
+      "loss": 0.1835,
+      "step": 35170
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.188336439809917e-06,
+      "loss": 0.2816,
+      "step": 35171
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.188000604503552e-06,
+      "loss": 0.2386,
+      "step": 35172
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.187664769197187e-06,
+      "loss": 0.3408,
+      "step": 35173
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.18732893389082e-06,
+      "loss": 0.2925,
+      "step": 35174
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.186993098584454e-06,
+      "loss": 0.2108,
+      "step": 35175
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.186657263278089e-06,
+      "loss": 0.1776,
+      "step": 35176
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.186321427971724e-06,
+      "loss": 0.1689,
+      "step": 35177
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.185985592665358e-06,
+      "loss": 0.2381,
+      "step": 35178
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.185649757358993e-06,
+      "loss": 0.246,
+      "step": 35179
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.185313922052626e-06,
+      "loss": 0.3769,
+      "step": 35180
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.18497808674626e-06,
+      "loss": 0.4422,
+      "step": 35181
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.184642251439895e-06,
+      "loss": 0.2559,
+      "step": 35182
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.184306416133528e-06,
+      "loss": 0.2737,
+      "step": 35183
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.183970580827163e-06,
+      "loss": 0.1766,
+      "step": 35184
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.183634745520797e-06,
+      "loss": 0.295,
+      "step": 35185
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.18329891021443e-06,
+      "loss": 0.1871,
+      "step": 35186
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.182963074908065e-06,
+      "loss": 0.2815,
+      "step": 35187
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.1826272396017e-06,
+      "loss": 0.2259,
+      "step": 35188
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.182291404295334e-06,
+      "loss": 0.3377,
+      "step": 35189
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.181955568988969e-06,
+      "loss": 0.283,
+      "step": 35190
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.181619733682604e-06,
+      "loss": 0.3681,
+      "step": 35191
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.181283898376237e-06,
+      "loss": 0.1886,
+      "step": 35192
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.180948063069871e-06,
+      "loss": 0.3252,
+      "step": 35193
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.180612227763506e-06,
+      "loss": 0.2836,
+      "step": 35194
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.180276392457139e-06,
+      "loss": 0.3329,
+      "step": 35195
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.179940557150773e-06,
+      "loss": 0.3848,
+      "step": 35196
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.179604721844408e-06,
+      "loss": 0.2799,
+      "step": 35197
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.179268886538043e-06,
+      "loss": 0.1617,
+      "step": 35198
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.178933051231677e-06,
+      "loss": 0.1462,
+      "step": 35199
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.178597215925312e-06,
+      "loss": 0.3096,
+      "step": 35200
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.178261380618945e-06,
+      "loss": 0.295,
+      "step": 35201
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.17792554531258e-06,
+      "loss": 0.4509,
+      "step": 35202
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.177589710006214e-06,
+      "loss": 0.2448,
+      "step": 35203
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.177253874699847e-06,
+      "loss": 0.2862,
+      "step": 35204
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.176918039393482e-06,
+      "loss": 0.2848,
+      "step": 35205
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.176582204087117e-06,
+      "loss": 0.3579,
+      "step": 35206
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.17624636878075e-06,
+      "loss": 0.4389,
+      "step": 35207
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.175910533474384e-06,
+      "loss": 0.4509,
+      "step": 35208
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.175574698168019e-06,
+      "loss": 0.3127,
+      "step": 35209
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.175238862861653e-06,
+      "loss": 0.2338,
+      "step": 35210
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.174903027555288e-06,
+      "loss": 0.31,
+      "step": 35211
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.174567192248923e-06,
+      "loss": 0.2444,
+      "step": 35212
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.174231356942556e-06,
+      "loss": 0.1976,
+      "step": 35213
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.17389552163619e-06,
+      "loss": 0.3188,
+      "step": 35214
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.173559686329825e-06,
+      "loss": 0.3568,
+      "step": 35215
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.173223851023458e-06,
+      "loss": 0.282,
+      "step": 35216
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.172888015717093e-06,
+      "loss": 0.3611,
+      "step": 35217
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.172552180410727e-06,
+      "loss": 0.2047,
+      "step": 35218
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.172216345104362e-06,
+      "loss": 0.1513,
+      "step": 35219
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.171880509797996e-06,
+      "loss": 0.3826,
+      "step": 35220
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.171544674491631e-06,
+      "loss": 0.2711,
+      "step": 35221
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.171208839185264e-06,
+      "loss": 0.2781,
+      "step": 35222
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.170873003878899e-06,
+      "loss": 0.3122,
+      "step": 35223
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.170537168572533e-06,
+      "loss": 0.2892,
+      "step": 35224
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.170201333266166e-06,
+      "loss": 0.1619,
+      "step": 35225
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.169865497959801e-06,
+      "loss": 0.6282,
+      "step": 35226
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.169529662653436e-06,
+      "loss": 0.2751,
+      "step": 35227
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.16919382734707e-06,
+      "loss": 0.2547,
+      "step": 35228
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.168857992040703e-06,
+      "loss": 0.4698,
+      "step": 35229
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.168522156734338e-06,
+      "loss": 0.4009,
+      "step": 35230
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.168186321427973e-06,
+      "loss": 0.3523,
+      "step": 35231
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.167850486121607e-06,
+      "loss": 0.4205,
+      "step": 35232
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.167514650815242e-06,
+      "loss": 0.235,
+      "step": 35233
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.167178815508875e-06,
+      "loss": 0.1493,
+      "step": 35234
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 8.16684298020251e-06,
+      "loss": 0.1942,
+      "step": 35235
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.166507144896142e-06,
+      "loss": 0.2877,
+      "step": 35236
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.166171309589777e-06,
+      "loss": 0.3708,
+      "step": 35237
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.165835474283412e-06,
+      "loss": 0.2045,
+      "step": 35238
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.165499638977046e-06,
+      "loss": 0.2826,
+      "step": 35239
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.165163803670681e-06,
+      "loss": 0.1235,
+      "step": 35240
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.164827968364316e-06,
+      "loss": 0.297,
+      "step": 35241
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.164492133057949e-06,
+      "loss": 0.3156,
+      "step": 35242
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.164156297751583e-06,
+      "loss": 0.246,
+      "step": 35243
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.163820462445218e-06,
+      "loss": 0.1672,
+      "step": 35244
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.16348462713885e-06,
+      "loss": 0.2018,
+      "step": 35245
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.163148791832485e-06,
+      "loss": 0.2471,
+      "step": 35246
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.16281295652612e-06,
+      "loss": 0.2906,
+      "step": 35247
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.162477121219755e-06,
+      "loss": 0.3146,
+      "step": 35248
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.16214128591339e-06,
+      "loss": 0.3884,
+      "step": 35249
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.161805450607024e-06,
+      "loss": 0.3106,
+      "step": 35250
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.161469615300657e-06,
+      "loss": 0.1884,
+      "step": 35251
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.161133779994292e-06,
+      "loss": 0.4563,
+      "step": 35252
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.160797944687926e-06,
+      "loss": 0.3298,
+      "step": 35253
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.16046210938156e-06,
+      "loss": 0.17,
+      "step": 35254
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.160126274075194e-06,
+      "loss": 0.426,
+      "step": 35255
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.159790438768829e-06,
+      "loss": 0.3593,
+      "step": 35256
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.159454603462462e-06,
+      "loss": 0.2323,
+      "step": 35257
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.159118768156096e-06,
+      "loss": 0.303,
+      "step": 35258
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.15878293284973e-06,
+      "loss": 0.2028,
+      "step": 35259
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.158447097543365e-06,
+      "loss": 0.3067,
+      "step": 35260
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.158111262237e-06,
+      "loss": 0.2911,
+      "step": 35261
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.157775426930635e-06,
+      "loss": 0.1529,
+      "step": 35262
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.157439591624268e-06,
+      "loss": 0.2693,
+      "step": 35263
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.157103756317902e-06,
+      "loss": 0.3591,
+      "step": 35264
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.156767921011537e-06,
+      "loss": 0.2714,
+      "step": 35265
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.15643208570517e-06,
+      "loss": 0.5264,
+      "step": 35266
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.156096250398805e-06,
+      "loss": 0.1906,
+      "step": 35267
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.15576041509244e-06,
+      "loss": 0.3224,
+      "step": 35268
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.155424579786074e-06,
+      "loss": 0.3093,
+      "step": 35269
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.155088744479709e-06,
+      "loss": 0.3067,
+      "step": 35270
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.154752909173343e-06,
+      "loss": 0.1997,
+      "step": 35271
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.154417073866976e-06,
+      "loss": 0.1823,
+      "step": 35272
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.15408123856061e-06,
+      "loss": 0.2122,
+      "step": 35273
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.153745403254245e-06,
+      "loss": 0.1675,
+      "step": 35274
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.153409567947878e-06,
+      "loss": 0.3688,
+      "step": 35275
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.153073732641513e-06,
+      "loss": 0.1826,
+      "step": 35276
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.152737897335148e-06,
+      "loss": 0.1894,
+      "step": 35277
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.15240206202878e-06,
+      "loss": 0.3665,
+      "step": 35278
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.152066226722415e-06,
+      "loss": 0.4666,
+      "step": 35279
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.15173039141605e-06,
+      "loss": 0.285,
+      "step": 35280
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.151394556109685e-06,
+      "loss": 0.3595,
+      "step": 35281
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.15105872080332e-06,
+      "loss": 0.2356,
+      "step": 35282
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.150722885496954e-06,
+      "loss": 0.2195,
+      "step": 35283
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.150387050190587e-06,
+      "loss": 0.1294,
+      "step": 35284
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.150051214884221e-06,
+      "loss": 0.2667,
+      "step": 35285
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.149715379577856e-06,
+      "loss": 0.3654,
+      "step": 35286
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.149379544271489e-06,
+      "loss": 0.4554,
+      "step": 35287
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.149043708965124e-06,
+      "loss": 0.2836,
+      "step": 35288
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.148707873658758e-06,
+      "loss": 0.2388,
+      "step": 35289
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.148372038352393e-06,
+      "loss": 0.1238,
+      "step": 35290
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.148036203046028e-06,
+      "loss": 0.3203,
+      "step": 35291
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.147700367739662e-06,
+      "loss": 0.4224,
+      "step": 35292
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.147364532433295e-06,
+      "loss": 0.3222,
+      "step": 35293
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.14702869712693e-06,
+      "loss": 0.2621,
+      "step": 35294
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.146692861820565e-06,
+      "loss": 0.2082,
+      "step": 35295
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.146357026514198e-06,
+      "loss": 0.4075,
+      "step": 35296
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.146021191207832e-06,
+      "loss": 0.4064,
+      "step": 35297
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.145685355901467e-06,
+      "loss": 0.3082,
+      "step": 35298
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.1453495205951e-06,
+      "loss": 0.3071,
+      "step": 35299
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.145013685288734e-06,
+      "loss": 0.2437,
+      "step": 35300
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.144677849982369e-06,
+      "loss": 0.3615,
+      "step": 35301
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.144342014676004e-06,
+      "loss": 0.3831,
+      "step": 35302
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.144006179369638e-06,
+      "loss": 0.2501,
+      "step": 35303
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.143670344063273e-06,
+      "loss": 0.2355,
+      "step": 35304
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.143334508756906e-06,
+      "loss": 0.2405,
+      "step": 35305
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.14299867345054e-06,
+      "loss": 0.1905,
+      "step": 35306
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.142662838144175e-06,
+      "loss": 0.2577,
+      "step": 35307
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.142327002837808e-06,
+      "loss": 0.4284,
+      "step": 35308
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.141991167531443e-06,
+      "loss": 0.3218,
+      "step": 35309
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.141655332225077e-06,
+      "loss": 0.4565,
+      "step": 35310
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.141319496918712e-06,
+      "loss": 0.2083,
+      "step": 35311
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.140983661612347e-06,
+      "loss": 0.4029,
+      "step": 35312
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.140647826305981e-06,
+      "loss": 0.2261,
+      "step": 35313
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.140311990999614e-06,
+      "loss": 0.3902,
+      "step": 35314
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.139976155693249e-06,
+      "loss": 0.2262,
+      "step": 35315
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.139640320386884e-06,
+      "loss": 0.2526,
+      "step": 35316
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.139304485080517e-06,
+      "loss": 0.3424,
+      "step": 35317
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.138968649774151e-06,
+      "loss": 0.3093,
+      "step": 35318
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.138632814467786e-06,
+      "loss": 0.18,
+      "step": 35319
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.138296979161419e-06,
+      "loss": 0.2076,
+      "step": 35320
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.137961143855054e-06,
+      "loss": 0.259,
+      "step": 35321
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.137625308548688e-06,
+      "loss": 0.2292,
+      "step": 35322
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.137289473242323e-06,
+      "loss": 0.3166,
+      "step": 35323
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.136953637935957e-06,
+      "loss": 0.3188,
+      "step": 35324
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.136617802629592e-06,
+      "loss": 0.4986,
+      "step": 35325
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.136281967323225e-06,
+      "loss": 0.2032,
+      "step": 35326
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.13594613201686e-06,
+      "loss": 0.162,
+      "step": 35327
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.135610296710493e-06,
+      "loss": 0.2821,
+      "step": 35328
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.135274461404127e-06,
+      "loss": 0.2081,
+      "step": 35329
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.134938626097762e-06,
+      "loss": 0.3861,
+      "step": 35330
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.134602790791397e-06,
+      "loss": 0.3229,
+      "step": 35331
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.134266955485031e-06,
+      "loss": 0.2687,
+      "step": 35332
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.133931120178666e-06,
+      "loss": 0.1897,
+      "step": 35333
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.133595284872299e-06,
+      "loss": 0.2643,
+      "step": 35334
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.133259449565934e-06,
+      "loss": 0.3543,
+      "step": 35335
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.132923614259568e-06,
+      "loss": 0.235,
+      "step": 35336
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.132587778953201e-06,
+      "loss": 0.302,
+      "step": 35337
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.132251943646836e-06,
+      "loss": 0.1671,
+      "step": 35338
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.13191610834047e-06,
+      "loss": 0.3028,
+      "step": 35339
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.131580273034105e-06,
+      "loss": 0.3748,
+      "step": 35340
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.13124443772774e-06,
+      "loss": 0.219,
+      "step": 35341
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.130908602421373e-06,
+      "loss": 0.4906,
+      "step": 35342
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.130572767115007e-06,
+      "loss": 0.3105,
+      "step": 35343
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.130236931808642e-06,
+      "loss": 0.2414,
+      "step": 35344
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.129901096502277e-06,
+      "loss": 0.404,
+      "step": 35345
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.12956526119591e-06,
+      "loss": 0.3512,
+      "step": 35346
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.129229425889544e-06,
+      "loss": 0.3583,
+      "step": 35347
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.128893590583179e-06,
+      "loss": 0.2859,
+      "step": 35348
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.128557755276812e-06,
+      "loss": 0.5484,
+      "step": 35349
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.128221919970446e-06,
+      "loss": 0.3419,
+      "step": 35350
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.127886084664081e-06,
+      "loss": 0.2308,
+      "step": 35351
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.127550249357716e-06,
+      "loss": 0.3705,
+      "step": 35352
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.12721441405135e-06,
+      "loss": 0.1107,
+      "step": 35353
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.126878578744985e-06,
+      "loss": 0.196,
+      "step": 35354
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.126542743438618e-06,
+      "loss": 0.359,
+      "step": 35355
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.126206908132253e-06,
+      "loss": 0.2701,
+      "step": 35356
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.125871072825887e-06,
+      "loss": 0.2323,
+      "step": 35357
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.12553523751952e-06,
+      "loss": 0.1647,
+      "step": 35358
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.125199402213155e-06,
+      "loss": 0.1896,
+      "step": 35359
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.12486356690679e-06,
+      "loss": 0.3262,
+      "step": 35360
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.124527731600424e-06,
+      "loss": 0.1933,
+      "step": 35361
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.124191896294059e-06,
+      "loss": 0.2985,
+      "step": 35362
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.123856060987693e-06,
+      "loss": 0.1832,
+      "step": 35363
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.123520225681326e-06,
+      "loss": 0.2337,
+      "step": 35364
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.123184390374961e-06,
+      "loss": 0.384,
+      "step": 35365
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.122848555068596e-06,
+      "loss": 0.2273,
+      "step": 35366
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.122512719762229e-06,
+      "loss": 0.3577,
+      "step": 35367
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.122176884455863e-06,
+      "loss": 0.2899,
+      "step": 35368
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.121841049149498e-06,
+      "loss": 0.4121,
+      "step": 35369
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.121505213843131e-06,
+      "loss": 0.3072,
+      "step": 35370
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.121169378536766e-06,
+      "loss": 0.3238,
+      "step": 35371
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.1208335432304e-06,
+      "loss": 0.2157,
+      "step": 35372
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.120497707924035e-06,
+      "loss": 0.2617,
+      "step": 35373
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.12016187261767e-06,
+      "loss": 0.1288,
+      "step": 35374
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.119826037311304e-06,
+      "loss": 0.4174,
+      "step": 35375
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.119490202004937e-06,
+      "loss": 0.3596,
+      "step": 35376
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.119154366698572e-06,
+      "loss": 0.2076,
+      "step": 35377
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.118818531392206e-06,
+      "loss": 0.3228,
+      "step": 35378
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.11848269608584e-06,
+      "loss": 0.2141,
+      "step": 35379
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.118146860779474e-06,
+      "loss": 0.3649,
+      "step": 35380
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.117811025473109e-06,
+      "loss": 0.1671,
+      "step": 35381
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.117475190166743e-06,
+      "loss": 0.2772,
+      "step": 35382
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.117139354860378e-06,
+      "loss": 0.3923,
+      "step": 35383
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.116803519554013e-06,
+      "loss": 0.2659,
+      "step": 35384
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.116467684247646e-06,
+      "loss": 0.2881,
+      "step": 35385
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.11613184894128e-06,
+      "loss": 0.293,
+      "step": 35386
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.115796013634915e-06,
+      "loss": 0.2704,
+      "step": 35387
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.115460178328548e-06,
+      "loss": 0.2882,
+      "step": 35388
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.115124343022182e-06,
+      "loss": 0.2716,
+      "step": 35389
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.114788507715817e-06,
+      "loss": 0.4042,
+      "step": 35390
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.11445267240945e-06,
+      "loss": 0.2036,
+      "step": 35391
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.114116837103085e-06,
+      "loss": 0.2729,
+      "step": 35392
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.11378100179672e-06,
+      "loss": 0.2497,
+      "step": 35393
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.113445166490354e-06,
+      "loss": 0.2823,
+      "step": 35394
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.113109331183989e-06,
+      "loss": 0.3065,
+      "step": 35395
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.112773495877623e-06,
+      "loss": 0.1951,
+      "step": 35396
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.112437660571256e-06,
+      "loss": 0.1307,
+      "step": 35397
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.112101825264891e-06,
+      "loss": 0.277,
+      "step": 35398
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.111765989958526e-06,
+      "loss": 0.3435,
+      "step": 35399
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.111430154652158e-06,
+      "loss": 0.3373,
+      "step": 35400
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.111094319345793e-06,
+      "loss": 0.2461,
+      "step": 35401
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.110758484039428e-06,
+      "loss": 0.2625,
+      "step": 35402
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.110422648733062e-06,
+      "loss": 0.3617,
+      "step": 35403
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.110086813426697e-06,
+      "loss": 0.4773,
+      "step": 35404
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.109750978120332e-06,
+      "loss": 0.2645,
+      "step": 35405
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.109415142813965e-06,
+      "loss": 0.3073,
+      "step": 35406
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.1090793075076e-06,
+      "loss": 0.308,
+      "step": 35407
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.108743472201234e-06,
+      "loss": 0.423,
+      "step": 35408
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.108407636894867e-06,
+      "loss": 0.1925,
+      "step": 35409
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.108071801588502e-06,
+      "loss": 0.3656,
+      "step": 35410
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.107735966282136e-06,
+      "loss": 0.271,
+      "step": 35411
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.10740013097577e-06,
+      "loss": 0.3445,
+      "step": 35412
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.107064295669404e-06,
+      "loss": 0.2352,
+      "step": 35413
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.106728460363038e-06,
+      "loss": 0.2261,
+      "step": 35414
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.106392625056673e-06,
+      "loss": 0.2827,
+      "step": 35415
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.106056789750308e-06,
+      "loss": 0.2629,
+      "step": 35416
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.10572095444394e-06,
+      "loss": 0.3252,
+      "step": 35417
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.105385119137575e-06,
+      "loss": 0.3,
+      "step": 35418
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.10504928383121e-06,
+      "loss": 0.2164,
+      "step": 35419
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.104713448524843e-06,
+      "loss": 0.2765,
+      "step": 35420
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.104377613218478e-06,
+      "loss": 0.3711,
+      "step": 35421
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.104041777912112e-06,
+      "loss": 0.3705,
+      "step": 35422
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.103705942605747e-06,
+      "loss": 0.2561,
+      "step": 35423
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.103370107299382e-06,
+      "loss": 0.2426,
+      "step": 35424
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.103034271993016e-06,
+      "loss": 0.214,
+      "step": 35425
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.102698436686649e-06,
+      "loss": 0.2403,
+      "step": 35426
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.102362601380284e-06,
+      "loss": 0.2634,
+      "step": 35427
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.102026766073918e-06,
+      "loss": 0.1912,
+      "step": 35428
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.101690930767551e-06,
+      "loss": 0.2218,
+      "step": 35429
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.101355095461186e-06,
+      "loss": 0.2698,
+      "step": 35430
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.10101926015482e-06,
+      "loss": 0.267,
+      "step": 35431
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.100683424848454e-06,
+      "loss": 0.2499,
+      "step": 35432
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.100347589542088e-06,
+      "loss": 0.506,
+      "step": 35433
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 8.100011754235723e-06,
+      "loss": 0.2798,
+      "step": 35434
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.099675918929358e-06,
+      "loss": 0.3296,
+      "step": 35435
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.099340083622992e-06,
+      "loss": 0.2473,
+      "step": 35436
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.099004248316627e-06,
+      "loss": 0.315,
+      "step": 35437
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.09866841301026e-06,
+      "loss": 0.2091,
+      "step": 35438
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.098332577703894e-06,
+      "loss": 0.3235,
+      "step": 35439
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.097996742397529e-06,
+      "loss": 0.3378,
+      "step": 35440
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.097660907091162e-06,
+      "loss": 0.2431,
+      "step": 35441
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.097325071784797e-06,
+      "loss": 0.307,
+      "step": 35442
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.096989236478431e-06,
+      "loss": 0.2862,
+      "step": 35443
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.096653401172066e-06,
+      "loss": 0.2676,
+      "step": 35444
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.0963175658657e-06,
+      "loss": 0.4352,
+      "step": 35445
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.095981730559335e-06,
+      "loss": 0.3219,
+      "step": 35446
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.095645895252968e-06,
+      "loss": 0.2575,
+      "step": 35447
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.095310059946603e-06,
+      "loss": 0.2592,
+      "step": 35448
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.094974224640238e-06,
+      "loss": 0.3568,
+      "step": 35449
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.09463838933387e-06,
+      "loss": 0.3208,
+      "step": 35450
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.094302554027505e-06,
+      "loss": 0.3024,
+      "step": 35451
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.09396671872114e-06,
+      "loss": 0.2315,
+      "step": 35452
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.093630883414774e-06,
+      "loss": 0.32,
+      "step": 35453
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.093295048108409e-06,
+      "loss": 0.3606,
+      "step": 35454
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.092959212802042e-06,
+      "loss": 0.216,
+      "step": 35455
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.092623377495677e-06,
+      "loss": 0.3177,
+      "step": 35456
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.092287542189311e-06,
+      "loss": 0.3545,
+      "step": 35457
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.091951706882946e-06,
+      "loss": 0.3,
+      "step": 35458
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.091615871576579e-06,
+      "loss": 0.3718,
+      "step": 35459
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.091280036270214e-06,
+      "loss": 0.3027,
+      "step": 35460
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.090944200963848e-06,
+      "loss": 0.2157,
+      "step": 35461
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.090608365657481e-06,
+      "loss": 0.2287,
+      "step": 35462
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.090272530351116e-06,
+      "loss": 0.2198,
+      "step": 35463
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.08993669504475e-06,
+      "loss": 0.2165,
+      "step": 35464
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.089600859738385e-06,
+      "loss": 0.3277,
+      "step": 35465
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.08926502443202e-06,
+      "loss": 0.26,
+      "step": 35466
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.088929189125654e-06,
+      "loss": 0.239,
+      "step": 35467
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.088593353819287e-06,
+      "loss": 0.413,
+      "step": 35468
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.088257518512922e-06,
+      "loss": 0.2886,
+      "step": 35469
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.087921683206557e-06,
+      "loss": 0.2715,
+      "step": 35470
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.08758584790019e-06,
+      "loss": 0.2208,
+      "step": 35471
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.087250012593824e-06,
+      "loss": 0.5632,
+      "step": 35472
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.086914177287459e-06,
+      "loss": 0.4526,
+      "step": 35473
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.086578341981094e-06,
+      "loss": 0.2755,
+      "step": 35474
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.086242506674728e-06,
+      "loss": 0.3438,
+      "step": 35475
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.085906671368363e-06,
+      "loss": 0.3061,
+      "step": 35476
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.085570836061996e-06,
+      "loss": 0.3368,
+      "step": 35477
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.08523500075563e-06,
+      "loss": 0.3135,
+      "step": 35478
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.084899165449265e-06,
+      "loss": 0.2539,
+      "step": 35479
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.084563330142898e-06,
+      "loss": 0.4869,
+      "step": 35480
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.084227494836533e-06,
+      "loss": 0.3922,
+      "step": 35481
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.083891659530167e-06,
+      "loss": 0.3351,
+      "step": 35482
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.0835558242238e-06,
+      "loss": 0.1613,
+      "step": 35483
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.083219988917435e-06,
+      "loss": 0.273,
+      "step": 35484
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.08288415361107e-06,
+      "loss": 0.3974,
+      "step": 35485
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.082548318304704e-06,
+      "loss": 0.3847,
+      "step": 35486
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.082212482998339e-06,
+      "loss": 0.2433,
+      "step": 35487
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.081876647691974e-06,
+      "loss": 0.2283,
+      "step": 35488
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.081540812385607e-06,
+      "loss": 0.1663,
+      "step": 35489
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.081204977079241e-06,
+      "loss": 0.179,
+      "step": 35490
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.080869141772876e-06,
+      "loss": 0.2936,
+      "step": 35491
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.080533306466509e-06,
+      "loss": 0.3798,
+      "step": 35492
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.080197471160143e-06,
+      "loss": 0.3261,
+      "step": 35493
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.079861635853778e-06,
+      "loss": 0.4055,
+      "step": 35494
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.079525800547413e-06,
+      "loss": 0.243,
+      "step": 35495
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.079189965241047e-06,
+      "loss": 0.2333,
+      "step": 35496
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.078854129934682e-06,
+      "loss": 0.1824,
+      "step": 35497
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.078518294628315e-06,
+      "loss": 0.2525,
+      "step": 35498
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.07818245932195e-06,
+      "loss": 0.366,
+      "step": 35499
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.077846624015584e-06,
+      "loss": 0.3681,
+      "step": 35500
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.077510788709217e-06,
+      "loss": 0.2744,
+      "step": 35501
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.077174953402852e-06,
+      "loss": 0.2675,
+      "step": 35502
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.076839118096487e-06,
+      "loss": 0.2807,
+      "step": 35503
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.07650328279012e-06,
+      "loss": 0.2924,
+      "step": 35504
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.076167447483754e-06,
+      "loss": 0.3,
+      "step": 35505
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.075831612177389e-06,
+      "loss": 0.194,
+      "step": 35506
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.075495776871023e-06,
+      "loss": 0.199,
+      "step": 35507
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.075159941564658e-06,
+      "loss": 0.2528,
+      "step": 35508
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.074824106258291e-06,
+      "loss": 0.5687,
+      "step": 35509
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.074488270951926e-06,
+      "loss": 0.3303,
+      "step": 35510
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.07415243564556e-06,
+      "loss": 0.3046,
+      "step": 35511
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.073816600339193e-06,
+      "loss": 0.5166,
+      "step": 35512
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.073480765032828e-06,
+      "loss": 0.1853,
+      "step": 35513
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.073144929726463e-06,
+      "loss": 0.2679,
+      "step": 35514
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.072809094420097e-06,
+      "loss": 0.2418,
+      "step": 35515
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.072473259113732e-06,
+      "loss": 0.3478,
+      "step": 35516
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.072137423807366e-06,
+      "loss": 0.2821,
+      "step": 35517
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.071801588501e-06,
+      "loss": 0.2043,
+      "step": 35518
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.071465753194634e-06,
+      "loss": 0.1755,
+      "step": 35519
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.071129917888269e-06,
+      "loss": 0.2498,
+      "step": 35520
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.070794082581902e-06,
+      "loss": 0.4353,
+      "step": 35521
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.070458247275536e-06,
+      "loss": 0.2125,
+      "step": 35522
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.070122411969171e-06,
+      "loss": 0.3801,
+      "step": 35523
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.069786576662804e-06,
+      "loss": 0.3908,
+      "step": 35524
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.069450741356439e-06,
+      "loss": 0.3821,
+      "step": 35525
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.069114906050073e-06,
+      "loss": 0.2218,
+      "step": 35526
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.068779070743708e-06,
+      "loss": 0.2402,
+      "step": 35527
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.068443235437343e-06,
+      "loss": 0.1881,
+      "step": 35528
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.068107400130977e-06,
+      "loss": 0.3687,
+      "step": 35529
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.06777156482461e-06,
+      "loss": 0.2741,
+      "step": 35530
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.067435729518245e-06,
+      "loss": 0.2657,
+      "step": 35531
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.06709989421188e-06,
+      "loss": 0.2558,
+      "step": 35532
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.066764058905512e-06,
+      "loss": 0.2913,
+      "step": 35533
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.066428223599147e-06,
+      "loss": 0.2595,
+      "step": 35534
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.066092388292782e-06,
+      "loss": 0.3618,
+      "step": 35535
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.065756552986416e-06,
+      "loss": 0.4701,
+      "step": 35536
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.065420717680051e-06,
+      "loss": 0.2401,
+      "step": 35537
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.065084882373686e-06,
+      "loss": 0.1967,
+      "step": 35538
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.064749047067319e-06,
+      "loss": 0.1754,
+      "step": 35539
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.064413211760953e-06,
+      "loss": 0.2907,
+      "step": 35540
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.064077376454588e-06,
+      "loss": 0.2599,
+      "step": 35541
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.06374154114822e-06,
+      "loss": 0.522,
+      "step": 35542
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.063405705841855e-06,
+      "loss": 0.2869,
+      "step": 35543
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.06306987053549e-06,
+      "loss": 0.3423,
+      "step": 35544
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.062734035229123e-06,
+      "loss": 0.2965,
+      "step": 35545
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.062398199922758e-06,
+      "loss": 0.4336,
+      "step": 35546
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.062062364616392e-06,
+      "loss": 0.2643,
+      "step": 35547
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.061726529310027e-06,
+      "loss": 0.3248,
+      "step": 35548
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.061390694003662e-06,
+      "loss": 0.2764,
+      "step": 35549
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.061054858697296e-06,
+      "loss": 0.2835,
+      "step": 35550
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.06071902339093e-06,
+      "loss": 0.1966,
+      "step": 35551
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.060383188084564e-06,
+      "loss": 0.3609,
+      "step": 35552
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.060047352778199e-06,
+      "loss": 0.3526,
+      "step": 35553
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.059711517471832e-06,
+      "loss": 0.2434,
+      "step": 35554
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.059375682165466e-06,
+      "loss": 0.2937,
+      "step": 35555
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.0590398468591e-06,
+      "loss": 0.2954,
+      "step": 35556
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.058704011552735e-06,
+      "loss": 0.2296,
+      "step": 35557
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.05836817624637e-06,
+      "loss": 0.205,
+      "step": 35558
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.058032340940005e-06,
+      "loss": 0.1862,
+      "step": 35559
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.057696505633638e-06,
+      "loss": 0.322,
+      "step": 35560
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.057360670327272e-06,
+      "loss": 0.153,
+      "step": 35561
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.057024835020907e-06,
+      "loss": 0.3502,
+      "step": 35562
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.05668899971454e-06,
+      "loss": 0.2367,
+      "step": 35563
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.056353164408175e-06,
+      "loss": 0.2755,
+      "step": 35564
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.05601732910181e-06,
+      "loss": 0.2258,
+      "step": 35565
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.055681493795444e-06,
+      "loss": 0.2598,
+      "step": 35566
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.055345658489077e-06,
+      "loss": 0.2573,
+      "step": 35567
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.055009823182711e-06,
+      "loss": 0.4283,
+      "step": 35568
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.054673987876346e-06,
+      "loss": 0.2509,
+      "step": 35569
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.05433815256998e-06,
+      "loss": 0.3171,
+      "step": 35570
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.054002317263615e-06,
+      "loss": 0.4489,
+      "step": 35571
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.053666481957248e-06,
+      "loss": 0.2834,
+      "step": 35572
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.053330646650883e-06,
+      "loss": 0.2632,
+      "step": 35573
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.052994811344518e-06,
+      "loss": 0.3348,
+      "step": 35574
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.05265897603815e-06,
+      "loss": 0.3449,
+      "step": 35575
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.052323140731785e-06,
+      "loss": 0.2493,
+      "step": 35576
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.05198730542542e-06,
+      "loss": 0.2489,
+      "step": 35577
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.051651470119055e-06,
+      "loss": 0.1829,
+      "step": 35578
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.05131563481269e-06,
+      "loss": 0.2652,
+      "step": 35579
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.050979799506324e-06,
+      "loss": 0.2565,
+      "step": 35580
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.050643964199957e-06,
+      "loss": 0.4698,
+      "step": 35581
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.050308128893591e-06,
+      "loss": 0.4189,
+      "step": 35582
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.049972293587226e-06,
+      "loss": 0.3174,
+      "step": 35583
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.049636458280859e-06,
+      "loss": 0.2987,
+      "step": 35584
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.049300622974494e-06,
+      "loss": 0.1579,
+      "step": 35585
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.048964787668128e-06,
+      "loss": 0.179,
+      "step": 35586
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.048628952361763e-06,
+      "loss": 0.4248,
+      "step": 35587
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.048293117055398e-06,
+      "loss": 0.341,
+      "step": 35588
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.047957281749032e-06,
+      "loss": 0.2695,
+      "step": 35589
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.047621446442665e-06,
+      "loss": 0.3386,
+      "step": 35590
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.0472856111363e-06,
+      "loss": 0.3572,
+      "step": 35591
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.046949775829935e-06,
+      "loss": 0.2809,
+      "step": 35592
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.046613940523568e-06,
+      "loss": 0.3666,
+      "step": 35593
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.046278105217202e-06,
+      "loss": 0.2018,
+      "step": 35594
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.045942269910837e-06,
+      "loss": 0.3596,
+      "step": 35595
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.04560643460447e-06,
+      "loss": 0.2908,
+      "step": 35596
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.045270599298104e-06,
+      "loss": 0.2496,
+      "step": 35597
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.044934763991739e-06,
+      "loss": 0.3164,
+      "step": 35598
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.044598928685374e-06,
+      "loss": 0.3126,
+      "step": 35599
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.044263093379008e-06,
+      "loss": 0.2792,
+      "step": 35600
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.043927258072641e-06,
+      "loss": 0.2808,
+      "step": 35601
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.043591422766276e-06,
+      "loss": 0.264,
+      "step": 35602
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.04325558745991e-06,
+      "loss": 0.4198,
+      "step": 35603
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.042919752153544e-06,
+      "loss": 0.2301,
+      "step": 35604
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.042583916847178e-06,
+      "loss": 0.3385,
+      "step": 35605
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.042248081540813e-06,
+      "loss": 0.272,
+      "step": 35606
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.041912246234447e-06,
+      "loss": 0.404,
+      "step": 35607
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.041576410928082e-06,
+      "loss": 0.2536,
+      "step": 35608
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.041240575621717e-06,
+      "loss": 0.4967,
+      "step": 35609
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.04090474031535e-06,
+      "loss": 0.5583,
+      "step": 35610
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.040568905008984e-06,
+      "loss": 0.1923,
+      "step": 35611
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.040233069702619e-06,
+      "loss": 0.18,
+      "step": 35612
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.039897234396252e-06,
+      "loss": 0.344,
+      "step": 35613
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.039561399089887e-06,
+      "loss": 0.2856,
+      "step": 35614
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.039225563783521e-06,
+      "loss": 0.5888,
+      "step": 35615
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.038889728477154e-06,
+      "loss": 0.3182,
+      "step": 35616
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.038553893170789e-06,
+      "loss": 0.2228,
+      "step": 35617
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.038218057864424e-06,
+      "loss": 0.1701,
+      "step": 35618
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.037882222558058e-06,
+      "loss": 0.53,
+      "step": 35619
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.037546387251693e-06,
+      "loss": 0.292,
+      "step": 35620
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.037210551945327e-06,
+      "loss": 0.2373,
+      "step": 35621
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.03687471663896e-06,
+      "loss": 0.276,
+      "step": 35622
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.036538881332595e-06,
+      "loss": 0.2172,
+      "step": 35623
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.03620304602623e-06,
+      "loss": 0.2618,
+      "step": 35624
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.035867210719863e-06,
+      "loss": 0.1953,
+      "step": 35625
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.035531375413497e-06,
+      "loss": 0.2272,
+      "step": 35626
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.035195540107132e-06,
+      "loss": 0.4295,
+      "step": 35627
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.034859704800767e-06,
+      "loss": 0.2674,
+      "step": 35628
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.034523869494401e-06,
+      "loss": 0.3211,
+      "step": 35629
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.034188034188036e-06,
+      "loss": 0.4003,
+      "step": 35630
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.033852198881669e-06,
+      "loss": 0.2996,
+      "step": 35631
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 8.033516363575304e-06,
+      "loss": 0.2867,
+      "step": 35632
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.033180528268938e-06,
+      "loss": 0.2104,
+      "step": 35633
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.032844692962571e-06,
+      "loss": 0.2607,
+      "step": 35634
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.032508857656206e-06,
+      "loss": 0.5868,
+      "step": 35635
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.03217302234984e-06,
+      "loss": 0.1675,
+      "step": 35636
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.031837187043473e-06,
+      "loss": 0.4009,
+      "step": 35637
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.031501351737108e-06,
+      "loss": 0.1824,
+      "step": 35638
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.031165516430743e-06,
+      "loss": 0.2807,
+      "step": 35639
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.030829681124377e-06,
+      "loss": 0.2121,
+      "step": 35640
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.030493845818012e-06,
+      "loss": 0.2473,
+      "step": 35641
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.030158010511647e-06,
+      "loss": 0.2535,
+      "step": 35642
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.02982217520528e-06,
+      "loss": 0.2251,
+      "step": 35643
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.029486339898914e-06,
+      "loss": 0.296,
+      "step": 35644
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.029150504592549e-06,
+      "loss": 0.2986,
+      "step": 35645
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.028814669286182e-06,
+      "loss": 0.2923,
+      "step": 35646
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.028478833979816e-06,
+      "loss": 0.1662,
+      "step": 35647
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.028142998673451e-06,
+      "loss": 0.3105,
+      "step": 35648
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.027807163367086e-06,
+      "loss": 0.3858,
+      "step": 35649
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.02747132806072e-06,
+      "loss": 0.1885,
+      "step": 35650
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.027135492754355e-06,
+      "loss": 0.1844,
+      "step": 35651
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.026799657447988e-06,
+      "loss": 0.3538,
+      "step": 35652
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.026463822141623e-06,
+      "loss": 0.2581,
+      "step": 35653
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.026127986835257e-06,
+      "loss": 0.4901,
+      "step": 35654
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.02579215152889e-06,
+      "loss": 0.2534,
+      "step": 35655
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.025456316222525e-06,
+      "loss": 0.1663,
+      "step": 35656
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.02512048091616e-06,
+      "loss": 0.2595,
+      "step": 35657
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.024784645609792e-06,
+      "loss": 0.2517,
+      "step": 35658
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.024448810303427e-06,
+      "loss": 0.2811,
+      "step": 35659
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.024112974997062e-06,
+      "loss": 0.2622,
+      "step": 35660
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.023777139690696e-06,
+      "loss": 0.3071,
+      "step": 35661
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.023441304384331e-06,
+      "loss": 0.2669,
+      "step": 35662
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.023105469077966e-06,
+      "loss": 0.3814,
+      "step": 35663
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.022769633771599e-06,
+      "loss": 0.2852,
+      "step": 35664
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.022433798465233e-06,
+      "loss": 0.3312,
+      "step": 35665
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.022097963158868e-06,
+      "loss": 0.4308,
+      "step": 35666
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.021762127852501e-06,
+      "loss": 0.3592,
+      "step": 35667
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.021426292546136e-06,
+      "loss": 0.2099,
+      "step": 35668
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.02109045723977e-06,
+      "loss": 0.3074,
+      "step": 35669
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.020754621933405e-06,
+      "loss": 0.3018,
+      "step": 35670
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.02041878662704e-06,
+      "loss": 0.3985,
+      "step": 35671
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.020082951320674e-06,
+      "loss": 0.1595,
+      "step": 35672
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.019747116014307e-06,
+      "loss": 0.2412,
+      "step": 35673
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.019411280707942e-06,
+      "loss": 0.5327,
+      "step": 35674
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.019075445401576e-06,
+      "loss": 0.458,
+      "step": 35675
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.01873961009521e-06,
+      "loss": 0.3152,
+      "step": 35676
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.018403774788844e-06,
+      "loss": 0.3515,
+      "step": 35677
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.018067939482479e-06,
+      "loss": 0.2076,
+      "step": 35678
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.017732104176113e-06,
+      "loss": 0.3724,
+      "step": 35679
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.017396268869746e-06,
+      "loss": 0.2332,
+      "step": 35680
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.017060433563381e-06,
+      "loss": 0.2994,
+      "step": 35681
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.016724598257016e-06,
+      "loss": 0.3096,
+      "step": 35682
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.01638876295065e-06,
+      "loss": 0.3402,
+      "step": 35683
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.016052927644285e-06,
+      "loss": 0.2029,
+      "step": 35684
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.015717092337918e-06,
+      "loss": 0.2733,
+      "step": 35685
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.015381257031552e-06,
+      "loss": 0.2885,
+      "step": 35686
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.015045421725185e-06,
+      "loss": 0.2282,
+      "step": 35687
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.01470958641882e-06,
+      "loss": 0.143,
+      "step": 35688
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.014373751112455e-06,
+      "loss": 0.3015,
+      "step": 35689
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.01403791580609e-06,
+      "loss": 0.2903,
+      "step": 35690
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.013702080499724e-06,
+      "loss": 0.2386,
+      "step": 35691
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.013366245193359e-06,
+      "loss": 0.3782,
+      "step": 35692
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.013030409886992e-06,
+      "loss": 0.3482,
+      "step": 35693
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.012694574580626e-06,
+      "loss": 0.3283,
+      "step": 35694
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.012358739274261e-06,
+      "loss": 0.2196,
+      "step": 35695
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.012022903967894e-06,
+      "loss": 0.2022,
+      "step": 35696
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.011687068661528e-06,
+      "loss": 0.3196,
+      "step": 35697
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.011351233355163e-06,
+      "loss": 0.5417,
+      "step": 35698
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.011015398048798e-06,
+      "loss": 0.1843,
+      "step": 35699
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.010679562742432e-06,
+      "loss": 0.32,
+      "step": 35700
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.010343727436067e-06,
+      "loss": 0.2753,
+      "step": 35701
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.0100078921297e-06,
+      "loss": 0.4208,
+      "step": 35702
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.009672056823335e-06,
+      "loss": 0.2515,
+      "step": 35703
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.00933622151697e-06,
+      "loss": 0.222,
+      "step": 35704
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.009000386210602e-06,
+      "loss": 0.2141,
+      "step": 35705
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.008664550904237e-06,
+      "loss": 0.1483,
+      "step": 35706
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.008328715597872e-06,
+      "loss": 0.5009,
+      "step": 35707
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.007992880291505e-06,
+      "loss": 0.2617,
+      "step": 35708
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.00765704498514e-06,
+      "loss": 0.2744,
+      "step": 35709
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.007321209678774e-06,
+      "loss": 0.4552,
+      "step": 35710
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.006985374372408e-06,
+      "loss": 0.3854,
+      "step": 35711
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.006649539066043e-06,
+      "loss": 0.3379,
+      "step": 35712
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.006313703759678e-06,
+      "loss": 0.2818,
+      "step": 35713
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.00597786845331e-06,
+      "loss": 0.3771,
+      "step": 35714
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.005642033146945e-06,
+      "loss": 0.3977,
+      "step": 35715
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.00530619784058e-06,
+      "loss": 0.4796,
+      "step": 35716
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.004970362534213e-06,
+      "loss": 0.2826,
+      "step": 35717
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.004634527227848e-06,
+      "loss": 0.2802,
+      "step": 35718
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.004298691921482e-06,
+      "loss": 0.2669,
+      "step": 35719
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.003962856615117e-06,
+      "loss": 0.2397,
+      "step": 35720
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.003627021308752e-06,
+      "loss": 0.1964,
+      "step": 35721
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.003291186002386e-06,
+      "loss": 0.1405,
+      "step": 35722
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.002955350696019e-06,
+      "loss": 0.1212,
+      "step": 35723
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.002619515389654e-06,
+      "loss": 0.3147,
+      "step": 35724
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.002283680083288e-06,
+      "loss": 0.214,
+      "step": 35725
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.001947844776921e-06,
+      "loss": 0.2929,
+      "step": 35726
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.001612009470556e-06,
+      "loss": 0.3354,
+      "step": 35727
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.00127617416419e-06,
+      "loss": 0.3408,
+      "step": 35728
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.000940338857824e-06,
+      "loss": 0.2444,
+      "step": 35729
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.000604503551458e-06,
+      "loss": 0.3185,
+      "step": 35730
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 8.000268668245093e-06,
+      "loss": 0.1967,
+      "step": 35731
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.999932832938728e-06,
+      "loss": 0.1803,
+      "step": 35732
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.999596997632362e-06,
+      "loss": 0.309,
+      "step": 35733
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.999261162325997e-06,
+      "loss": 0.1507,
+      "step": 35734
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.99892532701963e-06,
+      "loss": 0.1471,
+      "step": 35735
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.998589491713264e-06,
+      "loss": 0.3002,
+      "step": 35736
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.998253656406899e-06,
+      "loss": 0.4108,
+      "step": 35737
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.997917821100532e-06,
+      "loss": 0.3303,
+      "step": 35738
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.997581985794167e-06,
+      "loss": 0.2962,
+      "step": 35739
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.997246150487801e-06,
+      "loss": 0.3612,
+      "step": 35740
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.996910315181436e-06,
+      "loss": 0.2869,
+      "step": 35741
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.99657447987507e-06,
+      "loss": 0.2834,
+      "step": 35742
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.996238644568705e-06,
+      "loss": 0.3452,
+      "step": 35743
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.995902809262338e-06,
+      "loss": 0.2216,
+      "step": 35744
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.995566973955973e-06,
+      "loss": 0.2557,
+      "step": 35745
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.995231138649608e-06,
+      "loss": 0.2891,
+      "step": 35746
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.99489530334324e-06,
+      "loss": 0.4076,
+      "step": 35747
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.994559468036875e-06,
+      "loss": 0.2779,
+      "step": 35748
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.99422363273051e-06,
+      "loss": 0.3687,
+      "step": 35749
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.993887797424143e-06,
+      "loss": 0.2231,
+      "step": 35750
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.993551962117777e-06,
+      "loss": 0.1745,
+      "step": 35751
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.993216126811412e-06,
+      "loss": 0.3781,
+      "step": 35752
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.992880291505047e-06,
+      "loss": 0.3798,
+      "step": 35753
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.992544456198681e-06,
+      "loss": 0.3207,
+      "step": 35754
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.992208620892316e-06,
+      "loss": 0.2863,
+      "step": 35755
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.991872785585949e-06,
+      "loss": 0.2758,
+      "step": 35756
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.991536950279584e-06,
+      "loss": 0.1964,
+      "step": 35757
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.991201114973218e-06,
+      "loss": 0.2586,
+      "step": 35758
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.990865279666851e-06,
+      "loss": 0.2866,
+      "step": 35759
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.990529444360486e-06,
+      "loss": 0.3497,
+      "step": 35760
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.99019360905412e-06,
+      "loss": 0.2222,
+      "step": 35761
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.989857773747755e-06,
+      "loss": 0.4788,
+      "step": 35762
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.98952193844139e-06,
+      "loss": 0.3148,
+      "step": 35763
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.989186103135024e-06,
+      "loss": 0.262,
+      "step": 35764
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.988850267828657e-06,
+      "loss": 0.2953,
+      "step": 35765
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.988514432522292e-06,
+      "loss": 0.4499,
+      "step": 35766
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.988178597215927e-06,
+      "loss": 0.3843,
+      "step": 35767
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.98784276190956e-06,
+      "loss": 0.3578,
+      "step": 35768
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.987506926603194e-06,
+      "loss": 0.313,
+      "step": 35769
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.987171091296829e-06,
+      "loss": 0.2524,
+      "step": 35770
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.986835255990462e-06,
+      "loss": 0.1778,
+      "step": 35771
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.986499420684097e-06,
+      "loss": 0.2252,
+      "step": 35772
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.986163585377731e-06,
+      "loss": 0.2921,
+      "step": 35773
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.985827750071366e-06,
+      "loss": 0.2376,
+      "step": 35774
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.985491914765e-06,
+      "loss": 0.3707,
+      "step": 35775
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.985156079458635e-06,
+      "loss": 0.5032,
+      "step": 35776
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.984820244152268e-06,
+      "loss": 0.531,
+      "step": 35777
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.984484408845903e-06,
+      "loss": 0.1911,
+      "step": 35778
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.984148573539536e-06,
+      "loss": 0.3848,
+      "step": 35779
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.98381273823317e-06,
+      "loss": 0.2708,
+      "step": 35780
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.983476902926805e-06,
+      "loss": 0.1852,
+      "step": 35781
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.98314106762044e-06,
+      "loss": 0.3153,
+      "step": 35782
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.982805232314074e-06,
+      "loss": 0.2174,
+      "step": 35783
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.982469397007709e-06,
+      "loss": 0.2525,
+      "step": 35784
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.982133561701342e-06,
+      "loss": 0.2672,
+      "step": 35785
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.981797726394977e-06,
+      "loss": 0.303,
+      "step": 35786
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.981461891088611e-06,
+      "loss": 0.2111,
+      "step": 35787
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.981126055782244e-06,
+      "loss": 0.343,
+      "step": 35788
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.980790220475879e-06,
+      "loss": 0.3547,
+      "step": 35789
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.980454385169513e-06,
+      "loss": 0.2458,
+      "step": 35790
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.980118549863148e-06,
+      "loss": 0.3912,
+      "step": 35791
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.979782714556783e-06,
+      "loss": 0.3335,
+      "step": 35792
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.979446879250416e-06,
+      "loss": 0.2171,
+      "step": 35793
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.97911104394405e-06,
+      "loss": 0.3099,
+      "step": 35794
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.978775208637685e-06,
+      "loss": 0.3395,
+      "step": 35795
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.97843937333132e-06,
+      "loss": 0.245,
+      "step": 35796
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.978103538024953e-06,
+      "loss": 0.2231,
+      "step": 35797
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.977767702718587e-06,
+      "loss": 0.2599,
+      "step": 35798
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.977431867412222e-06,
+      "loss": 0.419,
+      "step": 35799
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.977096032105855e-06,
+      "loss": 0.1815,
+      "step": 35800
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.97676019679949e-06,
+      "loss": 0.3658,
+      "step": 35801
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.976424361493124e-06,
+      "loss": 0.1622,
+      "step": 35802
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.976088526186759e-06,
+      "loss": 0.2811,
+      "step": 35803
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.975752690880393e-06,
+      "loss": 0.412,
+      "step": 35804
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.975416855574028e-06,
+      "loss": 0.4403,
+      "step": 35805
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.975081020267661e-06,
+      "loss": 0.4616,
+      "step": 35806
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.974745184961296e-06,
+      "loss": 0.2268,
+      "step": 35807
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.97440934965493e-06,
+      "loss": 0.3036,
+      "step": 35808
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.974073514348563e-06,
+      "loss": 0.2172,
+      "step": 35809
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.973737679042198e-06,
+      "loss": 0.2903,
+      "step": 35810
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.973401843735833e-06,
+      "loss": 0.2869,
+      "step": 35811
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.973066008429467e-06,
+      "loss": 0.3444,
+      "step": 35812
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.972730173123102e-06,
+      "loss": 0.2639,
+      "step": 35813
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.972394337816736e-06,
+      "loss": 0.3139,
+      "step": 35814
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.97205850251037e-06,
+      "loss": 0.2371,
+      "step": 35815
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.971722667204004e-06,
+      "loss": 0.1757,
+      "step": 35816
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.971386831897639e-06,
+      "loss": 0.4318,
+      "step": 35817
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.971050996591272e-06,
+      "loss": 0.199,
+      "step": 35818
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.970715161284906e-06,
+      "loss": 0.311,
+      "step": 35819
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.970379325978541e-06,
+      "loss": 0.2877,
+      "step": 35820
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.970043490672174e-06,
+      "loss": 0.2289,
+      "step": 35821
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.969707655365809e-06,
+      "loss": 0.256,
+      "step": 35822
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.969371820059443e-06,
+      "loss": 0.3373,
+      "step": 35823
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.969035984753078e-06,
+      "loss": 0.3665,
+      "step": 35824
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.968700149446713e-06,
+      "loss": 0.2944,
+      "step": 35825
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.968364314140347e-06,
+      "loss": 0.231,
+      "step": 35826
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.96802847883398e-06,
+      "loss": 0.2738,
+      "step": 35827
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.967692643527615e-06,
+      "loss": 0.249,
+      "step": 35828
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.96735680822125e-06,
+      "loss": 0.2694,
+      "step": 35829
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.967020972914882e-06,
+      "loss": 0.2907,
+      "step": 35830
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 7.966685137608517e-06,
+      "loss": 0.286,
+      "step": 35831
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.966349302302152e-06,
+      "loss": 0.3721,
+      "step": 35832
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.966013466995786e-06,
+      "loss": 0.3381,
+      "step": 35833
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.965677631689421e-06,
+      "loss": 0.3324,
+      "step": 35834
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.965341796383056e-06,
+      "loss": 0.3489,
+      "step": 35835
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.965005961076689e-06,
+      "loss": 0.3073,
+      "step": 35836
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.964670125770323e-06,
+      "loss": 0.368,
+      "step": 35837
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.964334290463958e-06,
+      "loss": 0.4111,
+      "step": 35838
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.96399845515759e-06,
+      "loss": 0.3262,
+      "step": 35839
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.963662619851225e-06,
+      "loss": 0.3352,
+      "step": 35840
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.96332678454486e-06,
+      "loss": 0.3034,
+      "step": 35841
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.962990949238493e-06,
+      "loss": 0.2119,
+      "step": 35842
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.962655113932128e-06,
+      "loss": 0.3807,
+      "step": 35843
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.962319278625762e-06,
+      "loss": 0.2456,
+      "step": 35844
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.961983443319397e-06,
+      "loss": 0.2499,
+      "step": 35845
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.961647608013032e-06,
+      "loss": 0.1783,
+      "step": 35846
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.961311772706666e-06,
+      "loss": 0.2354,
+      "step": 35847
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.9609759374003e-06,
+      "loss": 0.3098,
+      "step": 35848
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.960640102093934e-06,
+      "loss": 0.3154,
+      "step": 35849
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.960304266787569e-06,
+      "loss": 0.3151,
+      "step": 35850
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.959968431481201e-06,
+      "loss": 0.2812,
+      "step": 35851
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.959632596174836e-06,
+      "loss": 0.3706,
+      "step": 35852
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.95929676086847e-06,
+      "loss": 0.1302,
+      "step": 35853
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.958960925562105e-06,
+      "loss": 0.3874,
+      "step": 35854
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.95862509025574e-06,
+      "loss": 0.176,
+      "step": 35855
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.958289254949375e-06,
+      "loss": 0.249,
+      "step": 35856
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.957953419643008e-06,
+      "loss": 0.216,
+      "step": 35857
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.957617584336642e-06,
+      "loss": 0.2452,
+      "step": 35858
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.957281749030277e-06,
+      "loss": 0.3339,
+      "step": 35859
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.95694591372391e-06,
+      "loss": 0.3645,
+      "step": 35860
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.956610078417545e-06,
+      "loss": 0.3611,
+      "step": 35861
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.95627424311118e-06,
+      "loss": 0.1154,
+      "step": 35862
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.955938407804812e-06,
+      "loss": 0.4186,
+      "step": 35863
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.955602572498447e-06,
+      "loss": 0.2012,
+      "step": 35864
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.955266737192081e-06,
+      "loss": 0.2867,
+      "step": 35865
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.954930901885716e-06,
+      "loss": 0.1951,
+      "step": 35866
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.95459506657935e-06,
+      "loss": 0.3585,
+      "step": 35867
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.954259231272984e-06,
+      "loss": 0.3703,
+      "step": 35868
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.953923395966618e-06,
+      "loss": 0.2746,
+      "step": 35869
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.953587560660253e-06,
+      "loss": 0.3166,
+      "step": 35870
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.953251725353886e-06,
+      "loss": 0.2866,
+      "step": 35871
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.95291589004752e-06,
+      "loss": 0.2223,
+      "step": 35872
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.952580054741155e-06,
+      "loss": 0.2518,
+      "step": 35873
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.95224421943479e-06,
+      "loss": 0.343,
+      "step": 35874
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.951908384128425e-06,
+      "loss": 0.313,
+      "step": 35875
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.95157254882206e-06,
+      "loss": 0.2559,
+      "step": 35876
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.951236713515692e-06,
+      "loss": 0.2401,
+      "step": 35877
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.950900878209327e-06,
+      "loss": 0.2192,
+      "step": 35878
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.950565042902961e-06,
+      "loss": 0.4491,
+      "step": 35879
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.950229207596594e-06,
+      "loss": 0.3931,
+      "step": 35880
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.949893372290229e-06,
+      "loss": 0.2922,
+      "step": 35881
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.949557536983864e-06,
+      "loss": 0.2299,
+      "step": 35882
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.949221701677498e-06,
+      "loss": 0.1593,
+      "step": 35883
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.948885866371131e-06,
+      "loss": 0.2982,
+      "step": 35884
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.948550031064766e-06,
+      "loss": 0.4222,
+      "step": 35885
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.9482141957584e-06,
+      "loss": 0.248,
+      "step": 35886
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.947878360452035e-06,
+      "loss": 0.3305,
+      "step": 35887
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.94754252514567e-06,
+      "loss": 0.2128,
+      "step": 35888
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.947206689839303e-06,
+      "loss": 0.2612,
+      "step": 35889
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.946870854532937e-06,
+      "loss": 0.1279,
+      "step": 35890
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.946535019226572e-06,
+      "loss": 0.2996,
+      "step": 35891
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.946199183920205e-06,
+      "loss": 0.2904,
+      "step": 35892
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.94586334861384e-06,
+      "loss": 0.2786,
+      "step": 35893
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.945527513307474e-06,
+      "loss": 0.3369,
+      "step": 35894
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.945191678001109e-06,
+      "loss": 0.299,
+      "step": 35895
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.944855842694744e-06,
+      "loss": 0.2944,
+      "step": 35896
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.944520007388378e-06,
+      "loss": 0.3254,
+      "step": 35897
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.944184172082011e-06,
+      "loss": 0.3467,
+      "step": 35898
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.943848336775646e-06,
+      "loss": 0.2931,
+      "step": 35899
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.94351250146928e-06,
+      "loss": 0.484,
+      "step": 35900
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.943176666162914e-06,
+      "loss": 0.2142,
+      "step": 35901
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.942840830856548e-06,
+      "loss": 0.1629,
+      "step": 35902
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.942504995550183e-06,
+      "loss": 0.3312,
+      "step": 35903
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.942169160243817e-06,
+      "loss": 0.1522,
+      "step": 35904
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.941833324937452e-06,
+      "loss": 0.2473,
+      "step": 35905
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.941497489631085e-06,
+      "loss": 0.5147,
+      "step": 35906
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.94116165432472e-06,
+      "loss": 0.2431,
+      "step": 35907
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.940825819018354e-06,
+      "loss": 0.2298,
+      "step": 35908
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.940489983711989e-06,
+      "loss": 0.4083,
+      "step": 35909
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.940154148405622e-06,
+      "loss": 0.4448,
+      "step": 35910
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.939818313099257e-06,
+      "loss": 0.3959,
+      "step": 35911
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.939482477792891e-06,
+      "loss": 0.3284,
+      "step": 35912
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.939146642486524e-06,
+      "loss": 0.2044,
+      "step": 35913
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.938810807180159e-06,
+      "loss": 0.3682,
+      "step": 35914
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.938474971873794e-06,
+      "loss": 0.2486,
+      "step": 35915
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.938139136567428e-06,
+      "loss": 0.2959,
+      "step": 35916
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.937803301261063e-06,
+      "loss": 0.2314,
+      "step": 35917
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.937467465954697e-06,
+      "loss": 0.3228,
+      "step": 35918
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.93713163064833e-06,
+      "loss": 0.2083,
+      "step": 35919
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.936795795341965e-06,
+      "loss": 0.3998,
+      "step": 35920
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.9364599600356e-06,
+      "loss": 0.1991,
+      "step": 35921
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.936124124729233e-06,
+      "loss": 0.162,
+      "step": 35922
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.935788289422867e-06,
+      "loss": 0.323,
+      "step": 35923
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.935452454116502e-06,
+      "loss": 0.2015,
+      "step": 35924
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.935116618810137e-06,
+      "loss": 0.2455,
+      "step": 35925
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.934780783503771e-06,
+      "loss": 0.4206,
+      "step": 35926
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.934444948197406e-06,
+      "loss": 0.3278,
+      "step": 35927
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.934109112891039e-06,
+      "loss": 0.2994,
+      "step": 35928
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.933773277584673e-06,
+      "loss": 0.4545,
+      "step": 35929
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.933437442278308e-06,
+      "loss": 0.2289,
+      "step": 35930
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.933101606971941e-06,
+      "loss": 0.2054,
+      "step": 35931
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.932765771665576e-06,
+      "loss": 0.2717,
+      "step": 35932
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.93242993635921e-06,
+      "loss": 0.3258,
+      "step": 35933
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.932094101052843e-06,
+      "loss": 0.3307,
+      "step": 35934
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.931758265746478e-06,
+      "loss": 0.3192,
+      "step": 35935
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.931422430440113e-06,
+      "loss": 0.4676,
+      "step": 35936
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.931086595133747e-06,
+      "loss": 0.3431,
+      "step": 35937
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.930750759827382e-06,
+      "loss": 0.3652,
+      "step": 35938
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.930414924521017e-06,
+      "loss": 0.2647,
+      "step": 35939
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.93007908921465e-06,
+      "loss": 0.4626,
+      "step": 35940
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.929743253908284e-06,
+      "loss": 0.2194,
+      "step": 35941
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.929407418601919e-06,
+      "loss": 0.2905,
+      "step": 35942
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.929071583295552e-06,
+      "loss": 0.2597,
+      "step": 35943
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.928735747989186e-06,
+      "loss": 0.4517,
+      "step": 35944
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.928399912682821e-06,
+      "loss": 0.3447,
+      "step": 35945
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.928064077376456e-06,
+      "loss": 0.3393,
+      "step": 35946
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.92772824207009e-06,
+      "loss": 0.3357,
+      "step": 35947
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.927392406763725e-06,
+      "loss": 0.3741,
+      "step": 35948
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.927056571457358e-06,
+      "loss": 0.3192,
+      "step": 35949
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.926720736150993e-06,
+      "loss": 0.4126,
+      "step": 35950
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.926384900844627e-06,
+      "loss": 0.2523,
+      "step": 35951
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.92604906553826e-06,
+      "loss": 0.1337,
+      "step": 35952
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.925713230231895e-06,
+      "loss": 0.4321,
+      "step": 35953
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.92537739492553e-06,
+      "loss": 0.2809,
+      "step": 35954
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.925041559619162e-06,
+      "loss": 0.2589,
+      "step": 35955
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.924705724312797e-06,
+      "loss": 0.2359,
+      "step": 35956
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.924369889006432e-06,
+      "loss": 0.3121,
+      "step": 35957
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.924034053700066e-06,
+      "loss": 0.2242,
+      "step": 35958
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.923698218393701e-06,
+      "loss": 0.283,
+      "step": 35959
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.923362383087334e-06,
+      "loss": 0.2079,
+      "step": 35960
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.923026547780969e-06,
+      "loss": 0.3181,
+      "step": 35961
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.922690712474603e-06,
+      "loss": 0.3084,
+      "step": 35962
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.922354877168236e-06,
+      "loss": 0.245,
+      "step": 35963
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.922019041861871e-06,
+      "loss": 0.3513,
+      "step": 35964
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.921683206555506e-06,
+      "loss": 0.2671,
+      "step": 35965
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.92134737124914e-06,
+      "loss": 0.3329,
+      "step": 35966
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.921011535942775e-06,
+      "loss": 0.2628,
+      "step": 35967
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.92067570063641e-06,
+      "loss": 0.2269,
+      "step": 35968
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.920339865330042e-06,
+      "loss": 0.2469,
+      "step": 35969
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.920004030023677e-06,
+      "loss": 0.4965,
+      "step": 35970
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.919668194717312e-06,
+      "loss": 0.2882,
+      "step": 35971
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.919332359410945e-06,
+      "loss": 0.3301,
+      "step": 35972
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.91899652410458e-06,
+      "loss": 0.2886,
+      "step": 35973
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.918660688798214e-06,
+      "loss": 0.2789,
+      "step": 35974
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.918324853491847e-06,
+      "loss": 0.1029,
+      "step": 35975
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.917989018185482e-06,
+      "loss": 0.1971,
+      "step": 35976
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.917653182879116e-06,
+      "loss": 0.1793,
+      "step": 35977
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.917317347572751e-06,
+      "loss": 0.1808,
+      "step": 35978
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.916981512266386e-06,
+      "loss": 0.3596,
+      "step": 35979
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.91664567696002e-06,
+      "loss": 0.4157,
+      "step": 35980
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.916309841653653e-06,
+      "loss": 0.1674,
+      "step": 35981
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.915974006347288e-06,
+      "loss": 0.2233,
+      "step": 35982
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.915638171040922e-06,
+      "loss": 0.2724,
+      "step": 35983
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.915302335734555e-06,
+      "loss": 0.3306,
+      "step": 35984
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.91496650042819e-06,
+      "loss": 0.4599,
+      "step": 35985
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.914630665121825e-06,
+      "loss": 0.3332,
+      "step": 35986
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.91429482981546e-06,
+      "loss": 0.3802,
+      "step": 35987
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.913958994509094e-06,
+      "loss": 0.2589,
+      "step": 35988
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.913623159202729e-06,
+      "loss": 0.3916,
+      "step": 35989
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.913287323896362e-06,
+      "loss": 0.2377,
+      "step": 35990
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.912951488589996e-06,
+      "loss": 0.3782,
+      "step": 35991
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.912615653283631e-06,
+      "loss": 0.2232,
+      "step": 35992
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.912279817977264e-06,
+      "loss": 0.3693,
+      "step": 35993
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.911943982670898e-06,
+      "loss": 0.4185,
+      "step": 35994
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.911608147364533e-06,
+      "loss": 0.309,
+      "step": 35995
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.911272312058168e-06,
+      "loss": 0.1588,
+      "step": 35996
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.9109364767518e-06,
+      "loss": 0.3873,
+      "step": 35997
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.910600641445435e-06,
+      "loss": 0.2963,
+      "step": 35998
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.91026480613907e-06,
+      "loss": 0.3529,
+      "step": 35999
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.909928970832705e-06,
+      "loss": 0.31,
+      "step": 36000
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.90959313552634e-06,
+      "loss": 0.2715,
+      "step": 36001
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.909257300219972e-06,
+      "loss": 0.2759,
+      "step": 36002
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.908921464913607e-06,
+      "loss": 0.2368,
+      "step": 36003
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.908585629607242e-06,
+      "loss": 0.3728,
+      "step": 36004
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.908249794300875e-06,
+      "loss": 0.4409,
+      "step": 36005
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.90791395899451e-06,
+      "loss": 0.2328,
+      "step": 36006
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.907578123688144e-06,
+      "loss": 0.2264,
+      "step": 36007
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.907242288381778e-06,
+      "loss": 0.3332,
+      "step": 36008
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.906906453075413e-06,
+      "loss": 0.3363,
+      "step": 36009
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.906570617769048e-06,
+      "loss": 0.2808,
+      "step": 36010
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.90623478246268e-06,
+      "loss": 0.2215,
+      "step": 36011
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.905898947156315e-06,
+      "loss": 0.3404,
+      "step": 36012
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.90556311184995e-06,
+      "loss": 0.2344,
+      "step": 36013
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.905227276543583e-06,
+      "loss": 0.3321,
+      "step": 36014
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.904891441237218e-06,
+      "loss": 0.2524,
+      "step": 36015
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.904555605930852e-06,
+      "loss": 0.2175,
+      "step": 36016
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.904219770624487e-06,
+      "loss": 0.3125,
+      "step": 36017
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.903883935318122e-06,
+      "loss": 0.2232,
+      "step": 36018
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.903548100011754e-06,
+      "loss": 0.1697,
+      "step": 36019
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.903212264705389e-06,
+      "loss": 0.3983,
+      "step": 36020
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.902876429399024e-06,
+      "loss": 0.3622,
+      "step": 36021
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.902540594092658e-06,
+      "loss": 0.3594,
+      "step": 36022
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.902204758786291e-06,
+      "loss": 0.4201,
+      "step": 36023
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.901868923479926e-06,
+      "loss": 0.4837,
+      "step": 36024
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.90153308817356e-06,
+      "loss": 0.2193,
+      "step": 36025
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.901197252867194e-06,
+      "loss": 0.2224,
+      "step": 36026
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.900861417560828e-06,
+      "loss": 0.2376,
+      "step": 36027
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.900525582254463e-06,
+      "loss": 0.3368,
+      "step": 36028
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 7.900189746948098e-06,
+      "loss": 0.4699,
+      "step": 36029
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.899853911641732e-06,
+      "loss": 0.2239,
+      "step": 36030
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.899518076335367e-06,
+      "loss": 0.2339,
+      "step": 36031
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.899182241029e-06,
+      "loss": 0.2485,
+      "step": 36032
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.898846405722634e-06,
+      "loss": 0.3008,
+      "step": 36033
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.898510570416269e-06,
+      "loss": 0.2632,
+      "step": 36034
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.898174735109902e-06,
+      "loss": 0.252,
+      "step": 36035
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.897838899803537e-06,
+      "loss": 0.2395,
+      "step": 36036
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.897503064497171e-06,
+      "loss": 0.3789,
+      "step": 36037
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.897167229190806e-06,
+      "loss": 0.2729,
+      "step": 36038
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.89683139388444e-06,
+      "loss": 0.4396,
+      "step": 36039
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.896495558578075e-06,
+      "loss": 0.4041,
+      "step": 36040
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.896159723271708e-06,
+      "loss": 0.3535,
+      "step": 36041
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.895823887965343e-06,
+      "loss": 0.3436,
+      "step": 36042
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.895488052658978e-06,
+      "loss": 0.3784,
+      "step": 36043
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.89515221735261e-06,
+      "loss": 0.342,
+      "step": 36044
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.894816382046245e-06,
+      "loss": 0.3549,
+      "step": 36045
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.89448054673988e-06,
+      "loss": 0.3283,
+      "step": 36046
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.894144711433513e-06,
+      "loss": 0.2518,
+      "step": 36047
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.893808876127147e-06,
+      "loss": 0.2644,
+      "step": 36048
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.893473040820782e-06,
+      "loss": 0.2934,
+      "step": 36049
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.893137205514417e-06,
+      "loss": 0.3013,
+      "step": 36050
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.892801370208051e-06,
+      "loss": 0.3115,
+      "step": 36051
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.892465534901684e-06,
+      "loss": 0.2293,
+      "step": 36052
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.892129699595319e-06,
+      "loss": 0.2692,
+      "step": 36053
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.891793864288954e-06,
+      "loss": 0.1804,
+      "step": 36054
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.891458028982587e-06,
+      "loss": 0.2269,
+      "step": 36055
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.891122193676221e-06,
+      "loss": 0.3628,
+      "step": 36056
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.890786358369856e-06,
+      "loss": 0.2827,
+      "step": 36057
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.89045052306349e-06,
+      "loss": 0.2225,
+      "step": 36058
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.890114687757125e-06,
+      "loss": 0.4341,
+      "step": 36059
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.88977885245076e-06,
+      "loss": 0.2515,
+      "step": 36060
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.889443017144393e-06,
+      "loss": 0.2769,
+      "step": 36061
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.889107181838027e-06,
+      "loss": 0.1568,
+      "step": 36062
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.888771346531662e-06,
+      "loss": 0.1928,
+      "step": 36063
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.888435511225295e-06,
+      "loss": 0.3103,
+      "step": 36064
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.88809967591893e-06,
+      "loss": 0.2216,
+      "step": 36065
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.887763840612564e-06,
+      "loss": 0.386,
+      "step": 36066
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.887428005306197e-06,
+      "loss": 0.3131,
+      "step": 36067
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.887092169999832e-06,
+      "loss": 0.1802,
+      "step": 36068
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.886756334693467e-06,
+      "loss": 0.3145,
+      "step": 36069
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.886420499387101e-06,
+      "loss": 0.277,
+      "step": 36070
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.886084664080736e-06,
+      "loss": 0.3006,
+      "step": 36071
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.88574882877437e-06,
+      "loss": 0.2721,
+      "step": 36072
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.885412993468003e-06,
+      "loss": 0.2966,
+      "step": 36073
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.885077158161638e-06,
+      "loss": 0.4162,
+      "step": 36074
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.884741322855273e-06,
+      "loss": 0.3287,
+      "step": 36075
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.884405487548906e-06,
+      "loss": 0.3891,
+      "step": 36076
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.88406965224254e-06,
+      "loss": 0.2255,
+      "step": 36077
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.883733816936175e-06,
+      "loss": 0.299,
+      "step": 36078
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.88339798162981e-06,
+      "loss": 0.3429,
+      "step": 36079
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.883062146323444e-06,
+      "loss": 0.4048,
+      "step": 36080
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.882726311017079e-06,
+      "loss": 0.2485,
+      "step": 36081
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.882390475710712e-06,
+      "loss": 0.4134,
+      "step": 36082
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.882054640404347e-06,
+      "loss": 0.1608,
+      "step": 36083
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.881718805097981e-06,
+      "loss": 0.3712,
+      "step": 36084
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.881382969791614e-06,
+      "loss": 0.2607,
+      "step": 36085
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.881047134485249e-06,
+      "loss": 0.4765,
+      "step": 36086
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.880711299178883e-06,
+      "loss": 0.3163,
+      "step": 36087
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.880375463872516e-06,
+      "loss": 0.2465,
+      "step": 36088
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.880039628566151e-06,
+      "loss": 0.2556,
+      "step": 36089
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.879703793259786e-06,
+      "loss": 0.2993,
+      "step": 36090
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.87936795795342e-06,
+      "loss": 0.4131,
+      "step": 36091
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.879032122647055e-06,
+      "loss": 0.2376,
+      "step": 36092
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.87869628734069e-06,
+      "loss": 0.1829,
+      "step": 36093
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.878360452034323e-06,
+      "loss": 0.2463,
+      "step": 36094
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.878024616727957e-06,
+      "loss": 0.2513,
+      "step": 36095
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.877688781421592e-06,
+      "loss": 0.2441,
+      "step": 36096
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.877352946115225e-06,
+      "loss": 0.371,
+      "step": 36097
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.87701711080886e-06,
+      "loss": 0.25,
+      "step": 36098
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.876681275502494e-06,
+      "loss": 0.2352,
+      "step": 36099
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.876345440196129e-06,
+      "loss": 0.2631,
+      "step": 36100
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.876009604889763e-06,
+      "loss": 0.2106,
+      "step": 36101
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.875673769583398e-06,
+      "loss": 0.2021,
+      "step": 36102
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.875337934277031e-06,
+      "loss": 0.236,
+      "step": 36103
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.875002098970666e-06,
+      "loss": 0.4253,
+      "step": 36104
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.8746662636643e-06,
+      "loss": 0.4764,
+      "step": 36105
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.874330428357933e-06,
+      "loss": 0.1949,
+      "step": 36106
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.873994593051568e-06,
+      "loss": 0.4643,
+      "step": 36107
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.873658757745203e-06,
+      "loss": 0.2265,
+      "step": 36108
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.873322922438837e-06,
+      "loss": 0.3005,
+      "step": 36109
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.87298708713247e-06,
+      "loss": 0.4092,
+      "step": 36110
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.872651251826105e-06,
+      "loss": 0.2597,
+      "step": 36111
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.87231541651974e-06,
+      "loss": 0.2631,
+      "step": 36112
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.871979581213374e-06,
+      "loss": 0.3754,
+      "step": 36113
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.871643745907009e-06,
+      "loss": 0.3024,
+      "step": 36114
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.871307910600642e-06,
+      "loss": 0.2686,
+      "step": 36115
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.870972075294276e-06,
+      "loss": 0.4458,
+      "step": 36116
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.870636239987911e-06,
+      "loss": 0.5477,
+      "step": 36117
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.870300404681544e-06,
+      "loss": 0.3618,
+      "step": 36118
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.869964569375179e-06,
+      "loss": 0.411,
+      "step": 36119
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.869628734068813e-06,
+      "loss": 0.3729,
+      "step": 36120
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.869292898762448e-06,
+      "loss": 0.2463,
+      "step": 36121
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.868957063456083e-06,
+      "loss": 0.3762,
+      "step": 36122
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.868621228149717e-06,
+      "loss": 0.2361,
+      "step": 36123
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.86828539284335e-06,
+      "loss": 0.3367,
+      "step": 36124
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.867949557536985e-06,
+      "loss": 0.3072,
+      "step": 36125
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.86761372223062e-06,
+      "loss": 0.2753,
+      "step": 36126
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.867277886924252e-06,
+      "loss": 0.3796,
+      "step": 36127
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.866942051617887e-06,
+      "loss": 0.3541,
+      "step": 36128
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.866606216311522e-06,
+      "loss": 0.3067,
+      "step": 36129
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.866270381005156e-06,
+      "loss": 0.2973,
+      "step": 36130
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.865934545698791e-06,
+      "loss": 0.2837,
+      "step": 36131
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.865598710392424e-06,
+      "loss": 0.334,
+      "step": 36132
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.865262875086059e-06,
+      "loss": 0.3655,
+      "step": 36133
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.864927039779693e-06,
+      "loss": 0.3468,
+      "step": 36134
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.864591204473328e-06,
+      "loss": 0.2374,
+      "step": 36135
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.86425536916696e-06,
+      "loss": 0.4369,
+      "step": 36136
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.863919533860595e-06,
+      "loss": 0.2007,
+      "step": 36137
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.863583698554228e-06,
+      "loss": 0.1781,
+      "step": 36138
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.863247863247863e-06,
+      "loss": 0.206,
+      "step": 36139
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.862912027941498e-06,
+      "loss": 0.3493,
+      "step": 36140
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.862576192635132e-06,
+      "loss": 0.3268,
+      "step": 36141
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.862240357328767e-06,
+      "loss": 0.4127,
+      "step": 36142
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.861904522022402e-06,
+      "loss": 0.294,
+      "step": 36143
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.861568686716035e-06,
+      "loss": 0.3569,
+      "step": 36144
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.86123285140967e-06,
+      "loss": 0.2478,
+      "step": 36145
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.860897016103304e-06,
+      "loss": 0.1637,
+      "step": 36146
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.860561180796937e-06,
+      "loss": 0.3506,
+      "step": 36147
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.860225345490571e-06,
+      "loss": 0.4342,
+      "step": 36148
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.859889510184206e-06,
+      "loss": 0.2847,
+      "step": 36149
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.85955367487784e-06,
+      "loss": 0.2498,
+      "step": 36150
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.859217839571475e-06,
+      "loss": 0.2046,
+      "step": 36151
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.85888200426511e-06,
+      "loss": 0.2138,
+      "step": 36152
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.858546168958743e-06,
+      "loss": 0.2111,
+      "step": 36153
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.858210333652378e-06,
+      "loss": 0.267,
+      "step": 36154
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.857874498346012e-06,
+      "loss": 0.3525,
+      "step": 36155
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.857538663039645e-06,
+      "loss": 0.3668,
+      "step": 36156
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.85720282773328e-06,
+      "loss": 0.2156,
+      "step": 36157
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.856866992426915e-06,
+      "loss": 0.2276,
+      "step": 36158
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.856531157120548e-06,
+      "loss": 0.1774,
+      "step": 36159
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.856195321814182e-06,
+      "loss": 0.2478,
+      "step": 36160
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.855859486507817e-06,
+      "loss": 0.3687,
+      "step": 36161
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.855523651201451e-06,
+      "loss": 0.2426,
+      "step": 36162
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.855187815895086e-06,
+      "loss": 0.3253,
+      "step": 36163
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.85485198058872e-06,
+      "loss": 0.2107,
+      "step": 36164
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.854516145282354e-06,
+      "loss": 0.2127,
+      "step": 36165
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.854180309975988e-06,
+      "loss": 0.2835,
+      "step": 36166
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.853844474669623e-06,
+      "loss": 0.2914,
+      "step": 36167
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.853508639363256e-06,
+      "loss": 0.3271,
+      "step": 36168
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.85317280405689e-06,
+      "loss": 0.3021,
+      "step": 36169
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.852836968750525e-06,
+      "loss": 0.3414,
+      "step": 36170
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.85250113344416e-06,
+      "loss": 0.1798,
+      "step": 36171
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.852165298137795e-06,
+      "loss": 0.1799,
+      "step": 36172
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.85182946283143e-06,
+      "loss": 0.3289,
+      "step": 36173
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.851493627525062e-06,
+      "loss": 0.363,
+      "step": 36174
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.851157792218697e-06,
+      "loss": 0.1838,
+      "step": 36175
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.850821956912331e-06,
+      "loss": 0.2158,
+      "step": 36176
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.850486121605964e-06,
+      "loss": 0.1472,
+      "step": 36177
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.850150286299599e-06,
+      "loss": 0.3209,
+      "step": 36178
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.849814450993234e-06,
+      "loss": 0.225,
+      "step": 36179
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.849478615686867e-06,
+      "loss": 0.1401,
+      "step": 36180
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.849142780380501e-06,
+      "loss": 0.2061,
+      "step": 36181
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.848806945074136e-06,
+      "loss": 0.483,
+      "step": 36182
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.84847110976777e-06,
+      "loss": 0.2806,
+      "step": 36183
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.848135274461405e-06,
+      "loss": 0.3538,
+      "step": 36184
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.84779943915504e-06,
+      "loss": 0.3134,
+      "step": 36185
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.847463603848673e-06,
+      "loss": 0.2242,
+      "step": 36186
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.847127768542307e-06,
+      "loss": 0.2907,
+      "step": 36187
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.846791933235942e-06,
+      "loss": 0.2761,
+      "step": 36188
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.846456097929575e-06,
+      "loss": 0.2031,
+      "step": 36189
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.84612026262321e-06,
+      "loss": 0.2483,
+      "step": 36190
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.845784427316844e-06,
+      "loss": 0.3265,
+      "step": 36191
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.845448592010479e-06,
+      "loss": 0.2499,
+      "step": 36192
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.845112756704114e-06,
+      "loss": 0.2703,
+      "step": 36193
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.844776921397748e-06,
+      "loss": 0.3501,
+      "step": 36194
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.844441086091381e-06,
+      "loss": 0.3487,
+      "step": 36195
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.844105250785016e-06,
+      "loss": 0.398,
+      "step": 36196
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.84376941547865e-06,
+      "loss": 0.3973,
+      "step": 36197
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.843433580172284e-06,
+      "loss": 0.1245,
+      "step": 36198
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.843097744865918e-06,
+      "loss": 0.323,
+      "step": 36199
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.842761909559553e-06,
+      "loss": 0.3082,
+      "step": 36200
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.842426074253186e-06,
+      "loss": 0.318,
+      "step": 36201
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.84209023894682e-06,
+      "loss": 0.2604,
+      "step": 36202
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.841754403640455e-06,
+      "loss": 0.1812,
+      "step": 36203
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.84141856833409e-06,
+      "loss": 0.3808,
+      "step": 36204
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.841082733027724e-06,
+      "loss": 0.1966,
+      "step": 36205
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.840746897721359e-06,
+      "loss": 0.2808,
+      "step": 36206
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.840411062414992e-06,
+      "loss": 0.2831,
+      "step": 36207
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.840075227108627e-06,
+      "loss": 0.2316,
+      "step": 36208
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.839739391802261e-06,
+      "loss": 0.4099,
+      "step": 36209
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.839403556495894e-06,
+      "loss": 0.51,
+      "step": 36210
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.839067721189529e-06,
+      "loss": 0.5321,
+      "step": 36211
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.838731885883164e-06,
+      "loss": 0.4861,
+      "step": 36212
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.838396050576798e-06,
+      "loss": 0.4968,
+      "step": 36213
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.838060215270433e-06,
+      "loss": 0.2266,
+      "step": 36214
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.837724379964067e-06,
+      "loss": 0.2277,
+      "step": 36215
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.8373885446577e-06,
+      "loss": 0.3062,
+      "step": 36216
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.837052709351335e-06,
+      "loss": 0.2959,
+      "step": 36217
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.83671687404497e-06,
+      "loss": 0.2512,
+      "step": 36218
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.836381038738603e-06,
+      "loss": 0.3231,
+      "step": 36219
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.836045203432237e-06,
+      "loss": 0.3829,
+      "step": 36220
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.835709368125872e-06,
+      "loss": 0.2358,
+      "step": 36221
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.835373532819505e-06,
+      "loss": 0.4014,
+      "step": 36222
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.83503769751314e-06,
+      "loss": 0.2169,
+      "step": 36223
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.834701862206774e-06,
+      "loss": 0.2194,
+      "step": 36224
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.834366026900409e-06,
+      "loss": 0.2173,
+      "step": 36225
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.834030191594043e-06,
+      "loss": 0.2146,
+      "step": 36226
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.833694356287678e-06,
+      "loss": 0.1577,
+      "step": 36227
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 7.833358520981311e-06,
+      "loss": 0.1628,
+      "step": 36228
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.833022685674946e-06,
+      "loss": 0.4092,
+      "step": 36229
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.832686850368579e-06,
+      "loss": 0.3469,
+      "step": 36230
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.832351015062213e-06,
+      "loss": 0.3597,
+      "step": 36231
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.832015179755848e-06,
+      "loss": 0.3724,
+      "step": 36232
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.831679344449483e-06,
+      "loss": 0.2132,
+      "step": 36233
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.831343509143117e-06,
+      "loss": 0.3006,
+      "step": 36234
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.831007673836752e-06,
+      "loss": 0.1573,
+      "step": 36235
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.830671838530385e-06,
+      "loss": 0.3357,
+      "step": 36236
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.83033600322402e-06,
+      "loss": 0.3139,
+      "step": 36237
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.830000167917654e-06,
+      "loss": 0.2049,
+      "step": 36238
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.829664332611287e-06,
+      "loss": 0.2338,
+      "step": 36239
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.829328497304922e-06,
+      "loss": 0.2899,
+      "step": 36240
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.828992661998556e-06,
+      "loss": 0.3165,
+      "step": 36241
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.828656826692191e-06,
+      "loss": 0.3195,
+      "step": 36242
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.828320991385826e-06,
+      "loss": 0.2064,
+      "step": 36243
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.82798515607946e-06,
+      "loss": 0.2061,
+      "step": 36244
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.827649320773093e-06,
+      "loss": 0.3152,
+      "step": 36245
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.827313485466728e-06,
+      "loss": 0.5632,
+      "step": 36246
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.826977650160363e-06,
+      "loss": 0.3132,
+      "step": 36247
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.826641814853996e-06,
+      "loss": 0.1827,
+      "step": 36248
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.82630597954763e-06,
+      "loss": 0.2071,
+      "step": 36249
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.825970144241265e-06,
+      "loss": 0.3143,
+      "step": 36250
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.825634308934898e-06,
+      "loss": 0.324,
+      "step": 36251
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.825298473628532e-06,
+      "loss": 0.2919,
+      "step": 36252
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.824962638322167e-06,
+      "loss": 0.345,
+      "step": 36253
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.824626803015802e-06,
+      "loss": 0.1489,
+      "step": 36254
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.824290967709436e-06,
+      "loss": 0.3387,
+      "step": 36255
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.823955132403071e-06,
+      "loss": 0.1828,
+      "step": 36256
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.823619297096704e-06,
+      "loss": 0.3245,
+      "step": 36257
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.823283461790339e-06,
+      "loss": 0.34,
+      "step": 36258
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.822947626483973e-06,
+      "loss": 0.2291,
+      "step": 36259
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.822611791177606e-06,
+      "loss": 0.4371,
+      "step": 36260
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.822275955871241e-06,
+      "loss": 0.2941,
+      "step": 36261
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.821940120564876e-06,
+      "loss": 0.5103,
+      "step": 36262
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.82160428525851e-06,
+      "loss": 0.3441,
+      "step": 36263
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.821268449952145e-06,
+      "loss": 0.1618,
+      "step": 36264
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.82093261464578e-06,
+      "loss": 0.2808,
+      "step": 36265
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.820596779339412e-06,
+      "loss": 0.2886,
+      "step": 36266
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.820260944033047e-06,
+      "loss": 0.3048,
+      "step": 36267
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.819925108726682e-06,
+      "loss": 0.2375,
+      "step": 36268
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.819589273420315e-06,
+      "loss": 0.2995,
+      "step": 36269
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.81925343811395e-06,
+      "loss": 0.3061,
+      "step": 36270
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.818917602807584e-06,
+      "loss": 0.3858,
+      "step": 36271
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.818581767501217e-06,
+      "loss": 0.1983,
+      "step": 36272
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.818245932194852e-06,
+      "loss": 0.2116,
+      "step": 36273
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.817910096888486e-06,
+      "loss": 0.2058,
+      "step": 36274
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.817574261582121e-06,
+      "loss": 0.2616,
+      "step": 36275
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.817238426275756e-06,
+      "loss": 0.3375,
+      "step": 36276
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.81690259096939e-06,
+      "loss": 0.2348,
+      "step": 36277
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.816566755663023e-06,
+      "loss": 0.2797,
+      "step": 36278
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.816230920356658e-06,
+      "loss": 0.2766,
+      "step": 36279
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.815895085050292e-06,
+      "loss": 0.2772,
+      "step": 36280
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.815559249743925e-06,
+      "loss": 0.3628,
+      "step": 36281
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.81522341443756e-06,
+      "loss": 0.3213,
+      "step": 36282
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.814887579131195e-06,
+      "loss": 0.5071,
+      "step": 36283
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.81455174382483e-06,
+      "loss": 0.2187,
+      "step": 36284
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.814215908518464e-06,
+      "loss": 0.4086,
+      "step": 36285
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.813880073212099e-06,
+      "loss": 0.4319,
+      "step": 36286
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.813544237905732e-06,
+      "loss": 0.4769,
+      "step": 36287
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.813208402599366e-06,
+      "loss": 0.1603,
+      "step": 36288
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.812872567293001e-06,
+      "loss": 0.2619,
+      "step": 36289
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.812536731986634e-06,
+      "loss": 0.3379,
+      "step": 36290
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.812200896680268e-06,
+      "loss": 0.2029,
+      "step": 36291
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.811865061373903e-06,
+      "loss": 0.1976,
+      "step": 36292
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.811529226067536e-06,
+      "loss": 0.1724,
+      "step": 36293
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.81119339076117e-06,
+      "loss": 0.2923,
+      "step": 36294
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.810857555454805e-06,
+      "loss": 0.326,
+      "step": 36295
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.81052172014844e-06,
+      "loss": 0.2416,
+      "step": 36296
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.810185884842075e-06,
+      "loss": 0.2231,
+      "step": 36297
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.80985004953571e-06,
+      "loss": 0.4265,
+      "step": 36298
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.809514214229342e-06,
+      "loss": 0.3007,
+      "step": 36299
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.809178378922977e-06,
+      "loss": 0.2588,
+      "step": 36300
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.808842543616612e-06,
+      "loss": 0.3472,
+      "step": 36301
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.808506708310245e-06,
+      "loss": 0.4859,
+      "step": 36302
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.80817087300388e-06,
+      "loss": 0.284,
+      "step": 36303
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.807835037697514e-06,
+      "loss": 0.2478,
+      "step": 36304
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.807499202391148e-06,
+      "loss": 0.2791,
+      "step": 36305
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.807163367084783e-06,
+      "loss": 0.4002,
+      "step": 36306
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.806827531778418e-06,
+      "loss": 0.2784,
+      "step": 36307
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.80649169647205e-06,
+      "loss": 0.235,
+      "step": 36308
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.806155861165685e-06,
+      "loss": 0.2314,
+      "step": 36309
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.80582002585932e-06,
+      "loss": 0.3404,
+      "step": 36310
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.805484190552953e-06,
+      "loss": 0.3097,
+      "step": 36311
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.805148355246588e-06,
+      "loss": 0.2979,
+      "step": 36312
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.804812519940222e-06,
+      "loss": 0.2808,
+      "step": 36313
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.804476684633855e-06,
+      "loss": 0.2348,
+      "step": 36314
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.80414084932749e-06,
+      "loss": 0.2915,
+      "step": 36315
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.803805014021124e-06,
+      "loss": 0.2652,
+      "step": 36316
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.803469178714759e-06,
+      "loss": 0.2434,
+      "step": 36317
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.803133343408394e-06,
+      "loss": 0.2872,
+      "step": 36318
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.802797508102027e-06,
+      "loss": 0.1184,
+      "step": 36319
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.802461672795661e-06,
+      "loss": 0.2502,
+      "step": 36320
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.802125837489296e-06,
+      "loss": 0.3142,
+      "step": 36321
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.801790002182929e-06,
+      "loss": 0.1718,
+      "step": 36322
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.801454166876564e-06,
+      "loss": 0.2813,
+      "step": 36323
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.801118331570198e-06,
+      "loss": 0.2939,
+      "step": 36324
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.800782496263833e-06,
+      "loss": 0.334,
+      "step": 36325
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.800446660957468e-06,
+      "loss": 0.446,
+      "step": 36326
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.800110825651102e-06,
+      "loss": 0.2724,
+      "step": 36327
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.799774990344735e-06,
+      "loss": 0.2635,
+      "step": 36328
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.79943915503837e-06,
+      "loss": 0.2608,
+      "step": 36329
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.799103319732004e-06,
+      "loss": 0.3174,
+      "step": 36330
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.798767484425637e-06,
+      "loss": 0.2301,
+      "step": 36331
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.798431649119272e-06,
+      "loss": 0.2923,
+      "step": 36332
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.798095813812907e-06,
+      "loss": 0.2802,
+      "step": 36333
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.797759978506541e-06,
+      "loss": 0.3953,
+      "step": 36334
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.797424143200174e-06,
+      "loss": 0.3455,
+      "step": 36335
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.797088307893809e-06,
+      "loss": 0.2545,
+      "step": 36336
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.796752472587444e-06,
+      "loss": 0.2715,
+      "step": 36337
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.796416637281078e-06,
+      "loss": 0.2158,
+      "step": 36338
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.796080801974713e-06,
+      "loss": 0.2499,
+      "step": 36339
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.795744966668346e-06,
+      "loss": 0.3053,
+      "step": 36340
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.79540913136198e-06,
+      "loss": 0.2129,
+      "step": 36341
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.795073296055615e-06,
+      "loss": 0.2973,
+      "step": 36342
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.794737460749248e-06,
+      "loss": 0.21,
+      "step": 36343
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.794401625442883e-06,
+      "loss": 0.2958,
+      "step": 36344
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.794065790136517e-06,
+      "loss": 0.3256,
+      "step": 36345
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.793729954830152e-06,
+      "loss": 0.3414,
+      "step": 36346
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.793394119523787e-06,
+      "loss": 0.2999,
+      "step": 36347
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.793058284217421e-06,
+      "loss": 0.3107,
+      "step": 36348
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.792722448911054e-06,
+      "loss": 0.5093,
+      "step": 36349
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.792386613604689e-06,
+      "loss": 0.3341,
+      "step": 36350
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.792050778298324e-06,
+      "loss": 0.3132,
+      "step": 36351
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.791714942991957e-06,
+      "loss": 0.2077,
+      "step": 36352
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.791379107685591e-06,
+      "loss": 0.3181,
+      "step": 36353
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.791043272379226e-06,
+      "loss": 0.2444,
+      "step": 36354
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.79070743707286e-06,
+      "loss": 0.2039,
+      "step": 36355
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.790371601766495e-06,
+      "loss": 0.339,
+      "step": 36356
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.790035766460128e-06,
+      "loss": 0.4038,
+      "step": 36357
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.789699931153763e-06,
+      "loss": 0.2249,
+      "step": 36358
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.789364095847397e-06,
+      "loss": 0.3038,
+      "step": 36359
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.789028260541032e-06,
+      "loss": 0.268,
+      "step": 36360
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.788692425234665e-06,
+      "loss": 0.2052,
+      "step": 36361
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.7883565899283e-06,
+      "loss": 0.2769,
+      "step": 36362
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.788020754621934e-06,
+      "loss": 0.1843,
+      "step": 36363
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.787684919315567e-06,
+      "loss": 0.5091,
+      "step": 36364
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.787349084009202e-06,
+      "loss": 0.2396,
+      "step": 36365
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.787013248702837e-06,
+      "loss": 0.2091,
+      "step": 36366
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.786677413396471e-06,
+      "loss": 0.4078,
+      "step": 36367
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.786341578090106e-06,
+      "loss": 0.1365,
+      "step": 36368
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.78600574278374e-06,
+      "loss": 0.2796,
+      "step": 36369
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.785669907477373e-06,
+      "loss": 0.2765,
+      "step": 36370
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.785334072171008e-06,
+      "loss": 0.224,
+      "step": 36371
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.784998236864643e-06,
+      "loss": 0.2556,
+      "step": 36372
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.784662401558276e-06,
+      "loss": 0.3346,
+      "step": 36373
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.78432656625191e-06,
+      "loss": 0.3213,
+      "step": 36374
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.783990730945545e-06,
+      "loss": 0.3388,
+      "step": 36375
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.78365489563918e-06,
+      "loss": 0.3265,
+      "step": 36376
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.783319060332814e-06,
+      "loss": 0.2286,
+      "step": 36377
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.782983225026449e-06,
+      "loss": 0.3348,
+      "step": 36378
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.782647389720082e-06,
+      "loss": 0.3783,
+      "step": 36379
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.782311554413717e-06,
+      "loss": 0.1955,
+      "step": 36380
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.781975719107351e-06,
+      "loss": 0.399,
+      "step": 36381
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.781639883800984e-06,
+      "loss": 0.4367,
+      "step": 36382
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.781304048494619e-06,
+      "loss": 0.4072,
+      "step": 36383
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.780968213188253e-06,
+      "loss": 0.4035,
+      "step": 36384
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.780632377881886e-06,
+      "loss": 0.2299,
+      "step": 36385
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.780296542575521e-06,
+      "loss": 0.1239,
+      "step": 36386
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.779960707269156e-06,
+      "loss": 0.2949,
+      "step": 36387
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.77962487196279e-06,
+      "loss": 0.3136,
+      "step": 36388
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.779289036656425e-06,
+      "loss": 0.1716,
+      "step": 36389
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.77895320135006e-06,
+      "loss": 0.2333,
+      "step": 36390
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.778617366043693e-06,
+      "loss": 0.2023,
+      "step": 36391
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.778281530737327e-06,
+      "loss": 0.2297,
+      "step": 36392
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.777945695430962e-06,
+      "loss": 0.2986,
+      "step": 36393
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.777609860124595e-06,
+      "loss": 0.1877,
+      "step": 36394
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.77727402481823e-06,
+      "loss": 0.3051,
+      "step": 36395
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.776938189511864e-06,
+      "loss": 0.3389,
+      "step": 36396
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.776602354205499e-06,
+      "loss": 0.2562,
+      "step": 36397
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.776266518899133e-06,
+      "loss": 0.3191,
+      "step": 36398
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.775930683592768e-06,
+      "loss": 0.4351,
+      "step": 36399
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.775594848286401e-06,
+      "loss": 0.1534,
+      "step": 36400
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.775259012980036e-06,
+      "loss": 0.1762,
+      "step": 36401
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.77492317767367e-06,
+      "loss": 0.2513,
+      "step": 36402
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.774587342367303e-06,
+      "loss": 0.238,
+      "step": 36403
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.774251507060938e-06,
+      "loss": 0.345,
+      "step": 36404
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.773915671754573e-06,
+      "loss": 0.2282,
+      "step": 36405
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.773579836448205e-06,
+      "loss": 0.2438,
+      "step": 36406
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.77324400114184e-06,
+      "loss": 0.2286,
+      "step": 36407
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.772908165835475e-06,
+      "loss": 0.4152,
+      "step": 36408
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.77257233052911e-06,
+      "loss": 0.2253,
+      "step": 36409
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.772236495222744e-06,
+      "loss": 0.2806,
+      "step": 36410
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.771900659916377e-06,
+      "loss": 0.1891,
+      "step": 36411
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.771564824610012e-06,
+      "loss": 0.3201,
+      "step": 36412
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.771228989303646e-06,
+      "loss": 0.1858,
+      "step": 36413
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.77089315399728e-06,
+      "loss": 0.1787,
+      "step": 36414
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.770557318690914e-06,
+      "loss": 0.342,
+      "step": 36415
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.770221483384549e-06,
+      "loss": 0.2593,
+      "step": 36416
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.769885648078183e-06,
+      "loss": 0.216,
+      "step": 36417
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.769549812771818e-06,
+      "loss": 0.275,
+      "step": 36418
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.769213977465453e-06,
+      "loss": 0.2746,
+      "step": 36419
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.768878142159085e-06,
+      "loss": 0.2852,
+      "step": 36420
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.76854230685272e-06,
+      "loss": 0.3646,
+      "step": 36421
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.768206471546355e-06,
+      "loss": 0.1847,
+      "step": 36422
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.767870636239988e-06,
+      "loss": 0.2242,
+      "step": 36423
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.767534800933622e-06,
+      "loss": 0.1786,
+      "step": 36424
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.767198965627257e-06,
+      "loss": 0.364,
+      "step": 36425
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 7.76686313032089e-06,
+      "loss": 0.3075,
+      "step": 36426
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.766527295014525e-06,
+      "loss": 0.2887,
+      "step": 36427
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.76619145970816e-06,
+      "loss": 0.2863,
+      "step": 36428
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.765855624401794e-06,
+      "loss": 0.2798,
+      "step": 36429
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.765519789095429e-06,
+      "loss": 0.2463,
+      "step": 36430
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.765183953789063e-06,
+      "loss": 0.2809,
+      "step": 36431
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.764848118482696e-06,
+      "loss": 0.2475,
+      "step": 36432
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.76451228317633e-06,
+      "loss": 0.2799,
+      "step": 36433
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.764176447869965e-06,
+      "loss": 0.2621,
+      "step": 36434
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.763840612563598e-06,
+      "loss": 0.5063,
+      "step": 36435
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.763504777257233e-06,
+      "loss": 0.2495,
+      "step": 36436
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.763168941950868e-06,
+      "loss": 0.2892,
+      "step": 36437
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.762833106644502e-06,
+      "loss": 0.282,
+      "step": 36438
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.762497271338137e-06,
+      "loss": 0.3401,
+      "step": 36439
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.762161436031772e-06,
+      "loss": 0.3299,
+      "step": 36440
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.761825600725405e-06,
+      "loss": 0.128,
+      "step": 36441
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.76148976541904e-06,
+      "loss": 0.236,
+      "step": 36442
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.761153930112674e-06,
+      "loss": 0.2814,
+      "step": 36443
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.760818094806307e-06,
+      "loss": 0.2466,
+      "step": 36444
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.760482259499941e-06,
+      "loss": 0.3295,
+      "step": 36445
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.760146424193576e-06,
+      "loss": 0.2472,
+      "step": 36446
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.75981058888721e-06,
+      "loss": 0.2887,
+      "step": 36447
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.759474753580844e-06,
+      "loss": 0.5474,
+      "step": 36448
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.759138918274478e-06,
+      "loss": 0.2711,
+      "step": 36449
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.758803082968113e-06,
+      "loss": 0.218,
+      "step": 36450
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.758467247661748e-06,
+      "loss": 0.2745,
+      "step": 36451
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.758131412355382e-06,
+      "loss": 0.2686,
+      "step": 36452
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.757795577049015e-06,
+      "loss": 0.311,
+      "step": 36453
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.75745974174265e-06,
+      "loss": 0.2234,
+      "step": 36454
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.757123906436285e-06,
+      "loss": 0.3341,
+      "step": 36455
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.756788071129918e-06,
+      "loss": 0.2768,
+      "step": 36456
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.756452235823552e-06,
+      "loss": 0.4052,
+      "step": 36457
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.756116400517187e-06,
+      "loss": 0.3433,
+      "step": 36458
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.755780565210821e-06,
+      "loss": 0.3691,
+      "step": 36459
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.755444729904456e-06,
+      "loss": 0.2465,
+      "step": 36460
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.75510889459809e-06,
+      "loss": 0.4028,
+      "step": 36461
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.754773059291724e-06,
+      "loss": 0.2265,
+      "step": 36462
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.754437223985358e-06,
+      "loss": 0.3285,
+      "step": 36463
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.754101388678993e-06,
+      "loss": 0.2623,
+      "step": 36464
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.753765553372626e-06,
+      "loss": 0.2405,
+      "step": 36465
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.75342971806626e-06,
+      "loss": 0.3039,
+      "step": 36466
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.753093882759895e-06,
+      "loss": 0.4369,
+      "step": 36467
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.75275804745353e-06,
+      "loss": 0.2431,
+      "step": 36468
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.752422212147165e-06,
+      "loss": 0.2734,
+      "step": 36469
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.752086376840798e-06,
+      "loss": 0.3303,
+      "step": 36470
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.751750541534432e-06,
+      "loss": 0.2758,
+      "step": 36471
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.751414706228067e-06,
+      "loss": 0.1982,
+      "step": 36472
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.751078870921701e-06,
+      "loss": 0.3518,
+      "step": 36473
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.750743035615334e-06,
+      "loss": 0.336,
+      "step": 36474
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.750407200308969e-06,
+      "loss": 0.3598,
+      "step": 36475
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.750071365002604e-06,
+      "loss": 0.2895,
+      "step": 36476
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.749735529696237e-06,
+      "loss": 0.2687,
+      "step": 36477
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.749399694389871e-06,
+      "loss": 0.2647,
+      "step": 36478
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.749063859083506e-06,
+      "loss": 0.29,
+      "step": 36479
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.74872802377714e-06,
+      "loss": 0.2291,
+      "step": 36480
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.748392188470775e-06,
+      "loss": 0.1999,
+      "step": 36481
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.74805635316441e-06,
+      "loss": 0.3293,
+      "step": 36482
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.747720517858043e-06,
+      "loss": 0.2156,
+      "step": 36483
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.747384682551677e-06,
+      "loss": 0.3687,
+      "step": 36484
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.747048847245312e-06,
+      "loss": 0.5501,
+      "step": 36485
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.746713011938945e-06,
+      "loss": 0.2185,
+      "step": 36486
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.74637717663258e-06,
+      "loss": 0.1879,
+      "step": 36487
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.746041341326214e-06,
+      "loss": 0.4116,
+      "step": 36488
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.745705506019849e-06,
+      "loss": 0.3196,
+      "step": 36489
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.745369670713484e-06,
+      "loss": 0.3034,
+      "step": 36490
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.745033835407118e-06,
+      "loss": 0.2969,
+      "step": 36491
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.744698000100751e-06,
+      "loss": 0.3528,
+      "step": 36492
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.744362164794386e-06,
+      "loss": 0.14,
+      "step": 36493
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.74402632948802e-06,
+      "loss": 0.4739,
+      "step": 36494
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.743690494181654e-06,
+      "loss": 0.1734,
+      "step": 36495
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.743354658875288e-06,
+      "loss": 0.3631,
+      "step": 36496
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.743018823568923e-06,
+      "loss": 0.2302,
+      "step": 36497
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.742682988262556e-06,
+      "loss": 0.2502,
+      "step": 36498
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.74234715295619e-06,
+      "loss": 0.2451,
+      "step": 36499
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.742011317649825e-06,
+      "loss": 0.4499,
+      "step": 36500
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.74167548234346e-06,
+      "loss": 0.2433,
+      "step": 36501
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.741339647037094e-06,
+      "loss": 0.2693,
+      "step": 36502
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.741003811730727e-06,
+      "loss": 0.1626,
+      "step": 36503
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.740667976424362e-06,
+      "loss": 0.3329,
+      "step": 36504
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.740332141117997e-06,
+      "loss": 0.2579,
+      "step": 36505
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.73999630581163e-06,
+      "loss": 0.3034,
+      "step": 36506
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.739660470505264e-06,
+      "loss": 0.2946,
+      "step": 36507
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.739324635198899e-06,
+      "loss": 0.2516,
+      "step": 36508
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.738988799892534e-06,
+      "loss": 0.2251,
+      "step": 36509
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.738652964586168e-06,
+      "loss": 0.2274,
+      "step": 36510
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.738317129279803e-06,
+      "loss": 0.1636,
+      "step": 36511
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.737981293973436e-06,
+      "loss": 0.4012,
+      "step": 36512
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.73764545866707e-06,
+      "loss": 0.3574,
+      "step": 36513
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.737309623360705e-06,
+      "loss": 0.1927,
+      "step": 36514
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.736973788054338e-06,
+      "loss": 0.2215,
+      "step": 36515
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.736637952747973e-06,
+      "loss": 0.3996,
+      "step": 36516
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.736302117441607e-06,
+      "loss": 0.3058,
+      "step": 36517
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.73596628213524e-06,
+      "loss": 0.4285,
+      "step": 36518
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.735630446828875e-06,
+      "loss": 0.1803,
+      "step": 36519
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.73529461152251e-06,
+      "loss": 0.3825,
+      "step": 36520
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.734958776216144e-06,
+      "loss": 0.2951,
+      "step": 36521
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.734622940909779e-06,
+      "loss": 0.2935,
+      "step": 36522
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.734287105603413e-06,
+      "loss": 0.3607,
+      "step": 36523
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.733951270297046e-06,
+      "loss": 0.3635,
+      "step": 36524
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.733615434990681e-06,
+      "loss": 0.3049,
+      "step": 36525
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.733279599684316e-06,
+      "loss": 0.3559,
+      "step": 36526
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.732943764377949e-06,
+      "loss": 0.2005,
+      "step": 36527
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.732607929071583e-06,
+      "loss": 0.2199,
+      "step": 36528
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.732272093765218e-06,
+      "loss": 0.2495,
+      "step": 36529
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.731936258458853e-06,
+      "loss": 0.2528,
+      "step": 36530
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.731600423152487e-06,
+      "loss": 0.1909,
+      "step": 36531
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.731264587846122e-06,
+      "loss": 0.3398,
+      "step": 36532
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.730928752539755e-06,
+      "loss": 0.1387,
+      "step": 36533
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.73059291723339e-06,
+      "loss": 0.3048,
+      "step": 36534
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.730257081927024e-06,
+      "loss": 0.3664,
+      "step": 36535
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.729921246620657e-06,
+      "loss": 0.1021,
+      "step": 36536
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.729585411314292e-06,
+      "loss": 0.2431,
+      "step": 36537
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.729249576007926e-06,
+      "loss": 0.3124,
+      "step": 36538
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.72891374070156e-06,
+      "loss": 0.3902,
+      "step": 36539
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.728577905395194e-06,
+      "loss": 0.4406,
+      "step": 36540
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.728242070088829e-06,
+      "loss": 0.3919,
+      "step": 36541
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.727906234782463e-06,
+      "loss": 0.3287,
+      "step": 36542
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.727570399476098e-06,
+      "loss": 0.1594,
+      "step": 36543
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.727234564169733e-06,
+      "loss": 0.2066,
+      "step": 36544
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.726898728863366e-06,
+      "loss": 0.2119,
+      "step": 36545
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.726562893557e-06,
+      "loss": 0.219,
+      "step": 36546
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.726227058250635e-06,
+      "loss": 0.3274,
+      "step": 36547
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.725891222944268e-06,
+      "loss": 0.3079,
+      "step": 36548
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.725555387637902e-06,
+      "loss": 0.411,
+      "step": 36549
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.725219552331537e-06,
+      "loss": 0.1415,
+      "step": 36550
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.724883717025172e-06,
+      "loss": 0.1444,
+      "step": 36551
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.724547881718806e-06,
+      "loss": 0.214,
+      "step": 36552
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.724212046412441e-06,
+      "loss": 0.2795,
+      "step": 36553
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.723876211106074e-06,
+      "loss": 0.3076,
+      "step": 36554
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.723540375799709e-06,
+      "loss": 0.5824,
+      "step": 36555
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.723204540493343e-06,
+      "loss": 0.2649,
+      "step": 36556
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.722868705186976e-06,
+      "loss": 0.2649,
+      "step": 36557
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.722532869880611e-06,
+      "loss": 0.2023,
+      "step": 36558
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.722197034574246e-06,
+      "loss": 0.3063,
+      "step": 36559
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.72186119926788e-06,
+      "loss": 0.2945,
+      "step": 36560
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.721525363961513e-06,
+      "loss": 0.2316,
+      "step": 36561
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.721189528655148e-06,
+      "loss": 0.3754,
+      "step": 36562
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.720853693348782e-06,
+      "loss": 0.2504,
+      "step": 36563
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.720517858042417e-06,
+      "loss": 0.2315,
+      "step": 36564
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.720182022736052e-06,
+      "loss": 0.3304,
+      "step": 36565
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.719846187429685e-06,
+      "loss": 0.299,
+      "step": 36566
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.71951035212332e-06,
+      "loss": 0.5612,
+      "step": 36567
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.719174516816954e-06,
+      "loss": 0.2155,
+      "step": 36568
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.718838681510587e-06,
+      "loss": 0.3353,
+      "step": 36569
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.718502846204222e-06,
+      "loss": 0.2998,
+      "step": 36570
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.718167010897856e-06,
+      "loss": 0.304,
+      "step": 36571
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.717831175591491e-06,
+      "loss": 0.2207,
+      "step": 36572
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.717495340285126e-06,
+      "loss": 0.5057,
+      "step": 36573
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.71715950497876e-06,
+      "loss": 0.2268,
+      "step": 36574
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.716823669672393e-06,
+      "loss": 0.368,
+      "step": 36575
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.716487834366028e-06,
+      "loss": 0.3284,
+      "step": 36576
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.716151999059662e-06,
+      "loss": 0.3321,
+      "step": 36577
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.715816163753295e-06,
+      "loss": 0.2713,
+      "step": 36578
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.71548032844693e-06,
+      "loss": 0.3119,
+      "step": 36579
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.715144493140565e-06,
+      "loss": 0.2389,
+      "step": 36580
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.7148086578342e-06,
+      "loss": 0.3379,
+      "step": 36581
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.714472822527834e-06,
+      "loss": 0.2986,
+      "step": 36582
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.714136987221467e-06,
+      "loss": 0.1537,
+      "step": 36583
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.713801151915102e-06,
+      "loss": 0.1994,
+      "step": 36584
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.713465316608736e-06,
+      "loss": 0.3736,
+      "step": 36585
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.713129481302371e-06,
+      "loss": 0.4047,
+      "step": 36586
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.712793645996004e-06,
+      "loss": 0.2262,
+      "step": 36587
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.712457810689638e-06,
+      "loss": 0.3497,
+      "step": 36588
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.712121975383273e-06,
+      "loss": 0.2502,
+      "step": 36589
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.711786140076906e-06,
+      "loss": 0.2606,
+      "step": 36590
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.71145030477054e-06,
+      "loss": 0.2066,
+      "step": 36591
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.711114469464175e-06,
+      "loss": 0.3875,
+      "step": 36592
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.71077863415781e-06,
+      "loss": 0.2838,
+      "step": 36593
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.710442798851445e-06,
+      "loss": 0.2011,
+      "step": 36594
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.710106963545078e-06,
+      "loss": 0.2852,
+      "step": 36595
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.709771128238712e-06,
+      "loss": 0.394,
+      "step": 36596
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.709435292932347e-06,
+      "loss": 0.3926,
+      "step": 36597
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.70909945762598e-06,
+      "loss": 0.3234,
+      "step": 36598
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.708763622319615e-06,
+      "loss": 0.179,
+      "step": 36599
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.708427787013249e-06,
+      "loss": 0.1576,
+      "step": 36600
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.708091951706884e-06,
+      "loss": 0.2768,
+      "step": 36601
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.707756116400518e-06,
+      "loss": 0.2062,
+      "step": 36602
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.707420281094153e-06,
+      "loss": 0.1849,
+      "step": 36603
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.707084445787786e-06,
+      "loss": 0.3331,
+      "step": 36604
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.70674861048142e-06,
+      "loss": 0.2877,
+      "step": 36605
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.706412775175055e-06,
+      "loss": 0.3372,
+      "step": 36606
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.706076939868688e-06,
+      "loss": 0.2616,
+      "step": 36607
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.705741104562323e-06,
+      "loss": 0.2199,
+      "step": 36608
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.705405269255958e-06,
+      "loss": 0.2972,
+      "step": 36609
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.70506943394959e-06,
+      "loss": 0.2391,
+      "step": 36610
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.704733598643225e-06,
+      "loss": 0.1674,
+      "step": 36611
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.70439776333686e-06,
+      "loss": 0.3956,
+      "step": 36612
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.704061928030494e-06,
+      "loss": 0.1842,
+      "step": 36613
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.703726092724129e-06,
+      "loss": 0.3303,
+      "step": 36614
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.703390257417764e-06,
+      "loss": 0.2903,
+      "step": 36615
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.703054422111397e-06,
+      "loss": 0.3522,
+      "step": 36616
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.702718586805031e-06,
+      "loss": 0.2958,
+      "step": 36617
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.702382751498666e-06,
+      "loss": 0.2175,
+      "step": 36618
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.702046916192299e-06,
+      "loss": 0.3318,
+      "step": 36619
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.701711080885934e-06,
+      "loss": 0.2567,
+      "step": 36620
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.701375245579568e-06,
+      "loss": 0.2994,
+      "step": 36621
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.701039410273203e-06,
+      "loss": 0.2135,
+      "step": 36622
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.700703574966838e-06,
+      "loss": 0.2404,
+      "step": 36623
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.700367739660472e-06,
+      "loss": 0.156,
+      "step": 36624
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 7.700031904354105e-06,
+      "loss": 0.2337,
+      "step": 36625
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.69969606904774e-06,
+      "loss": 0.2821,
+      "step": 36626
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.699360233741374e-06,
+      "loss": 0.2881,
+      "step": 36627
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.699024398435007e-06,
+      "loss": 0.3055,
+      "step": 36628
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.698688563128642e-06,
+      "loss": 0.2825,
+      "step": 36629
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.698352727822277e-06,
+      "loss": 0.1278,
+      "step": 36630
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.69801689251591e-06,
+      "loss": 0.1045,
+      "step": 36631
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.697681057209544e-06,
+      "loss": 0.4261,
+      "step": 36632
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.697345221903179e-06,
+      "loss": 0.2254,
+      "step": 36633
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.697009386596814e-06,
+      "loss": 0.3636,
+      "step": 36634
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.696673551290448e-06,
+      "loss": 0.2326,
+      "step": 36635
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.696337715984083e-06,
+      "loss": 0.4445,
+      "step": 36636
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.696001880677716e-06,
+      "loss": 0.2601,
+      "step": 36637
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.69566604537135e-06,
+      "loss": 0.3626,
+      "step": 36638
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.695330210064985e-06,
+      "loss": 0.145,
+      "step": 36639
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.694994374758618e-06,
+      "loss": 0.4445,
+      "step": 36640
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.694658539452253e-06,
+      "loss": 0.2935,
+      "step": 36641
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.694322704145887e-06,
+      "loss": 0.3594,
+      "step": 36642
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.693986868839522e-06,
+      "loss": 0.2147,
+      "step": 36643
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.693651033533157e-06,
+      "loss": 0.3089,
+      "step": 36644
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.693315198226791e-06,
+      "loss": 0.4038,
+      "step": 36645
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.692979362920424e-06,
+      "loss": 0.3356,
+      "step": 36646
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.692643527614059e-06,
+      "loss": 0.341,
+      "step": 36647
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.692307692307694e-06,
+      "loss": 0.2887,
+      "step": 36648
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.691971857001327e-06,
+      "loss": 0.242,
+      "step": 36649
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.691636021694961e-06,
+      "loss": 0.1765,
+      "step": 36650
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.691300186388596e-06,
+      "loss": 0.2736,
+      "step": 36651
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.690964351082229e-06,
+      "loss": 0.3374,
+      "step": 36652
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.690628515775863e-06,
+      "loss": 0.3186,
+      "step": 36653
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.690292680469498e-06,
+      "loss": 0.2457,
+      "step": 36654
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.689956845163133e-06,
+      "loss": 0.2635,
+      "step": 36655
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.689621009856767e-06,
+      "loss": 0.2699,
+      "step": 36656
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.689285174550402e-06,
+      "loss": 0.498,
+      "step": 36657
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.688949339244035e-06,
+      "loss": 0.1044,
+      "step": 36658
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.68861350393767e-06,
+      "loss": 0.2549,
+      "step": 36659
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.688277668631304e-06,
+      "loss": 0.1533,
+      "step": 36660
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.687941833324937e-06,
+      "loss": 0.3151,
+      "step": 36661
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.687605998018572e-06,
+      "loss": 0.3069,
+      "step": 36662
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.687270162712207e-06,
+      "loss": 0.3205,
+      "step": 36663
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.686934327405841e-06,
+      "loss": 0.2045,
+      "step": 36664
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.686598492099476e-06,
+      "loss": 0.267,
+      "step": 36665
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.68626265679311e-06,
+      "loss": 0.2412,
+      "step": 36666
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.685926821486743e-06,
+      "loss": 0.3001,
+      "step": 36667
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.685590986180378e-06,
+      "loss": 0.3679,
+      "step": 36668
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.685255150874013e-06,
+      "loss": 0.4029,
+      "step": 36669
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.684919315567646e-06,
+      "loss": 0.3178,
+      "step": 36670
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.68458348026128e-06,
+      "loss": 0.2799,
+      "step": 36671
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.684247644954915e-06,
+      "loss": 0.1515,
+      "step": 36672
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.68391180964855e-06,
+      "loss": 0.3413,
+      "step": 36673
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.683575974342183e-06,
+      "loss": 0.2248,
+      "step": 36674
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.683240139035817e-06,
+      "loss": 0.2031,
+      "step": 36675
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.682904303729452e-06,
+      "loss": 0.2508,
+      "step": 36676
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.682568468423087e-06,
+      "loss": 0.3054,
+      "step": 36677
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.682232633116721e-06,
+      "loss": 0.4134,
+      "step": 36678
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.681896797810354e-06,
+      "loss": 0.306,
+      "step": 36679
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.681560962503989e-06,
+      "loss": 0.4576,
+      "step": 36680
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.681225127197622e-06,
+      "loss": 0.1722,
+      "step": 36681
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.680889291891256e-06,
+      "loss": 0.204,
+      "step": 36682
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.680553456584891e-06,
+      "loss": 0.687,
+      "step": 36683
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.680217621278526e-06,
+      "loss": 0.1779,
+      "step": 36684
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.67988178597216e-06,
+      "loss": 0.152,
+      "step": 36685
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.679545950665795e-06,
+      "loss": 0.3107,
+      "step": 36686
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.679210115359428e-06,
+      "loss": 0.252,
+      "step": 36687
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.678874280053063e-06,
+      "loss": 0.2105,
+      "step": 36688
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.678538444746697e-06,
+      "loss": 0.1739,
+      "step": 36689
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.67820260944033e-06,
+      "loss": 0.3004,
+      "step": 36690
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.677866774133965e-06,
+      "loss": 0.319,
+      "step": 36691
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.6775309388276e-06,
+      "loss": 0.2973,
+      "step": 36692
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.677195103521234e-06,
+      "loss": 0.323,
+      "step": 36693
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.676859268214869e-06,
+      "loss": 0.332,
+      "step": 36694
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.676523432908503e-06,
+      "loss": 0.2114,
+      "step": 36695
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.676187597602136e-06,
+      "loss": 0.46,
+      "step": 36696
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.675851762295771e-06,
+      "loss": 0.2966,
+      "step": 36697
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.675515926989406e-06,
+      "loss": 0.2366,
+      "step": 36698
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.675180091683039e-06,
+      "loss": 0.2344,
+      "step": 36699
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.674844256376673e-06,
+      "loss": 0.1447,
+      "step": 36700
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.674508421070308e-06,
+      "loss": 0.1781,
+      "step": 36701
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.67417258576394e-06,
+      "loss": 0.2864,
+      "step": 36702
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.673836750457575e-06,
+      "loss": 0.2354,
+      "step": 36703
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.67350091515121e-06,
+      "loss": 0.16,
+      "step": 36704
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.673165079844845e-06,
+      "loss": 0.2917,
+      "step": 36705
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.67282924453848e-06,
+      "loss": 0.3976,
+      "step": 36706
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.672493409232114e-06,
+      "loss": 0.1858,
+      "step": 36707
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.672157573925747e-06,
+      "loss": 0.328,
+      "step": 36708
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.671821738619382e-06,
+      "loss": 0.2868,
+      "step": 36709
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.671485903313016e-06,
+      "loss": 0.3315,
+      "step": 36710
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.67115006800665e-06,
+      "loss": 0.3605,
+      "step": 36711
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.670814232700284e-06,
+      "loss": 0.575,
+      "step": 36712
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.670478397393919e-06,
+      "loss": 0.291,
+      "step": 36713
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.670142562087553e-06,
+      "loss": 0.3,
+      "step": 36714
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.669806726781188e-06,
+      "loss": 0.2626,
+      "step": 36715
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.669470891474823e-06,
+      "loss": 0.4036,
+      "step": 36716
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.669135056168455e-06,
+      "loss": 0.2494,
+      "step": 36717
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.66879922086209e-06,
+      "loss": 0.1772,
+      "step": 36718
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.668463385555725e-06,
+      "loss": 0.2379,
+      "step": 36719
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.668127550249358e-06,
+      "loss": 0.3752,
+      "step": 36720
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.667791714942992e-06,
+      "loss": 0.3058,
+      "step": 36721
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.667455879636627e-06,
+      "loss": 0.5147,
+      "step": 36722
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.66712004433026e-06,
+      "loss": 0.2653,
+      "step": 36723
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.666784209023895e-06,
+      "loss": 0.2352,
+      "step": 36724
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.66644837371753e-06,
+      "loss": 0.4602,
+      "step": 36725
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.666112538411164e-06,
+      "loss": 0.4222,
+      "step": 36726
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.665776703104799e-06,
+      "loss": 0.2704,
+      "step": 36727
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.665440867798433e-06,
+      "loss": 0.2779,
+      "step": 36728
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.665105032492066e-06,
+      "loss": 0.1938,
+      "step": 36729
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.6647691971857e-06,
+      "loss": 0.366,
+      "step": 36730
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.664433361879335e-06,
+      "loss": 0.185,
+      "step": 36731
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.664097526572968e-06,
+      "loss": 0.2792,
+      "step": 36732
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.663761691266603e-06,
+      "loss": 0.2262,
+      "step": 36733
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.663425855960238e-06,
+      "loss": 0.1806,
+      "step": 36734
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.663090020653872e-06,
+      "loss": 0.3065,
+      "step": 36735
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.662754185347507e-06,
+      "loss": 0.3186,
+      "step": 36736
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.662418350041142e-06,
+      "loss": 0.4685,
+      "step": 36737
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.662082514734775e-06,
+      "loss": 0.4267,
+      "step": 36738
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.66174667942841e-06,
+      "loss": 0.2977,
+      "step": 36739
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.661410844122044e-06,
+      "loss": 0.2416,
+      "step": 36740
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.661075008815677e-06,
+      "loss": 0.3732,
+      "step": 36741
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.660739173509311e-06,
+      "loss": 0.2119,
+      "step": 36742
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.660403338202946e-06,
+      "loss": 0.2668,
+      "step": 36743
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.660067502896579e-06,
+      "loss": 0.3014,
+      "step": 36744
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.659731667590214e-06,
+      "loss": 0.4128,
+      "step": 36745
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.659395832283848e-06,
+      "loss": 0.3905,
+      "step": 36746
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.659059996977483e-06,
+      "loss": 0.4092,
+      "step": 36747
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.658724161671118e-06,
+      "loss": 0.533,
+      "step": 36748
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.658388326364752e-06,
+      "loss": 0.2594,
+      "step": 36749
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.658052491058385e-06,
+      "loss": 0.1512,
+      "step": 36750
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.65771665575202e-06,
+      "loss": 0.2796,
+      "step": 36751
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.657380820445655e-06,
+      "loss": 0.3551,
+      "step": 36752
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.657044985139288e-06,
+      "loss": 0.3456,
+      "step": 36753
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.656709149832922e-06,
+      "loss": 0.3443,
+      "step": 36754
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.656373314526557e-06,
+      "loss": 0.2188,
+      "step": 36755
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.656037479220191e-06,
+      "loss": 0.1274,
+      "step": 36756
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.655701643913826e-06,
+      "loss": 0.1787,
+      "step": 36757
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.65536580860746e-06,
+      "loss": 0.2004,
+      "step": 36758
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.655029973301094e-06,
+      "loss": 0.329,
+      "step": 36759
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.654694137994728e-06,
+      "loss": 0.3125,
+      "step": 36760
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.654358302688363e-06,
+      "loss": 0.271,
+      "step": 36761
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.654022467381996e-06,
+      "loss": 0.301,
+      "step": 36762
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.65368663207563e-06,
+      "loss": 0.2098,
+      "step": 36763
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.653350796769265e-06,
+      "loss": 0.2201,
+      "step": 36764
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.653014961462898e-06,
+      "loss": 0.5254,
+      "step": 36765
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.652679126156533e-06,
+      "loss": 0.2092,
+      "step": 36766
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.652343290850167e-06,
+      "loss": 0.2842,
+      "step": 36767
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.652007455543802e-06,
+      "loss": 0.2066,
+      "step": 36768
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.651671620237437e-06,
+      "loss": 0.1897,
+      "step": 36769
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.65133578493107e-06,
+      "loss": 0.2408,
+      "step": 36770
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.650999949624704e-06,
+      "loss": 0.378,
+      "step": 36771
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.650664114318339e-06,
+      "loss": 0.3935,
+      "step": 36772
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.650328279011972e-06,
+      "loss": 0.3788,
+      "step": 36773
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.649992443705607e-06,
+      "loss": 0.3381,
+      "step": 36774
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.649656608399241e-06,
+      "loss": 0.4117,
+      "step": 36775
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.649320773092876e-06,
+      "loss": 0.201,
+      "step": 36776
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.64898493778651e-06,
+      "loss": 0.2032,
+      "step": 36777
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.648649102480145e-06,
+      "loss": 0.2804,
+      "step": 36778
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.648313267173778e-06,
+      "loss": 0.2526,
+      "step": 36779
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.647977431867413e-06,
+      "loss": 0.4097,
+      "step": 36780
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.647641596561047e-06,
+      "loss": 0.3303,
+      "step": 36781
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.64730576125468e-06,
+      "loss": 0.2642,
+      "step": 36782
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.646969925948315e-06,
+      "loss": 0.2202,
+      "step": 36783
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.64663409064195e-06,
+      "loss": 0.2734,
+      "step": 36784
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.646298255335584e-06,
+      "loss": 0.5437,
+      "step": 36785
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.645962420029219e-06,
+      "loss": 0.2856,
+      "step": 36786
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.645626584722852e-06,
+      "loss": 0.3318,
+      "step": 36787
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.645290749416487e-06,
+      "loss": 0.2897,
+      "step": 36788
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.644954914110121e-06,
+      "loss": 0.2384,
+      "step": 36789
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.644619078803756e-06,
+      "loss": 0.3019,
+      "step": 36790
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.644283243497389e-06,
+      "loss": 0.3624,
+      "step": 36791
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.643947408191024e-06,
+      "loss": 0.3204,
+      "step": 36792
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.643611572884658e-06,
+      "loss": 0.3161,
+      "step": 36793
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.643275737578291e-06,
+      "loss": 0.4044,
+      "step": 36794
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.642939902271926e-06,
+      "loss": 0.3242,
+      "step": 36795
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.64260406696556e-06,
+      "loss": 0.1415,
+      "step": 36796
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.642268231659195e-06,
+      "loss": 0.31,
+      "step": 36797
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.64193239635283e-06,
+      "loss": 0.2767,
+      "step": 36798
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.641596561046464e-06,
+      "loss": 0.1738,
+      "step": 36799
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.641260725740097e-06,
+      "loss": 0.2422,
+      "step": 36800
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.640924890433732e-06,
+      "loss": 0.2888,
+      "step": 36801
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.640589055127367e-06,
+      "loss": 0.1892,
+      "step": 36802
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.640253219821e-06,
+      "loss": 0.1982,
+      "step": 36803
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.639917384514634e-06,
+      "loss": 0.3429,
+      "step": 36804
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.639581549208269e-06,
+      "loss": 0.2088,
+      "step": 36805
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.639245713901904e-06,
+      "loss": 0.2076,
+      "step": 36806
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.638909878595538e-06,
+      "loss": 0.1701,
+      "step": 36807
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.638574043289173e-06,
+      "loss": 0.2583,
+      "step": 36808
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.638238207982806e-06,
+      "loss": 0.1524,
+      "step": 36809
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.63790237267644e-06,
+      "loss": 0.4701,
+      "step": 36810
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.637566537370075e-06,
+      "loss": 0.403,
+      "step": 36811
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.637230702063708e-06,
+      "loss": 0.1895,
+      "step": 36812
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.636894866757343e-06,
+      "loss": 0.3,
+      "step": 36813
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.636559031450977e-06,
+      "loss": 0.3052,
+      "step": 36814
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.63622319614461e-06,
+      "loss": 0.2405,
+      "step": 36815
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.635887360838245e-06,
+      "loss": 0.4909,
+      "step": 36816
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.63555152553188e-06,
+      "loss": 0.215,
+      "step": 36817
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.635215690225514e-06,
+      "loss": 0.187,
+      "step": 36818
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.634879854919149e-06,
+      "loss": 0.2535,
+      "step": 36819
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.634544019612783e-06,
+      "loss": 0.2916,
+      "step": 36820
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.634208184306416e-06,
+      "loss": 0.3668,
+      "step": 36821
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.633872349000051e-06,
+      "loss": 0.3496,
+      "step": 36822
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 7.633536513693686e-06,
+      "loss": 0.1668,
+      "step": 36823
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.633200678387319e-06,
+      "loss": 0.0775,
+      "step": 36824
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.632864843080953e-06,
+      "loss": 0.3027,
+      "step": 36825
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.632529007774588e-06,
+      "loss": 0.232,
+      "step": 36826
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.632193172468223e-06,
+      "loss": 0.1839,
+      "step": 36827
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.631857337161857e-06,
+      "loss": 0.2694,
+      "step": 36828
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.631521501855492e-06,
+      "loss": 0.2054,
+      "step": 36829
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.631185666549125e-06,
+      "loss": 0.3145,
+      "step": 36830
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.63084983124276e-06,
+      "loss": 0.3022,
+      "step": 36831
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.630513995936394e-06,
+      "loss": 0.2854,
+      "step": 36832
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.630178160630027e-06,
+      "loss": 0.248,
+      "step": 36833
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.629842325323662e-06,
+      "loss": 0.3022,
+      "step": 36834
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.629506490017296e-06,
+      "loss": 0.2229,
+      "step": 36835
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.62917065471093e-06,
+      "loss": 0.3106,
+      "step": 36836
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.628834819404565e-06,
+      "loss": 0.3329,
+      "step": 36837
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.6284989840981995e-06,
+      "loss": 0.3035,
+      "step": 36838
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.6281631487918325e-06,
+      "loss": 0.2735,
+      "step": 36839
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.627827313485467e-06,
+      "loss": 0.1838,
+      "step": 36840
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.627491478179102e-06,
+      "loss": 0.2736,
+      "step": 36841
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.6271556428727356e-06,
+      "loss": 0.4274,
+      "step": 36842
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.62681980756637e-06,
+      "loss": 0.2312,
+      "step": 36843
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.626483972260005e-06,
+      "loss": 0.4513,
+      "step": 36844
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.626148136953639e-06,
+      "loss": 0.1659,
+      "step": 36845
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.625812301647273e-06,
+      "loss": 0.4598,
+      "step": 36846
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.625476466340908e-06,
+      "loss": 0.4204,
+      "step": 36847
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.625140631034541e-06,
+      "loss": 0.1673,
+      "step": 36848
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.6248047957281756e-06,
+      "loss": 0.1876,
+      "step": 36849
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.62446896042181e-06,
+      "loss": 0.2433,
+      "step": 36850
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.624133125115444e-06,
+      "loss": 0.2767,
+      "step": 36851
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.623797289809079e-06,
+      "loss": 0.4313,
+      "step": 36852
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.623461454502713e-06,
+      "loss": 0.1925,
+      "step": 36853
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.623125619196346e-06,
+      "loss": 0.264,
+      "step": 36854
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.622789783889981e-06,
+      "loss": 0.2577,
+      "step": 36855
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.6224539485836155e-06,
+      "loss": 0.3785,
+      "step": 36856
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.622118113277249e-06,
+      "loss": 0.215,
+      "step": 36857
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.621782277970884e-06,
+      "loss": 0.3797,
+      "step": 36858
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.621446442664519e-06,
+      "loss": 0.272,
+      "step": 36859
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.621110607358152e-06,
+      "loss": 0.2348,
+      "step": 36860
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.620774772051786e-06,
+      "loss": 0.2471,
+      "step": 36861
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.62043893674542e-06,
+      "loss": 0.227,
+      "step": 36862
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.620103101439055e-06,
+      "loss": 0.4073,
+      "step": 36863
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.619767266132689e-06,
+      "loss": 0.2108,
+      "step": 36864
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.619431430826323e-06,
+      "loss": 0.3254,
+      "step": 36865
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.619095595519958e-06,
+      "loss": 0.2267,
+      "step": 36866
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.6187597602135924e-06,
+      "loss": 0.3029,
+      "step": 36867
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.618423924907225e-06,
+      "loss": 0.2574,
+      "step": 36868
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.61808808960086e-06,
+      "loss": 0.2652,
+      "step": 36869
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.617752254294495e-06,
+      "loss": 0.2411,
+      "step": 36870
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.6174164189881285e-06,
+      "loss": 0.4691,
+      "step": 36871
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.617080583681763e-06,
+      "loss": 0.4008,
+      "step": 36872
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.616744748375398e-06,
+      "loss": 0.2204,
+      "step": 36873
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.616408913069031e-06,
+      "loss": 0.4603,
+      "step": 36874
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.616073077762665e-06,
+      "loss": 0.3134,
+      "step": 36875
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.6157372424563e-06,
+      "loss": 0.2861,
+      "step": 36876
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.615401407149934e-06,
+      "loss": 0.282,
+      "step": 36877
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.6150655718435685e-06,
+      "loss": 0.3163,
+      "step": 36878
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.614729736537203e-06,
+      "loss": 0.3388,
+      "step": 36879
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.614393901230837e-06,
+      "loss": 0.426,
+      "step": 36880
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.6140580659244716e-06,
+      "loss": 0.3543,
+      "step": 36881
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.613722230618105e-06,
+      "loss": 0.481,
+      "step": 36882
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.613386395311739e-06,
+      "loss": 0.4565,
+      "step": 36883
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.613050560005374e-06,
+      "loss": 0.2791,
+      "step": 36884
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.6127147246990085e-06,
+      "loss": 0.1864,
+      "step": 36885
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.612378889392642e-06,
+      "loss": 0.3418,
+      "step": 36886
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.612043054086277e-06,
+      "loss": 0.4623,
+      "step": 36887
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.6117072187799116e-06,
+      "loss": 0.3104,
+      "step": 36888
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.6113713834735445e-06,
+      "loss": 0.2453,
+      "step": 36889
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.611035548167179e-06,
+      "loss": 0.366,
+      "step": 36890
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.610699712860814e-06,
+      "loss": 0.2998,
+      "step": 36891
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.610363877554448e-06,
+      "loss": 0.2433,
+      "step": 36892
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.610028042248082e-06,
+      "loss": 0.1623,
+      "step": 36893
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.609692206941717e-06,
+      "loss": 0.3875,
+      "step": 36894
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.60935637163535e-06,
+      "loss": 0.2511,
+      "step": 36895
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.6090205363289845e-06,
+      "loss": 0.14,
+      "step": 36896
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.608684701022619e-06,
+      "loss": 0.3929,
+      "step": 36897
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.608348865716253e-06,
+      "loss": 0.4517,
+      "step": 36898
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.608013030409888e-06,
+      "loss": 0.319,
+      "step": 36899
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.607677195103522e-06,
+      "loss": 0.2466,
+      "step": 36900
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.607341359797156e-06,
+      "loss": 0.3086,
+      "step": 36901
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.607005524490791e-06,
+      "loss": 0.3602,
+      "step": 36902
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.606669689184425e-06,
+      "loss": 0.3236,
+      "step": 36903
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.606333853878058e-06,
+      "loss": 0.3839,
+      "step": 36904
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.605998018571693e-06,
+      "loss": 0.183,
+      "step": 36905
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.605662183265328e-06,
+      "loss": 0.3059,
+      "step": 36906
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.605326347958961e-06,
+      "loss": 0.2363,
+      "step": 36907
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.604990512652596e-06,
+      "loss": 0.1837,
+      "step": 36908
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.604654677346231e-06,
+      "loss": 0.3679,
+      "step": 36909
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.604318842039864e-06,
+      "loss": 0.4146,
+      "step": 36910
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.603983006733498e-06,
+      "loss": 0.2726,
+      "step": 36911
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.603647171427133e-06,
+      "loss": 0.3477,
+      "step": 36912
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.603311336120767e-06,
+      "loss": 0.2928,
+      "step": 36913
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.602975500814401e-06,
+      "loss": 0.3624,
+      "step": 36914
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.602639665508036e-06,
+      "loss": 0.2658,
+      "step": 36915
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.60230383020167e-06,
+      "loss": 0.1504,
+      "step": 36916
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.601967994895304e-06,
+      "loss": 0.2322,
+      "step": 36917
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.601632159588938e-06,
+      "loss": 0.2581,
+      "step": 36918
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.601296324282572e-06,
+      "loss": 0.2794,
+      "step": 36919
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.600960488976207e-06,
+      "loss": 0.4155,
+      "step": 36920
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.600624653669841e-06,
+      "loss": 0.1554,
+      "step": 36921
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.600288818363475e-06,
+      "loss": 0.4388,
+      "step": 36922
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.59995298305711e-06,
+      "loss": 0.3892,
+      "step": 36923
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.5996171477507445e-06,
+      "loss": 0.2775,
+      "step": 36924
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.599281312444377e-06,
+      "loss": 0.1943,
+      "step": 36925
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.598945477138012e-06,
+      "loss": 0.2396,
+      "step": 36926
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.598609641831647e-06,
+      "loss": 0.3804,
+      "step": 36927
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.5982738065252805e-06,
+      "loss": 0.3448,
+      "step": 36928
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.597937971218915e-06,
+      "loss": 0.2326,
+      "step": 36929
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.59760213591255e-06,
+      "loss": 0.3277,
+      "step": 36930
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.597266300606183e-06,
+      "loss": 0.2677,
+      "step": 36931
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.596930465299817e-06,
+      "loss": 0.2994,
+      "step": 36932
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.596594629993452e-06,
+      "loss": 0.3211,
+      "step": 36933
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.596258794687086e-06,
+      "loss": 0.3628,
+      "step": 36934
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.5959229593807205e-06,
+      "loss": 0.1447,
+      "step": 36935
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.595587124074355e-06,
+      "loss": 0.2126,
+      "step": 36936
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.595251288767989e-06,
+      "loss": 0.2079,
+      "step": 36937
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.594915453461624e-06,
+      "loss": 0.3931,
+      "step": 36938
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.594579618155257e-06,
+      "loss": 0.2799,
+      "step": 36939
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.594243782848891e-06,
+      "loss": 0.2557,
+      "step": 36940
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.593907947542526e-06,
+      "loss": 0.2861,
+      "step": 36941
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.5935721122361605e-06,
+      "loss": 0.1574,
+      "step": 36942
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.593236276929794e-06,
+      "loss": 0.1047,
+      "step": 36943
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.592900441623429e-06,
+      "loss": 0.3922,
+      "step": 36944
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.592564606317064e-06,
+      "loss": 0.35,
+      "step": 36945
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.5922287710106965e-06,
+      "loss": 0.1372,
+      "step": 36946
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.591892935704331e-06,
+      "loss": 0.3794,
+      "step": 36947
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.591557100397966e-06,
+      "loss": 0.2533,
+      "step": 36948
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.5912212650916e-06,
+      "loss": 0.303,
+      "step": 36949
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.590885429785234e-06,
+      "loss": 0.2598,
+      "step": 36950
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.590549594478867e-06,
+      "loss": 0.2629,
+      "step": 36951
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.590213759172502e-06,
+      "loss": 0.3376,
+      "step": 36952
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.5898779238661365e-06,
+      "loss": 0.3399,
+      "step": 36953
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.58954208855977e-06,
+      "loss": 0.3502,
+      "step": 36954
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.589206253253405e-06,
+      "loss": 0.3491,
+      "step": 36955
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.58887041794704e-06,
+      "loss": 0.2382,
+      "step": 36956
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.588534582640673e-06,
+      "loss": 0.356,
+      "step": 36957
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.588198747334308e-06,
+      "loss": 0.4286,
+      "step": 36958
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.587862912027943e-06,
+      "loss": 0.5634,
+      "step": 36959
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.587527076721576e-06,
+      "loss": 0.3523,
+      "step": 36960
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.58719124141521e-06,
+      "loss": 0.1844,
+      "step": 36961
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.586855406108845e-06,
+      "loss": 0.1727,
+      "step": 36962
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.586519570802479e-06,
+      "loss": 0.4258,
+      "step": 36963
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.586183735496113e-06,
+      "loss": 0.2921,
+      "step": 36964
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.585847900189748e-06,
+      "loss": 0.2278,
+      "step": 36965
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.585512064883381e-06,
+      "loss": 0.3638,
+      "step": 36966
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.585176229577016e-06,
+      "loss": 0.2323,
+      "step": 36967
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.58484039427065e-06,
+      "loss": 0.2504,
+      "step": 36968
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.584504558964284e-06,
+      "loss": 0.3614,
+      "step": 36969
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.584168723657919e-06,
+      "loss": 0.2697,
+      "step": 36970
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.583832888351553e-06,
+      "loss": 0.1887,
+      "step": 36971
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.583497053045187e-06,
+      "loss": 0.3104,
+      "step": 36972
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.583161217738821e-06,
+      "loss": 0.453,
+      "step": 36973
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.582825382432456e-06,
+      "loss": 0.1666,
+      "step": 36974
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.5824895471260895e-06,
+      "loss": 0.3496,
+      "step": 36975
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.582153711819724e-06,
+      "loss": 0.26,
+      "step": 36976
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.581817876513359e-06,
+      "loss": 0.2617,
+      "step": 36977
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.5814820412069925e-06,
+      "loss": 0.2037,
+      "step": 36978
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.581146205900627e-06,
+      "loss": 0.178,
+      "step": 36979
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.580810370594262e-06,
+      "loss": 0.2477,
+      "step": 36980
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.580474535287895e-06,
+      "loss": 0.3272,
+      "step": 36981
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.5801386999815294e-06,
+      "loss": 0.1542,
+      "step": 36982
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.579802864675164e-06,
+      "loss": 0.1731,
+      "step": 36983
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.579467029368798e-06,
+      "loss": 0.3128,
+      "step": 36984
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.5791311940624325e-06,
+      "loss": 0.43,
+      "step": 36985
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.578795358756067e-06,
+      "loss": 0.2311,
+      "step": 36986
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.5784595234497e-06,
+      "loss": 0.476,
+      "step": 36987
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.578123688143335e-06,
+      "loss": 0.2715,
+      "step": 36988
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.5777878528369694e-06,
+      "loss": 0.3007,
+      "step": 36989
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.577452017530603e-06,
+      "loss": 0.2453,
+      "step": 36990
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.577116182224238e-06,
+      "loss": 0.1992,
+      "step": 36991
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.5767803469178725e-06,
+      "loss": 0.2522,
+      "step": 36992
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.576444511611506e-06,
+      "loss": 0.29,
+      "step": 36993
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.576108676305141e-06,
+      "loss": 0.4198,
+      "step": 36994
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.575772840998775e-06,
+      "loss": 0.2852,
+      "step": 36995
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.575437005692409e-06,
+      "loss": 0.1326,
+      "step": 36996
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.575101170386043e-06,
+      "loss": 0.214,
+      "step": 36997
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.574765335079678e-06,
+      "loss": 0.156,
+      "step": 36998
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.574429499773312e-06,
+      "loss": 0.385,
+      "step": 36999
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.574093664466946e-06,
+      "loss": 0.2528,
+      "step": 37000
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.573757829160581e-06,
+      "loss": 0.2469,
+      "step": 37001
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.573421993854214e-06,
+      "loss": 0.1935,
+      "step": 37002
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.5730861585478486e-06,
+      "loss": 0.322,
+      "step": 37003
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.572750323241483e-06,
+      "loss": 0.2956,
+      "step": 37004
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.572414487935117e-06,
+      "loss": 0.2381,
+      "step": 37005
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.572078652628752e-06,
+      "loss": 0.3204,
+      "step": 37006
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.571742817322386e-06,
+      "loss": 0.2561,
+      "step": 37007
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.571406982016019e-06,
+      "loss": 0.3025,
+      "step": 37008
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.571071146709654e-06,
+      "loss": 0.2091,
+      "step": 37009
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.5707353114032886e-06,
+      "loss": 0.4277,
+      "step": 37010
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.570399476096922e-06,
+      "loss": 0.2102,
+      "step": 37011
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.570063640790557e-06,
+      "loss": 0.3714,
+      "step": 37012
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.569727805484192e-06,
+      "loss": 0.2413,
+      "step": 37013
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.5693919701778255e-06,
+      "loss": 0.2369,
+      "step": 37014
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.56905613487146e-06,
+      "loss": 0.1901,
+      "step": 37015
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.568720299565095e-06,
+      "loss": 0.4599,
+      "step": 37016
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.568384464258728e-06,
+      "loss": 0.3081,
+      "step": 37017
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.568048628952362e-06,
+      "loss": 0.3004,
+      "step": 37018
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.567712793645997e-06,
+      "loss": 0.3745,
+      "step": 37019
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.567376958339631e-06,
+      "loss": 0.4326,
+      "step": 37020
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.5670411230332654e-06,
+      "loss": 0.1158,
+      "step": 37021
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 7.5667052877269e-06,
+      "loss": 0.1644,
+      "step": 37022
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.566369452420533e-06,
+      "loss": 0.2709,
+      "step": 37023
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.566033617114168e-06,
+      "loss": 0.2286,
+      "step": 37024
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.565697781807802e-06,
+      "loss": 0.2397,
+      "step": 37025
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.565361946501436e-06,
+      "loss": 0.255,
+      "step": 37026
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.565026111195071e-06,
+      "loss": 0.3793,
+      "step": 37027
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5646902758887054e-06,
+      "loss": 0.2231,
+      "step": 37028
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.564354440582338e-06,
+      "loss": 0.433,
+      "step": 37029
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.564018605275973e-06,
+      "loss": 0.2791,
+      "step": 37030
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.563682769969608e-06,
+      "loss": 0.2733,
+      "step": 37031
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5633469346632415e-06,
+      "loss": 0.4146,
+      "step": 37032
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.563011099356876e-06,
+      "loss": 0.2072,
+      "step": 37033
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.562675264050511e-06,
+      "loss": 0.4014,
+      "step": 37034
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.562339428744145e-06,
+      "loss": 0.2996,
+      "step": 37035
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.562003593437779e-06,
+      "loss": 0.2433,
+      "step": 37036
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.561667758131414e-06,
+      "loss": 0.4019,
+      "step": 37037
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.561331922825047e-06,
+      "loss": 0.2277,
+      "step": 37038
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5609960875186815e-06,
+      "loss": 0.1181,
+      "step": 37039
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.560660252212316e-06,
+      "loss": 0.1486,
+      "step": 37040
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.56032441690595e-06,
+      "loss": 0.3281,
+      "step": 37041
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5599885815995846e-06,
+      "loss": 0.2847,
+      "step": 37042
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5596527462932175e-06,
+      "loss": 0.1879,
+      "step": 37043
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.559316910986852e-06,
+      "loss": 0.3653,
+      "step": 37044
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.558981075680487e-06,
+      "loss": 0.2935,
+      "step": 37045
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.558645240374121e-06,
+      "loss": 0.1628,
+      "step": 37046
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.558309405067755e-06,
+      "loss": 0.2599,
+      "step": 37047
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.55797356976139e-06,
+      "loss": 0.3242,
+      "step": 37048
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.557637734455024e-06,
+      "loss": 0.3238,
+      "step": 37049
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.557301899148658e-06,
+      "loss": 0.2867,
+      "step": 37050
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.556966063842293e-06,
+      "loss": 0.2412,
+      "step": 37051
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.556630228535926e-06,
+      "loss": 0.2884,
+      "step": 37052
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.556294393229561e-06,
+      "loss": 0.376,
+      "step": 37053
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.555958557923195e-06,
+      "loss": 0.5066,
+      "step": 37054
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.555622722616829e-06,
+      "loss": 0.2822,
+      "step": 37055
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.555286887310464e-06,
+      "loss": 0.2297,
+      "step": 37056
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.554951052004098e-06,
+      "loss": 0.2432,
+      "step": 37057
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.554615216697731e-06,
+      "loss": 0.2901,
+      "step": 37058
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.554279381391366e-06,
+      "loss": 0.2337,
+      "step": 37059
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.553943546085001e-06,
+      "loss": 0.2126,
+      "step": 37060
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.553607710778634e-06,
+      "loss": 0.5806,
+      "step": 37061
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.553271875472269e-06,
+      "loss": 0.1931,
+      "step": 37062
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.552936040165904e-06,
+      "loss": 0.2505,
+      "step": 37063
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.552600204859537e-06,
+      "loss": 0.2977,
+      "step": 37064
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.552264369553171e-06,
+      "loss": 0.4862,
+      "step": 37065
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.551928534246806e-06,
+      "loss": 0.2862,
+      "step": 37066
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.55159269894044e-06,
+      "loss": 0.133,
+      "step": 37067
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.551256863634074e-06,
+      "loss": 0.1792,
+      "step": 37068
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.550921028327709e-06,
+      "loss": 0.2726,
+      "step": 37069
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.550585193021343e-06,
+      "loss": 0.6918,
+      "step": 37070
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5502493577149775e-06,
+      "loss": 0.3545,
+      "step": 37071
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.549913522408612e-06,
+      "loss": 0.3652,
+      "step": 37072
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.549577687102245e-06,
+      "loss": 0.2275,
+      "step": 37073
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.54924185179588e-06,
+      "loss": 0.4342,
+      "step": 37074
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.548906016489514e-06,
+      "loss": 0.2226,
+      "step": 37075
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.548570181183148e-06,
+      "loss": 0.2617,
+      "step": 37076
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.548234345876783e-06,
+      "loss": 0.1386,
+      "step": 37077
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5478985105704175e-06,
+      "loss": 0.2846,
+      "step": 37078
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5475626752640504e-06,
+      "loss": 0.279,
+      "step": 37079
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.547226839957685e-06,
+      "loss": 0.3881,
+      "step": 37080
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.54689100465132e-06,
+      "loss": 0.2503,
+      "step": 37081
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5465551693449535e-06,
+      "loss": 0.2099,
+      "step": 37082
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.546219334038588e-06,
+      "loss": 0.4352,
+      "step": 37083
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.545883498732223e-06,
+      "loss": 0.2109,
+      "step": 37084
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.545547663425857e-06,
+      "loss": 0.1632,
+      "step": 37085
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.54521182811949e-06,
+      "loss": 0.3109,
+      "step": 37086
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.544875992813125e-06,
+      "loss": 0.3273,
+      "step": 37087
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.544540157506759e-06,
+      "loss": 0.28,
+      "step": 37088
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5442043222003935e-06,
+      "loss": 0.3218,
+      "step": 37089
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.543868486894028e-06,
+      "loss": 0.4752,
+      "step": 37090
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.543532651587662e-06,
+      "loss": 0.2027,
+      "step": 37091
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.543196816281297e-06,
+      "loss": 0.2359,
+      "step": 37092
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.542860980974931e-06,
+      "loss": 0.3673,
+      "step": 37093
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.542525145668564e-06,
+      "loss": 0.3234,
+      "step": 37094
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.542189310362199e-06,
+      "loss": 0.1359,
+      "step": 37095
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5418534750558335e-06,
+      "loss": 0.1806,
+      "step": 37096
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.541517639749467e-06,
+      "loss": 0.3312,
+      "step": 37097
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.541181804443102e-06,
+      "loss": 0.3008,
+      "step": 37098
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.540845969136737e-06,
+      "loss": 0.2746,
+      "step": 37099
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5405101338303696e-06,
+      "loss": 0.2985,
+      "step": 37100
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.540174298524004e-06,
+      "loss": 0.2147,
+      "step": 37101
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.539838463217639e-06,
+      "loss": 0.2743,
+      "step": 37102
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.539502627911273e-06,
+      "loss": 0.3667,
+      "step": 37103
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.539166792604907e-06,
+      "loss": 0.2204,
+      "step": 37104
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.538830957298542e-06,
+      "loss": 0.3704,
+      "step": 37105
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.538495121992176e-06,
+      "loss": 0.1521,
+      "step": 37106
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.53815928668581e-06,
+      "loss": 0.1385,
+      "step": 37107
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.537823451379444e-06,
+      "loss": 0.158,
+      "step": 37108
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.537487616073078e-06,
+      "loss": 0.2756,
+      "step": 37109
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.537151780766713e-06,
+      "loss": 0.3099,
+      "step": 37110
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.536815945460347e-06,
+      "loss": 0.3299,
+      "step": 37111
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.536480110153981e-06,
+      "loss": 0.3411,
+      "step": 37112
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.536144274847616e-06,
+      "loss": 0.2648,
+      "step": 37113
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.53580843954125e-06,
+      "loss": 0.1684,
+      "step": 37114
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.535472604234883e-06,
+      "loss": 0.2681,
+      "step": 37115
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.535136768928518e-06,
+      "loss": 0.2657,
+      "step": 37116
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.534800933622153e-06,
+      "loss": 0.4229,
+      "step": 37117
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5344650983157864e-06,
+      "loss": 0.2351,
+      "step": 37118
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.534129263009421e-06,
+      "loss": 0.4234,
+      "step": 37119
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.533793427703056e-06,
+      "loss": 0.4291,
+      "step": 37120
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.533457592396689e-06,
+      "loss": 0.3814,
+      "step": 37121
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.533121757090323e-06,
+      "loss": 0.3983,
+      "step": 37122
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.532785921783958e-06,
+      "loss": 0.2134,
+      "step": 37123
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.532450086477592e-06,
+      "loss": 0.2939,
+      "step": 37124
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.532114251171226e-06,
+      "loss": 0.2324,
+      "step": 37125
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.531778415864861e-06,
+      "loss": 0.3367,
+      "step": 37126
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.531442580558495e-06,
+      "loss": 0.3269,
+      "step": 37127
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5311067452521295e-06,
+      "loss": 0.461,
+      "step": 37128
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.530770909945764e-06,
+      "loss": 0.2287,
+      "step": 37129
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.530435074639397e-06,
+      "loss": 0.1808,
+      "step": 37130
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.530099239333032e-06,
+      "loss": 0.2898,
+      "step": 37131
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5297634040266656e-06,
+      "loss": 0.1193,
+      "step": 37132
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5294275687203e-06,
+      "loss": 0.3027,
+      "step": 37133
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.529091733413935e-06,
+      "loss": 0.5092,
+      "step": 37134
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.528755898107568e-06,
+      "loss": 0.1906,
+      "step": 37135
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5284200628012025e-06,
+      "loss": 0.2713,
+      "step": 37136
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.528084227494837e-06,
+      "loss": 0.2847,
+      "step": 37137
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.527748392188471e-06,
+      "loss": 0.1926,
+      "step": 37138
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5274125568821056e-06,
+      "loss": 0.4322,
+      "step": 37139
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.52707672157574e-06,
+      "loss": 0.3377,
+      "step": 37140
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.526740886269374e-06,
+      "loss": 0.3318,
+      "step": 37141
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.526405050963008e-06,
+      "loss": 0.204,
+      "step": 37142
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5260692156566425e-06,
+      "loss": 0.2722,
+      "step": 37143
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.525733380350276e-06,
+      "loss": 0.4105,
+      "step": 37144
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.525397545043911e-06,
+      "loss": 0.2375,
+      "step": 37145
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5250617097375455e-06,
+      "loss": 0.4,
+      "step": 37146
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.524725874431179e-06,
+      "loss": 0.2429,
+      "step": 37147
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.524390039124814e-06,
+      "loss": 0.2481,
+      "step": 37148
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.524054203818449e-06,
+      "loss": 0.2284,
+      "step": 37149
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.523718368512082e-06,
+      "loss": 0.2879,
+      "step": 37150
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.523382533205716e-06,
+      "loss": 0.1286,
+      "step": 37151
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.523046697899351e-06,
+      "loss": 0.2162,
+      "step": 37152
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.522710862592985e-06,
+      "loss": 0.3075,
+      "step": 37153
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.522375027286619e-06,
+      "loss": 0.3857,
+      "step": 37154
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.522039191980254e-06,
+      "loss": 0.1623,
+      "step": 37155
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.521703356673887e-06,
+      "loss": 0.3099,
+      "step": 37156
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.521367521367522e-06,
+      "loss": 0.193,
+      "step": 37157
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.521031686061156e-06,
+      "loss": 0.246,
+      "step": 37158
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.52069585075479e-06,
+      "loss": 0.2775,
+      "step": 37159
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.520360015448425e-06,
+      "loss": 0.1968,
+      "step": 37160
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.520024180142059e-06,
+      "loss": 0.3037,
+      "step": 37161
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.519688344835693e-06,
+      "loss": 0.2821,
+      "step": 37162
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.519352509529328e-06,
+      "loss": 0.4124,
+      "step": 37163
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5190166742229616e-06,
+      "loss": 0.3394,
+      "step": 37164
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.518680838916595e-06,
+      "loss": 0.2757,
+      "step": 37165
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.51834500361023e-06,
+      "loss": 0.4227,
+      "step": 37166
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.518009168303865e-06,
+      "loss": 0.3424,
+      "step": 37167
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5176733329974985e-06,
+      "loss": 0.4009,
+      "step": 37168
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.517337497691133e-06,
+      "loss": 0.2342,
+      "step": 37169
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.517001662384768e-06,
+      "loss": 0.2052,
+      "step": 37170
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.516665827078401e-06,
+      "loss": 0.2394,
+      "step": 37171
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.516329991772035e-06,
+      "loss": 0.3492,
+      "step": 37172
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.51599415646567e-06,
+      "loss": 0.2017,
+      "step": 37173
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.515658321159304e-06,
+      "loss": 0.2959,
+      "step": 37174
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5153224858529385e-06,
+      "loss": 0.1907,
+      "step": 37175
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.514986650546573e-06,
+      "loss": 0.2205,
+      "step": 37176
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.514650815240206e-06,
+      "loss": 0.4867,
+      "step": 37177
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.514314979933841e-06,
+      "loss": 0.322,
+      "step": 37178
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.513979144627475e-06,
+      "loss": 0.3316,
+      "step": 37179
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.513643309321109e-06,
+      "loss": 0.179,
+      "step": 37180
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.513307474014744e-06,
+      "loss": 0.2545,
+      "step": 37181
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5129716387083785e-06,
+      "loss": 0.345,
+      "step": 37182
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.512635803402012e-06,
+      "loss": 0.2016,
+      "step": 37183
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.512299968095647e-06,
+      "loss": 0.2225,
+      "step": 37184
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5119641327892815e-06,
+      "loss": 0.2147,
+      "step": 37185
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5116282974829145e-06,
+      "loss": 0.2317,
+      "step": 37186
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.511292462176549e-06,
+      "loss": 0.2529,
+      "step": 37187
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.510956626870184e-06,
+      "loss": 0.2201,
+      "step": 37188
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.510620791563818e-06,
+      "loss": 0.2523,
+      "step": 37189
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.510284956257452e-06,
+      "loss": 0.3457,
+      "step": 37190
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.509949120951087e-06,
+      "loss": 0.195,
+      "step": 37191
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.50961328564472e-06,
+      "loss": 0.3723,
+      "step": 37192
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5092774503383545e-06,
+      "loss": 0.4068,
+      "step": 37193
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.508941615031989e-06,
+      "loss": 0.1708,
+      "step": 37194
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.508605779725623e-06,
+      "loss": 0.3316,
+      "step": 37195
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.508269944419258e-06,
+      "loss": 0.22,
+      "step": 37196
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.507934109112892e-06,
+      "loss": 0.3697,
+      "step": 37197
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.507598273806526e-06,
+      "loss": 0.3579,
+      "step": 37198
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.50726243850016e-06,
+      "loss": 0.2279,
+      "step": 37199
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.5069266031937945e-06,
+      "loss": 0.3085,
+      "step": 37200
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.506590767887428e-06,
+      "loss": 0.2042,
+      "step": 37201
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.506254932581063e-06,
+      "loss": 0.2995,
+      "step": 37202
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.505919097274698e-06,
+      "loss": 0.3001,
+      "step": 37203
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.505583261968331e-06,
+      "loss": 0.3282,
+      "step": 37204
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.505247426661966e-06,
+      "loss": 0.2384,
+      "step": 37205
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.504911591355601e-06,
+      "loss": 0.1472,
+      "step": 37206
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.504575756049234e-06,
+      "loss": 0.2172,
+      "step": 37207
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.504239920742868e-06,
+      "loss": 0.1174,
+      "step": 37208
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.503904085436503e-06,
+      "loss": 0.2781,
+      "step": 37209
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.503568250130137e-06,
+      "loss": 0.2533,
+      "step": 37210
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.503232414823771e-06,
+      "loss": 0.2595,
+      "step": 37211
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.502896579517406e-06,
+      "loss": 0.3078,
+      "step": 37212
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.502560744211039e-06,
+      "loss": 0.3813,
+      "step": 37213
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.502224908904674e-06,
+      "loss": 0.401,
+      "step": 37214
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.501889073598308e-06,
+      "loss": 0.3008,
+      "step": 37215
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.501553238291942e-06,
+      "loss": 0.2598,
+      "step": 37216
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.501217402985577e-06,
+      "loss": 0.3708,
+      "step": 37217
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.500881567679211e-06,
+      "loss": 0.4753,
+      "step": 37218
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.500545732372845e-06,
+      "loss": 0.2588,
+      "step": 37219
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 7.50020989706648e-06,
+      "loss": 0.2316,
+      "step": 37220
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.499874061760113e-06,
+      "loss": 0.3958,
+      "step": 37221
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.499538226453747e-06,
+      "loss": 0.278,
+      "step": 37222
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.499202391147382e-06,
+      "loss": 0.2754,
+      "step": 37223
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.498866555841016e-06,
+      "loss": 0.2127,
+      "step": 37224
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4985307205346505e-06,
+      "loss": 0.2874,
+      "step": 37225
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.498194885228285e-06,
+      "loss": 0.3746,
+      "step": 37226
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.497859049921918e-06,
+      "loss": 0.2131,
+      "step": 37227
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.497523214615553e-06,
+      "loss": 0.2682,
+      "step": 37228
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.497187379309187e-06,
+      "loss": 0.4539,
+      "step": 37229
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.496851544002821e-06,
+      "loss": 0.336,
+      "step": 37230
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.496515708696456e-06,
+      "loss": 0.3052,
+      "step": 37231
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4961798733900905e-06,
+      "loss": 0.2629,
+      "step": 37232
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4958440380837234e-06,
+      "loss": 0.3268,
+      "step": 37233
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.495508202777358e-06,
+      "loss": 0.1845,
+      "step": 37234
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.495172367470993e-06,
+      "loss": 0.3501,
+      "step": 37235
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4948365321646265e-06,
+      "loss": 0.2703,
+      "step": 37236
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.494500696858261e-06,
+      "loss": 0.277,
+      "step": 37237
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.494164861551896e-06,
+      "loss": 0.4699,
+      "step": 37238
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.49382902624553e-06,
+      "loss": 0.2459,
+      "step": 37239
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.493493190939164e-06,
+      "loss": 0.4209,
+      "step": 37240
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.493157355632799e-06,
+      "loss": 0.1777,
+      "step": 37241
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.492821520326432e-06,
+      "loss": 0.2982,
+      "step": 37242
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4924856850200665e-06,
+      "loss": 0.2457,
+      "step": 37243
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.492149849713701e-06,
+      "loss": 0.3515,
+      "step": 37244
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.491814014407335e-06,
+      "loss": 0.2676,
+      "step": 37245
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.49147817910097e-06,
+      "loss": 0.1875,
+      "step": 37246
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.491142343794604e-06,
+      "loss": 0.2019,
+      "step": 37247
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.490806508488237e-06,
+      "loss": 0.3623,
+      "step": 37248
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.490470673181872e-06,
+      "loss": 0.4597,
+      "step": 37249
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4901348378755065e-06,
+      "loss": 0.2518,
+      "step": 37250
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.48979900256914e-06,
+      "loss": 0.329,
+      "step": 37251
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.489463167262775e-06,
+      "loss": 0.4056,
+      "step": 37252
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.48912733195641e-06,
+      "loss": 0.407,
+      "step": 37253
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.488791496650043e-06,
+      "loss": 0.4756,
+      "step": 37254
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.488455661343677e-06,
+      "loss": 0.3746,
+      "step": 37255
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.488119826037312e-06,
+      "loss": 0.2234,
+      "step": 37256
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.487783990730946e-06,
+      "loss": 0.4277,
+      "step": 37257
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.48744815542458e-06,
+      "loss": 0.1847,
+      "step": 37258
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.487112320118215e-06,
+      "loss": 0.189,
+      "step": 37259
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.486776484811849e-06,
+      "loss": 0.3033,
+      "step": 37260
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.486440649505483e-06,
+      "loss": 0.3932,
+      "step": 37261
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.486104814199118e-06,
+      "loss": 0.2439,
+      "step": 37262
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.485768978892751e-06,
+      "loss": 0.3542,
+      "step": 37263
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.485433143586386e-06,
+      "loss": 0.2276,
+      "step": 37264
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.48509730828002e-06,
+      "loss": 0.3784,
+      "step": 37265
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.484761472973654e-06,
+      "loss": 0.1968,
+      "step": 37266
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.484425637667289e-06,
+      "loss": 0.2686,
+      "step": 37267
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.484089802360923e-06,
+      "loss": 0.1959,
+      "step": 37268
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.483753967054556e-06,
+      "loss": 0.2217,
+      "step": 37269
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.483418131748191e-06,
+      "loss": 0.2359,
+      "step": 37270
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.483082296441826e-06,
+      "loss": 0.2842,
+      "step": 37271
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4827464611354594e-06,
+      "loss": 0.195,
+      "step": 37272
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.482410625829094e-06,
+      "loss": 0.2744,
+      "step": 37273
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.482074790522729e-06,
+      "loss": 0.2714,
+      "step": 37274
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4817389552163625e-06,
+      "loss": 0.2474,
+      "step": 37275
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.481403119909997e-06,
+      "loss": 0.3556,
+      "step": 37276
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.481067284603631e-06,
+      "loss": 0.1827,
+      "step": 37277
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.480731449297265e-06,
+      "loss": 0.4392,
+      "step": 37278
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4803956139908994e-06,
+      "loss": 0.2406,
+      "step": 37279
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.480059778684534e-06,
+      "loss": 0.2422,
+      "step": 37280
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.479723943378168e-06,
+      "loss": 0.241,
+      "step": 37281
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4793881080718025e-06,
+      "loss": 0.2992,
+      "step": 37282
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.479052272765437e-06,
+      "loss": 0.389,
+      "step": 37283
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.47871643745907e-06,
+      "loss": 0.1993,
+      "step": 37284
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.478380602152705e-06,
+      "loss": 0.3241,
+      "step": 37285
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4780447668463394e-06,
+      "loss": 0.2316,
+      "step": 37286
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.477708931539973e-06,
+      "loss": 0.2696,
+      "step": 37287
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.477373096233608e-06,
+      "loss": 0.3964,
+      "step": 37288
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4770372609272425e-06,
+      "loss": 0.3229,
+      "step": 37289
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4767014256208755e-06,
+      "loss": 0.254,
+      "step": 37290
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.47636559031451e-06,
+      "loss": 0.1748,
+      "step": 37291
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.476029755008145e-06,
+      "loss": 0.3363,
+      "step": 37292
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4756939197017786e-06,
+      "loss": 0.2682,
+      "step": 37293
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.475358084395413e-06,
+      "loss": 0.3559,
+      "step": 37294
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.475022249089048e-06,
+      "loss": 0.313,
+      "step": 37295
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.474686413782682e-06,
+      "loss": 0.3053,
+      "step": 37296
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.474350578476316e-06,
+      "loss": 0.3099,
+      "step": 37297
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.474014743169951e-06,
+      "loss": 0.2503,
+      "step": 37298
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.473678907863584e-06,
+      "loss": 0.3653,
+      "step": 37299
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4733430725572186e-06,
+      "loss": 0.193,
+      "step": 37300
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.473007237250853e-06,
+      "loss": 0.3476,
+      "step": 37301
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.472671401944487e-06,
+      "loss": 0.2473,
+      "step": 37302
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.472335566638122e-06,
+      "loss": 0.2825,
+      "step": 37303
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.471999731331756e-06,
+      "loss": 0.3064,
+      "step": 37304
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.471663896025389e-06,
+      "loss": 0.3055,
+      "step": 37305
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.471328060719024e-06,
+      "loss": 0.2181,
+      "step": 37306
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4709922254126586e-06,
+      "loss": 0.2163,
+      "step": 37307
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.470656390106292e-06,
+      "loss": 0.2201,
+      "step": 37308
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.470320554799927e-06,
+      "loss": 0.1826,
+      "step": 37309
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.469984719493562e-06,
+      "loss": 0.3024,
+      "step": 37310
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4696488841871954e-06,
+      "loss": 0.335,
+      "step": 37311
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.469313048880829e-06,
+      "loss": 0.3385,
+      "step": 37312
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.468977213574463e-06,
+      "loss": 0.3494,
+      "step": 37313
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.468641378268098e-06,
+      "loss": 0.1868,
+      "step": 37314
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.468305542961732e-06,
+      "loss": 0.4427,
+      "step": 37315
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.467969707655366e-06,
+      "loss": 0.5269,
+      "step": 37316
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.467633872349001e-06,
+      "loss": 0.1469,
+      "step": 37317
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4672980370426354e-06,
+      "loss": 0.1915,
+      "step": 37318
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.466962201736268e-06,
+      "loss": 0.323,
+      "step": 37319
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.466626366429903e-06,
+      "loss": 0.3719,
+      "step": 37320
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.466290531123538e-06,
+      "loss": 0.4776,
+      "step": 37321
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4659546958171715e-06,
+      "loss": 0.4378,
+      "step": 37322
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.465618860510806e-06,
+      "loss": 0.2306,
+      "step": 37323
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.465283025204441e-06,
+      "loss": 0.2424,
+      "step": 37324
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.464947189898074e-06,
+      "loss": 0.2981,
+      "step": 37325
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.464611354591708e-06,
+      "loss": 0.5119,
+      "step": 37326
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.464275519285343e-06,
+      "loss": 0.3606,
+      "step": 37327
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.463939683978977e-06,
+      "loss": 0.2267,
+      "step": 37328
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4636038486726115e-06,
+      "loss": 0.4762,
+      "step": 37329
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.463268013366246e-06,
+      "loss": 0.2965,
+      "step": 37330
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.46293217805988e-06,
+      "loss": 0.2488,
+      "step": 37331
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4625963427535146e-06,
+      "loss": 0.187,
+      "step": 37332
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.462260507447149e-06,
+      "loss": 0.2447,
+      "step": 37333
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.461924672140782e-06,
+      "loss": 0.3,
+      "step": 37334
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.461588836834417e-06,
+      "loss": 0.1673,
+      "step": 37335
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4612530015280515e-06,
+      "loss": 0.327,
+      "step": 37336
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.460917166221685e-06,
+      "loss": 0.3913,
+      "step": 37337
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.46058133091532e-06,
+      "loss": 0.3187,
+      "step": 37338
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4602454956089546e-06,
+      "loss": 0.2092,
+      "step": 37339
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4599096603025875e-06,
+      "loss": 0.1939,
+      "step": 37340
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.459573824996222e-06,
+      "loss": 0.3954,
+      "step": 37341
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.459237989689857e-06,
+      "loss": 0.3416,
+      "step": 37342
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.458902154383491e-06,
+      "loss": 0.4444,
+      "step": 37343
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.458566319077125e-06,
+      "loss": 0.402,
+      "step": 37344
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.45823048377076e-06,
+      "loss": 0.3284,
+      "step": 37345
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.457894648464393e-06,
+      "loss": 0.3326,
+      "step": 37346
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4575588131580275e-06,
+      "loss": 0.3061,
+      "step": 37347
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.457222977851662e-06,
+      "loss": 0.2525,
+      "step": 37348
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.456887142545296e-06,
+      "loss": 0.4734,
+      "step": 37349
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.456551307238931e-06,
+      "loss": 0.4598,
+      "step": 37350
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.456215471932565e-06,
+      "loss": 0.3527,
+      "step": 37351
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.455879636626199e-06,
+      "loss": 0.3897,
+      "step": 37352
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.455543801319834e-06,
+      "loss": 0.2631,
+      "step": 37353
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.455207966013468e-06,
+      "loss": 0.3836,
+      "step": 37354
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.454872130707101e-06,
+      "loss": 0.2363,
+      "step": 37355
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.454536295400736e-06,
+      "loss": 0.3046,
+      "step": 37356
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.454200460094371e-06,
+      "loss": 0.379,
+      "step": 37357
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.453864624788004e-06,
+      "loss": 0.1969,
+      "step": 37358
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.453528789481639e-06,
+      "loss": 0.3659,
+      "step": 37359
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.453192954175274e-06,
+      "loss": 0.277,
+      "step": 37360
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.452857118868907e-06,
+      "loss": 0.3796,
+      "step": 37361
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.452521283562541e-06,
+      "loss": 0.2978,
+      "step": 37362
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.452185448256176e-06,
+      "loss": 0.2723,
+      "step": 37363
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.45184961294981e-06,
+      "loss": 0.2744,
+      "step": 37364
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.451513777643444e-06,
+      "loss": 0.1529,
+      "step": 37365
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.451177942337079e-06,
+      "loss": 0.3856,
+      "step": 37366
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.450842107030713e-06,
+      "loss": 0.1908,
+      "step": 37367
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.450506271724347e-06,
+      "loss": 0.2749,
+      "step": 37368
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.450170436417981e-06,
+      "loss": 0.559,
+      "step": 37369
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.449834601111615e-06,
+      "loss": 0.2877,
+      "step": 37370
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.44949876580525e-06,
+      "loss": 0.2284,
+      "step": 37371
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.449162930498884e-06,
+      "loss": 0.4027,
+      "step": 37372
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.448827095192518e-06,
+      "loss": 0.2806,
+      "step": 37373
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.448491259886153e-06,
+      "loss": 0.1762,
+      "step": 37374
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4481554245797875e-06,
+      "loss": 0.4236,
+      "step": 37375
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4478195892734204e-06,
+      "loss": 0.3812,
+      "step": 37376
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.447483753967055e-06,
+      "loss": 0.1837,
+      "step": 37377
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.44714791866069e-06,
+      "loss": 0.4943,
+      "step": 37378
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4468120833543235e-06,
+      "loss": 0.3889,
+      "step": 37379
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.446476248047958e-06,
+      "loss": 0.2629,
+      "step": 37380
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.446140412741593e-06,
+      "loss": 0.3194,
+      "step": 37381
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.445804577435226e-06,
+      "loss": 0.2143,
+      "step": 37382
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.44546874212886e-06,
+      "loss": 0.36,
+      "step": 37383
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.445132906822495e-06,
+      "loss": 0.2336,
+      "step": 37384
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.444797071516129e-06,
+      "loss": 0.5064,
+      "step": 37385
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4444612362097635e-06,
+      "loss": 0.2535,
+      "step": 37386
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.444125400903398e-06,
+      "loss": 0.4206,
+      "step": 37387
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.443789565597032e-06,
+      "loss": 0.2893,
+      "step": 37388
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.443453730290667e-06,
+      "loss": 0.1598,
+      "step": 37389
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4431178949843e-06,
+      "loss": 0.3726,
+      "step": 37390
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.442782059677934e-06,
+      "loss": 0.2909,
+      "step": 37391
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.442446224371569e-06,
+      "loss": 0.2156,
+      "step": 37392
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4421103890652035e-06,
+      "loss": 0.298,
+      "step": 37393
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.441774553758837e-06,
+      "loss": 0.3326,
+      "step": 37394
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.441438718452472e-06,
+      "loss": 0.2943,
+      "step": 37395
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.441102883146107e-06,
+      "loss": 0.3511,
+      "step": 37396
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4407670478397395e-06,
+      "loss": 0.3669,
+      "step": 37397
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.440431212533374e-06,
+      "loss": 0.3557,
+      "step": 37398
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.440095377227009e-06,
+      "loss": 0.4088,
+      "step": 37399
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.439759541920643e-06,
+      "loss": 0.2965,
+      "step": 37400
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.439423706614277e-06,
+      "loss": 0.1778,
+      "step": 37401
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.43908787130791e-06,
+      "loss": 0.3671,
+      "step": 37402
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.438752036001545e-06,
+      "loss": 0.3907,
+      "step": 37403
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4384162006951795e-06,
+      "loss": 0.2897,
+      "step": 37404
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.438080365388813e-06,
+      "loss": 0.2077,
+      "step": 37405
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.437744530082448e-06,
+      "loss": 0.2708,
+      "step": 37406
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.437408694776083e-06,
+      "loss": 0.2941,
+      "step": 37407
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4370728594697164e-06,
+      "loss": 0.226,
+      "step": 37408
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.436737024163351e-06,
+      "loss": 0.1518,
+      "step": 37409
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.436401188856986e-06,
+      "loss": 0.3906,
+      "step": 37410
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.436065353550619e-06,
+      "loss": 0.2681,
+      "step": 37411
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.435729518244253e-06,
+      "loss": 0.3818,
+      "step": 37412
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.435393682937888e-06,
+      "loss": 0.2118,
+      "step": 37413
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.435057847631522e-06,
+      "loss": 0.3096,
+      "step": 37414
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.4347220123251564e-06,
+      "loss": 0.3013,
+      "step": 37415
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.434386177018791e-06,
+      "loss": 0.1994,
+      "step": 37416
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.434050341712424e-06,
+      "loss": 0.3399,
+      "step": 37417
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.433714506406059e-06,
+      "loss": 0.2752,
+      "step": 37418
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 7.433378671099693e-06,
+      "loss": 0.3373,
+      "step": 37419
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.433042835793327e-06,
+      "loss": 0.2742,
+      "step": 37420
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.432707000486962e-06,
+      "loss": 0.2598,
+      "step": 37421
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.432371165180596e-06,
+      "loss": 0.2449,
+      "step": 37422
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.43203532987423e-06,
+      "loss": 0.3181,
+      "step": 37423
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.431699494567864e-06,
+      "loss": 0.2762,
+      "step": 37424
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.431363659261499e-06,
+      "loss": 0.3018,
+      "step": 37425
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.4310278239551325e-06,
+      "loss": 0.3686,
+      "step": 37426
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.430691988648767e-06,
+      "loss": 0.3335,
+      "step": 37427
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.430356153342402e-06,
+      "loss": 0.2147,
+      "step": 37428
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.4300203180360356e-06,
+      "loss": 0.2257,
+      "step": 37429
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.42968448272967e-06,
+      "loss": 0.4342,
+      "step": 37430
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.429348647423305e-06,
+      "loss": 0.1197,
+      "step": 37431
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.429012812116938e-06,
+      "loss": 0.1698,
+      "step": 37432
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.4286769768105725e-06,
+      "loss": 0.3457,
+      "step": 37433
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.428341141504207e-06,
+      "loss": 0.2999,
+      "step": 37434
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.428005306197841e-06,
+      "loss": 0.2736,
+      "step": 37435
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.4276694708914755e-06,
+      "loss": 0.4255,
+      "step": 37436
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.42733363558511e-06,
+      "loss": 0.399,
+      "step": 37437
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.426997800278743e-06,
+      "loss": 0.3829,
+      "step": 37438
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.426661964972378e-06,
+      "loss": 0.2837,
+      "step": 37439
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.4263261296660124e-06,
+      "loss": 0.2262,
+      "step": 37440
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.425990294359646e-06,
+      "loss": 0.422,
+      "step": 37441
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.425654459053281e-06,
+      "loss": 0.4374,
+      "step": 37442
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.4253186237469155e-06,
+      "loss": 0.429,
+      "step": 37443
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.424982788440549e-06,
+      "loss": 0.5805,
+      "step": 37444
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.424646953134184e-06,
+      "loss": 0.3069,
+      "step": 37445
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.424311117827819e-06,
+      "loss": 0.3874,
+      "step": 37446
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.423975282521452e-06,
+      "loss": 0.2235,
+      "step": 37447
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.423639447215086e-06,
+      "loss": 0.1762,
+      "step": 37448
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.423303611908721e-06,
+      "loss": 0.2911,
+      "step": 37449
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.422967776602355e-06,
+      "loss": 0.254,
+      "step": 37450
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.422631941295989e-06,
+      "loss": 0.3211,
+      "step": 37451
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.422296105989624e-06,
+      "loss": 0.1693,
+      "step": 37452
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.421960270683257e-06,
+      "loss": 0.2557,
+      "step": 37453
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.421624435376892e-06,
+      "loss": 0.3446,
+      "step": 37454
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.421288600070526e-06,
+      "loss": 0.252,
+      "step": 37455
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.42095276476416e-06,
+      "loss": 0.3851,
+      "step": 37456
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.420616929457795e-06,
+      "loss": 0.1555,
+      "step": 37457
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.420281094151429e-06,
+      "loss": 0.3487,
+      "step": 37458
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.419945258845062e-06,
+      "loss": 0.2759,
+      "step": 37459
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.419609423538697e-06,
+      "loss": 0.195,
+      "step": 37460
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.4192735882323316e-06,
+      "loss": 0.3612,
+      "step": 37461
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.418937752925965e-06,
+      "loss": 0.4632,
+      "step": 37462
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.4186019176196e-06,
+      "loss": 0.332,
+      "step": 37463
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.418266082313235e-06,
+      "loss": 0.2951,
+      "step": 37464
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.4179302470068685e-06,
+      "loss": 0.2013,
+      "step": 37465
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.417594411700503e-06,
+      "loss": 0.2833,
+      "step": 37466
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.417258576394138e-06,
+      "loss": 0.3112,
+      "step": 37467
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.416922741087771e-06,
+      "loss": 0.3395,
+      "step": 37468
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.416586905781405e-06,
+      "loss": 0.4403,
+      "step": 37469
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.41625107047504e-06,
+      "loss": 0.214,
+      "step": 37470
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.415915235168674e-06,
+      "loss": 0.3155,
+      "step": 37471
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.4155793998623085e-06,
+      "loss": 0.2416,
+      "step": 37472
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.415243564555943e-06,
+      "loss": 0.1917,
+      "step": 37473
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.414907729249576e-06,
+      "loss": 0.2886,
+      "step": 37474
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.414571893943211e-06,
+      "loss": 0.1918,
+      "step": 37475
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.414236058636845e-06,
+      "loss": 0.1995,
+      "step": 37476
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.413900223330479e-06,
+      "loss": 0.2458,
+      "step": 37477
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.413564388024114e-06,
+      "loss": 0.235,
+      "step": 37478
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.4132285527177484e-06,
+      "loss": 0.2683,
+      "step": 37479
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.412892717411382e-06,
+      "loss": 0.1564,
+      "step": 37480
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.412556882105016e-06,
+      "loss": 0.2649,
+      "step": 37481
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.412221046798651e-06,
+      "loss": 0.2491,
+      "step": 37482
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.4118852114922845e-06,
+      "loss": 0.3019,
+      "step": 37483
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.411549376185919e-06,
+      "loss": 0.3251,
+      "step": 37484
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.411213540879554e-06,
+      "loss": 0.282,
+      "step": 37485
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.410877705573188e-06,
+      "loss": 0.1873,
+      "step": 37486
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.410541870266822e-06,
+      "loss": 0.293,
+      "step": 37487
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.410206034960457e-06,
+      "loss": 0.2738,
+      "step": 37488
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.40987019965409e-06,
+      "loss": 0.4198,
+      "step": 37489
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.4095343643477245e-06,
+      "loss": 0.342,
+      "step": 37490
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.409198529041359e-06,
+      "loss": 0.1232,
+      "step": 37491
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.408862693734993e-06,
+      "loss": 0.2377,
+      "step": 37492
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.408526858428628e-06,
+      "loss": 0.2691,
+      "step": 37493
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.4081910231222605e-06,
+      "loss": 0.1345,
+      "step": 37494
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.407855187815895e-06,
+      "loss": 0.3173,
+      "step": 37495
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.40751935250953e-06,
+      "loss": 0.3354,
+      "step": 37496
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.407183517203164e-06,
+      "loss": 0.225,
+      "step": 37497
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.406847681896798e-06,
+      "loss": 0.1133,
+      "step": 37498
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.406511846590433e-06,
+      "loss": 0.3679,
+      "step": 37499
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.406176011284067e-06,
+      "loss": 0.4383,
+      "step": 37500
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.405840175977701e-06,
+      "loss": 0.4186,
+      "step": 37501
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.405504340671336e-06,
+      "loss": 0.1969,
+      "step": 37502
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.405168505364969e-06,
+      "loss": 0.3031,
+      "step": 37503
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.404832670058604e-06,
+      "loss": 0.2149,
+      "step": 37504
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.404496834752238e-06,
+      "loss": 0.2375,
+      "step": 37505
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.404160999445872e-06,
+      "loss": 0.5147,
+      "step": 37506
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.403825164139507e-06,
+      "loss": 0.3288,
+      "step": 37507
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.403489328833141e-06,
+      "loss": 0.3029,
+      "step": 37508
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.403153493526774e-06,
+      "loss": 0.1773,
+      "step": 37509
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.402817658220409e-06,
+      "loss": 0.2438,
+      "step": 37510
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.402481822914044e-06,
+      "loss": 0.274,
+      "step": 37511
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.402145987607677e-06,
+      "loss": 0.2773,
+      "step": 37512
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.401810152301312e-06,
+      "loss": 0.438,
+      "step": 37513
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.401474316994947e-06,
+      "loss": 0.107,
+      "step": 37514
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.40113848168858e-06,
+      "loss": 0.1273,
+      "step": 37515
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.400802646382214e-06,
+      "loss": 0.4329,
+      "step": 37516
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.400466811075849e-06,
+      "loss": 0.2733,
+      "step": 37517
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.400130975769483e-06,
+      "loss": 0.3113,
+      "step": 37518
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.399795140463117e-06,
+      "loss": 0.4594,
+      "step": 37519
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.399459305156752e-06,
+      "loss": 0.2757,
+      "step": 37520
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.399123469850386e-06,
+      "loss": 0.3434,
+      "step": 37521
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.3987876345440205e-06,
+      "loss": 0.2667,
+      "step": 37522
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.398451799237655e-06,
+      "loss": 0.4253,
+      "step": 37523
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.398115963931288e-06,
+      "loss": 0.3361,
+      "step": 37524
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.397780128624923e-06,
+      "loss": 0.2467,
+      "step": 37525
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.397444293318557e-06,
+      "loss": 0.234,
+      "step": 37526
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.397108458012191e-06,
+      "loss": 0.335,
+      "step": 37527
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.396772622705826e-06,
+      "loss": 0.2424,
+      "step": 37528
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.3964367873994605e-06,
+      "loss": 0.2465,
+      "step": 37529
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.3961009520930934e-06,
+      "loss": 0.293,
+      "step": 37530
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.395765116786728e-06,
+      "loss": 0.2775,
+      "step": 37531
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.395429281480363e-06,
+      "loss": 0.4826,
+      "step": 37532
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.3950934461739965e-06,
+      "loss": 0.4408,
+      "step": 37533
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.394757610867631e-06,
+      "loss": 0.2735,
+      "step": 37534
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.394421775561266e-06,
+      "loss": 0.427,
+      "step": 37535
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.3940859402549e-06,
+      "loss": 0.2912,
+      "step": 37536
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.3937501049485334e-06,
+      "loss": 0.3447,
+      "step": 37537
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.393414269642168e-06,
+      "loss": 0.1985,
+      "step": 37538
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.393078434335802e-06,
+      "loss": 0.2774,
+      "step": 37539
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.3927425990294365e-06,
+      "loss": 0.2082,
+      "step": 37540
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.392406763723071e-06,
+      "loss": 0.2478,
+      "step": 37541
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.392070928416705e-06,
+      "loss": 0.3794,
+      "step": 37542
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.39173509311034e-06,
+      "loss": 0.2133,
+      "step": 37543
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.391399257803974e-06,
+      "loss": 0.2108,
+      "step": 37544
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.391063422497607e-06,
+      "loss": 0.3463,
+      "step": 37545
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.390727587191242e-06,
+      "loss": 0.2488,
+      "step": 37546
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.3903917518848765e-06,
+      "loss": 0.277,
+      "step": 37547
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.39005591657851e-06,
+      "loss": 0.4396,
+      "step": 37548
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.389720081272145e-06,
+      "loss": 0.1846,
+      "step": 37549
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.38938424596578e-06,
+      "loss": 0.289,
+      "step": 37550
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.3890484106594126e-06,
+      "loss": 0.2855,
+      "step": 37551
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.388712575353047e-06,
+      "loss": 0.2671,
+      "step": 37552
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.388376740046682e-06,
+      "loss": 0.1517,
+      "step": 37553
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.388040904740316e-06,
+      "loss": 0.2815,
+      "step": 37554
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.38770506943395e-06,
+      "loss": 0.326,
+      "step": 37555
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.387369234127585e-06,
+      "loss": 0.2421,
+      "step": 37556
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.387033398821219e-06,
+      "loss": 0.3944,
+      "step": 37557
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.386697563514853e-06,
+      "loss": 0.1727,
+      "step": 37558
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.386361728208487e-06,
+      "loss": 0.2693,
+      "step": 37559
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.386025892902121e-06,
+      "loss": 0.2408,
+      "step": 37560
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.385690057595756e-06,
+      "loss": 0.2775,
+      "step": 37561
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.38535422228939e-06,
+      "loss": 0.2705,
+      "step": 37562
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.385018386983024e-06,
+      "loss": 0.2792,
+      "step": 37563
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.384682551676659e-06,
+      "loss": 0.3805,
+      "step": 37564
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.384346716370293e-06,
+      "loss": 0.4363,
+      "step": 37565
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.384010881063926e-06,
+      "loss": 0.3545,
+      "step": 37566
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.383675045757561e-06,
+      "loss": 0.2553,
+      "step": 37567
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.383339210451196e-06,
+      "loss": 0.3911,
+      "step": 37568
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.3830033751448294e-06,
+      "loss": 0.2271,
+      "step": 37569
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.382667539838464e-06,
+      "loss": 0.4014,
+      "step": 37570
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.382331704532099e-06,
+      "loss": 0.1958,
+      "step": 37571
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.381995869225732e-06,
+      "loss": 0.4467,
+      "step": 37572
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.381660033919366e-06,
+      "loss": 0.2358,
+      "step": 37573
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.381324198613001e-06,
+      "loss": 0.2068,
+      "step": 37574
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.380988363306635e-06,
+      "loss": 0.4176,
+      "step": 37575
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.3806525280002694e-06,
+      "loss": 0.2944,
+      "step": 37576
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.380316692693904e-06,
+      "loss": 0.355,
+      "step": 37577
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.379980857387538e-06,
+      "loss": 0.3236,
+      "step": 37578
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.3796450220811725e-06,
+      "loss": 0.3764,
+      "step": 37579
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.379309186774807e-06,
+      "loss": 0.3461,
+      "step": 37580
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.37897335146844e-06,
+      "loss": 0.2883,
+      "step": 37581
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.378637516162075e-06,
+      "loss": 0.2478,
+      "step": 37582
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.3783016808557086e-06,
+      "loss": 0.3757,
+      "step": 37583
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.377965845549343e-06,
+      "loss": 0.374,
+      "step": 37584
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.377630010242978e-06,
+      "loss": 0.3235,
+      "step": 37585
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.377294174936611e-06,
+      "loss": 0.2177,
+      "step": 37586
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.3769583396302455e-06,
+      "loss": 0.4184,
+      "step": 37587
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.37662250432388e-06,
+      "loss": 0.2334,
+      "step": 37588
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.376286669017514e-06,
+      "loss": 0.2101,
+      "step": 37589
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.3759508337111486e-06,
+      "loss": 0.2649,
+      "step": 37590
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.375614998404783e-06,
+      "loss": 0.2104,
+      "step": 37591
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.375279163098417e-06,
+      "loss": 0.1402,
+      "step": 37592
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.374943327792052e-06,
+      "loss": 0.3616,
+      "step": 37593
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.3746074924856855e-06,
+      "loss": 0.267,
+      "step": 37594
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.374271657179319e-06,
+      "loss": 0.4748,
+      "step": 37595
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.373935821872954e-06,
+      "loss": 0.4431,
+      "step": 37596
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.3735999865665886e-06,
+      "loss": 0.3716,
+      "step": 37597
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.373264151260222e-06,
+      "loss": 0.2299,
+      "step": 37598
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.372928315953857e-06,
+      "loss": 0.3228,
+      "step": 37599
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.372592480647492e-06,
+      "loss": 0.3661,
+      "step": 37600
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.372256645341125e-06,
+      "loss": 0.3589,
+      "step": 37601
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.371920810034759e-06,
+      "loss": 0.2264,
+      "step": 37602
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.371584974728394e-06,
+      "loss": 0.1924,
+      "step": 37603
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.371249139422028e-06,
+      "loss": 0.2921,
+      "step": 37604
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.370913304115662e-06,
+      "loss": 0.1877,
+      "step": 37605
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.370577468809297e-06,
+      "loss": 0.1308,
+      "step": 37606
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.37024163350293e-06,
+      "loss": 0.217,
+      "step": 37607
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.369905798196565e-06,
+      "loss": 0.1487,
+      "step": 37608
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.369569962890199e-06,
+      "loss": 0.2492,
+      "step": 37609
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.369234127583833e-06,
+      "loss": 0.2614,
+      "step": 37610
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.368898292277468e-06,
+      "loss": 0.3071,
+      "step": 37611
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.368562456971102e-06,
+      "loss": 0.254,
+      "step": 37612
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.368226621664736e-06,
+      "loss": 0.3497,
+      "step": 37613
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.367890786358371e-06,
+      "loss": 0.3136,
+      "step": 37614
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.3675549510520054e-06,
+      "loss": 0.3495,
+      "step": 37615
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.367219115745638e-06,
+      "loss": 0.4006,
+      "step": 37616
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 7.366883280439273e-06,
+      "loss": 0.2912,
+      "step": 37617
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.366547445132908e-06,
+      "loss": 0.3763,
+      "step": 37618
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3662116098265415e-06,
+      "loss": 0.2682,
+      "step": 37619
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.365875774520176e-06,
+      "loss": 0.2617,
+      "step": 37620
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.365539939213811e-06,
+      "loss": 0.2339,
+      "step": 37621
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.365204103907444e-06,
+      "loss": 0.2875,
+      "step": 37622
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.364868268601078e-06,
+      "loss": 0.1856,
+      "step": 37623
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.364532433294713e-06,
+      "loss": 0.1811,
+      "step": 37624
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.364196597988347e-06,
+      "loss": 0.2847,
+      "step": 37625
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3638607626819815e-06,
+      "loss": 0.2426,
+      "step": 37626
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.363524927375616e-06,
+      "loss": 0.3026,
+      "step": 37627
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.363189092069249e-06,
+      "loss": 0.3397,
+      "step": 37628
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.362853256762884e-06,
+      "loss": 0.4591,
+      "step": 37629
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.362517421456518e-06,
+      "loss": 0.4872,
+      "step": 37630
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.362181586150152e-06,
+      "loss": 0.2419,
+      "step": 37631
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.361845750843787e-06,
+      "loss": 0.2541,
+      "step": 37632
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3615099155374215e-06,
+      "loss": 0.1878,
+      "step": 37633
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.361174080231055e-06,
+      "loss": 0.2765,
+      "step": 37634
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.36083824492469e-06,
+      "loss": 0.1889,
+      "step": 37635
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3605024096183246e-06,
+      "loss": 0.2579,
+      "step": 37636
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3601665743119575e-06,
+      "loss": 0.4738,
+      "step": 37637
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.359830739005592e-06,
+      "loss": 0.3489,
+      "step": 37638
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.359494903699227e-06,
+      "loss": 0.2529,
+      "step": 37639
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.359159068392861e-06,
+      "loss": 0.2632,
+      "step": 37640
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.358823233086495e-06,
+      "loss": 0.293,
+      "step": 37641
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.35848739778013e-06,
+      "loss": 0.2375,
+      "step": 37642
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.358151562473763e-06,
+      "loss": 0.1908,
+      "step": 37643
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3578157271673975e-06,
+      "loss": 0.3309,
+      "step": 37644
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.357479891861032e-06,
+      "loss": 0.2422,
+      "step": 37645
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.357144056554666e-06,
+      "loss": 0.1653,
+      "step": 37646
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.356808221248301e-06,
+      "loss": 0.3402,
+      "step": 37647
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.356472385941935e-06,
+      "loss": 0.3492,
+      "step": 37648
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.356136550635569e-06,
+      "loss": 0.2066,
+      "step": 37649
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.355800715329203e-06,
+      "loss": 0.164,
+      "step": 37650
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3554648800228375e-06,
+      "loss": 0.1937,
+      "step": 37651
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.355129044716471e-06,
+      "loss": 0.3943,
+      "step": 37652
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.354793209410106e-06,
+      "loss": 0.1725,
+      "step": 37653
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.354457374103741e-06,
+      "loss": 0.2805,
+      "step": 37654
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.354121538797374e-06,
+      "loss": 0.204,
+      "step": 37655
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.353785703491009e-06,
+      "loss": 0.1223,
+      "step": 37656
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.353449868184644e-06,
+      "loss": 0.3319,
+      "step": 37657
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.353114032878277e-06,
+      "loss": 0.3112,
+      "step": 37658
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.352778197571911e-06,
+      "loss": 0.2764,
+      "step": 37659
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.352442362265546e-06,
+      "loss": 0.2693,
+      "step": 37660
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.35210652695918e-06,
+      "loss": 0.1643,
+      "step": 37661
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.351770691652814e-06,
+      "loss": 0.4139,
+      "step": 37662
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.351434856346449e-06,
+      "loss": 0.3427,
+      "step": 37663
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.351099021040082e-06,
+      "loss": 0.268,
+      "step": 37664
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.350763185733717e-06,
+      "loss": 0.3272,
+      "step": 37665
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.350427350427351e-06,
+      "loss": 0.3538,
+      "step": 37666
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.350091515120985e-06,
+      "loss": 0.1595,
+      "step": 37667
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.34975567981462e-06,
+      "loss": 0.3207,
+      "step": 37668
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.349419844508254e-06,
+      "loss": 0.3486,
+      "step": 37669
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.349084009201888e-06,
+      "loss": 0.3301,
+      "step": 37670
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.348748173895523e-06,
+      "loss": 0.2167,
+      "step": 37671
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.348412338589156e-06,
+      "loss": 0.3208,
+      "step": 37672
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.34807650328279e-06,
+      "loss": 0.2639,
+      "step": 37673
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.347740667976425e-06,
+      "loss": 0.3694,
+      "step": 37674
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.347404832670059e-06,
+      "loss": 0.3802,
+      "step": 37675
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3470689973636935e-06,
+      "loss": 0.2444,
+      "step": 37676
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.346733162057328e-06,
+      "loss": 0.4067,
+      "step": 37677
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.346397326750961e-06,
+      "loss": 0.2595,
+      "step": 37678
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.346061491444596e-06,
+      "loss": 0.2693,
+      "step": 37679
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.34572565613823e-06,
+      "loss": 0.3146,
+      "step": 37680
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.345389820831864e-06,
+      "loss": 0.141,
+      "step": 37681
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.345053985525499e-06,
+      "loss": 0.4323,
+      "step": 37682
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3447181502191335e-06,
+      "loss": 0.4214,
+      "step": 37683
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3443823149127665e-06,
+      "loss": 0.2288,
+      "step": 37684
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.344046479606401e-06,
+      "loss": 0.2476,
+      "step": 37685
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.343710644300036e-06,
+      "loss": 0.2407,
+      "step": 37686
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3433748089936696e-06,
+      "loss": 0.3045,
+      "step": 37687
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.343038973687304e-06,
+      "loss": 0.3571,
+      "step": 37688
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.342703138380939e-06,
+      "loss": 0.2633,
+      "step": 37689
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.342367303074573e-06,
+      "loss": 0.2946,
+      "step": 37690
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.342031467768207e-06,
+      "loss": 0.4305,
+      "step": 37691
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.341695632461842e-06,
+      "loss": 0.2232,
+      "step": 37692
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.341359797155475e-06,
+      "loss": 0.2065,
+      "step": 37693
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3410239618491095e-06,
+      "loss": 0.2204,
+      "step": 37694
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.340688126542744e-06,
+      "loss": 0.316,
+      "step": 37695
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.340352291236378e-06,
+      "loss": 0.1687,
+      "step": 37696
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.340016455930013e-06,
+      "loss": 0.3337,
+      "step": 37697
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.339680620623647e-06,
+      "loss": 0.4546,
+      "step": 37698
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.33934478531728e-06,
+      "loss": 0.2209,
+      "step": 37699
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.339008950010915e-06,
+      "loss": 0.2466,
+      "step": 37700
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3386731147045495e-06,
+      "loss": 0.2839,
+      "step": 37701
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.338337279398183e-06,
+      "loss": 0.1726,
+      "step": 37702
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.338001444091818e-06,
+      "loss": 0.2458,
+      "step": 37703
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.337665608785453e-06,
+      "loss": 0.2404,
+      "step": 37704
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3373297734790864e-06,
+      "loss": 0.2751,
+      "step": 37705
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.336993938172721e-06,
+      "loss": 0.2399,
+      "step": 37706
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.336658102866355e-06,
+      "loss": 0.2147,
+      "step": 37707
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.336322267559989e-06,
+      "loss": 0.2964,
+      "step": 37708
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.335986432253623e-06,
+      "loss": 0.5459,
+      "step": 37709
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.335650596947258e-06,
+      "loss": 0.2259,
+      "step": 37710
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.335314761640892e-06,
+      "loss": 0.2957,
+      "step": 37711
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.334978926334526e-06,
+      "loss": 0.3563,
+      "step": 37712
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.334643091028161e-06,
+      "loss": 0.3834,
+      "step": 37713
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.334307255721794e-06,
+      "loss": 0.6538,
+      "step": 37714
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.333971420415429e-06,
+      "loss": 0.2832,
+      "step": 37715
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.333635585109063e-06,
+      "loss": 0.3691,
+      "step": 37716
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.333299749802697e-06,
+      "loss": 0.3157,
+      "step": 37717
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.332963914496332e-06,
+      "loss": 0.2666,
+      "step": 37718
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.332628079189966e-06,
+      "loss": 0.2334,
+      "step": 37719
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.332292243883599e-06,
+      "loss": 0.4064,
+      "step": 37720
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.331956408577234e-06,
+      "loss": 0.3391,
+      "step": 37721
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.331620573270869e-06,
+      "loss": 0.3015,
+      "step": 37722
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3312847379645025e-06,
+      "loss": 0.3038,
+      "step": 37723
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.330948902658137e-06,
+      "loss": 0.3979,
+      "step": 37724
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.330613067351772e-06,
+      "loss": 0.457,
+      "step": 37725
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3302772320454056e-06,
+      "loss": 0.5007,
+      "step": 37726
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.32994139673904e-06,
+      "loss": 0.2366,
+      "step": 37727
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.329605561432675e-06,
+      "loss": 0.2892,
+      "step": 37728
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.329269726126308e-06,
+      "loss": 0.3058,
+      "step": 37729
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3289338908199424e-06,
+      "loss": 0.2787,
+      "step": 37730
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.328598055513577e-06,
+      "loss": 0.3916,
+      "step": 37731
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.328262220207211e-06,
+      "loss": 0.249,
+      "step": 37732
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3279263849008455e-06,
+      "loss": 0.1677,
+      "step": 37733
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.32759054959448e-06,
+      "loss": 0.3737,
+      "step": 37734
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.327254714288113e-06,
+      "loss": 0.2918,
+      "step": 37735
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.326918878981748e-06,
+      "loss": 0.3829,
+      "step": 37736
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3265830436753824e-06,
+      "loss": 0.2776,
+      "step": 37737
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.326247208369016e-06,
+      "loss": 0.2221,
+      "step": 37738
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.325911373062651e-06,
+      "loss": 0.3032,
+      "step": 37739
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3255755377562855e-06,
+      "loss": 0.2878,
+      "step": 37740
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3252397024499185e-06,
+      "loss": 0.3305,
+      "step": 37741
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.324903867143553e-06,
+      "loss": 0.1816,
+      "step": 37742
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.324568031837188e-06,
+      "loss": 0.3917,
+      "step": 37743
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.324232196530822e-06,
+      "loss": 0.2733,
+      "step": 37744
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.323896361224456e-06,
+      "loss": 0.5158,
+      "step": 37745
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.323560525918091e-06,
+      "loss": 0.2218,
+      "step": 37746
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.323224690611725e-06,
+      "loss": 0.2298,
+      "step": 37747
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.322888855305359e-06,
+      "loss": 0.2897,
+      "step": 37748
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.322553019998994e-06,
+      "loss": 0.4323,
+      "step": 37749
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.322217184692627e-06,
+      "loss": 0.2832,
+      "step": 37750
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3218813493862616e-06,
+      "loss": 0.3699,
+      "step": 37751
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.321545514079896e-06,
+      "loss": 0.2232,
+      "step": 37752
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.32120967877353e-06,
+      "loss": 0.3127,
+      "step": 37753
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.320873843467165e-06,
+      "loss": 0.1766,
+      "step": 37754
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.320538008160799e-06,
+      "loss": 0.192,
+      "step": 37755
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.320202172854432e-06,
+      "loss": 0.2583,
+      "step": 37756
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.319866337548067e-06,
+      "loss": 0.373,
+      "step": 37757
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3195305022417016e-06,
+      "loss": 0.2064,
+      "step": 37758
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.319194666935335e-06,
+      "loss": 0.2783,
+      "step": 37759
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.31885883162897e-06,
+      "loss": 0.1446,
+      "step": 37760
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.318522996322605e-06,
+      "loss": 0.4157,
+      "step": 37761
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3181871610162385e-06,
+      "loss": 0.2006,
+      "step": 37762
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.317851325709872e-06,
+      "loss": 0.4016,
+      "step": 37763
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.317515490403506e-06,
+      "loss": 0.4337,
+      "step": 37764
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.317179655097141e-06,
+      "loss": 0.217,
+      "step": 37765
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.316843819790775e-06,
+      "loss": 0.296,
+      "step": 37766
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.316507984484409e-06,
+      "loss": 0.4827,
+      "step": 37767
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.316172149178044e-06,
+      "loss": 0.2143,
+      "step": 37768
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3158363138716785e-06,
+      "loss": 0.3411,
+      "step": 37769
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.315500478565311e-06,
+      "loss": 0.3224,
+      "step": 37770
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.315164643258946e-06,
+      "loss": 0.1589,
+      "step": 37771
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.314828807952581e-06,
+      "loss": 0.2113,
+      "step": 37772
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3144929726462145e-06,
+      "loss": 0.3199,
+      "step": 37773
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.314157137339849e-06,
+      "loss": 0.1783,
+      "step": 37774
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.313821302033484e-06,
+      "loss": 0.276,
+      "step": 37775
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.313485466727117e-06,
+      "loss": 0.3626,
+      "step": 37776
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.313149631420751e-06,
+      "loss": 0.2186,
+      "step": 37777
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.312813796114386e-06,
+      "loss": 0.3131,
+      "step": 37778
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.31247796080802e-06,
+      "loss": 0.466,
+      "step": 37779
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3121421255016545e-06,
+      "loss": 0.4225,
+      "step": 37780
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.311806290195289e-06,
+      "loss": 0.5051,
+      "step": 37781
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.311470454888923e-06,
+      "loss": 0.4737,
+      "step": 37782
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.311134619582558e-06,
+      "loss": 0.3274,
+      "step": 37783
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.310798784276192e-06,
+      "loss": 0.3287,
+      "step": 37784
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.310462948969825e-06,
+      "loss": 0.2493,
+      "step": 37785
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.31012711366346e-06,
+      "loss": 0.2942,
+      "step": 37786
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3097912783570945e-06,
+      "loss": 0.3496,
+      "step": 37787
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.309455443050728e-06,
+      "loss": 0.2654,
+      "step": 37788
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.309119607744363e-06,
+      "loss": 0.3055,
+      "step": 37789
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3087837724379976e-06,
+      "loss": 0.2033,
+      "step": 37790
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3084479371316305e-06,
+      "loss": 0.361,
+      "step": 37791
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.308112101825265e-06,
+      "loss": 0.5401,
+      "step": 37792
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3077762665189e-06,
+      "loss": 0.2565,
+      "step": 37793
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.307440431212534e-06,
+      "loss": 0.2769,
+      "step": 37794
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.307104595906168e-06,
+      "loss": 0.4614,
+      "step": 37795
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.306768760599803e-06,
+      "loss": 0.2344,
+      "step": 37796
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.306432925293436e-06,
+      "loss": 0.2705,
+      "step": 37797
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.3060970899870705e-06,
+      "loss": 0.1883,
+      "step": 37798
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.305761254680705e-06,
+      "loss": 0.3177,
+      "step": 37799
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.305425419374339e-06,
+      "loss": 0.3646,
+      "step": 37800
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.305089584067974e-06,
+      "loss": 0.2876,
+      "step": 37801
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.304753748761608e-06,
+      "loss": 0.4382,
+      "step": 37802
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.304417913455242e-06,
+      "loss": 0.38,
+      "step": 37803
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.304082078148877e-06,
+      "loss": 0.335,
+      "step": 37804
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.303746242842511e-06,
+      "loss": 0.3275,
+      "step": 37805
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.303410407536144e-06,
+      "loss": 0.4297,
+      "step": 37806
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.303074572229779e-06,
+      "loss": 0.3186,
+      "step": 37807
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.302738736923414e-06,
+      "loss": 0.3545,
+      "step": 37808
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.302402901617047e-06,
+      "loss": 0.405,
+      "step": 37809
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.302067066310682e-06,
+      "loss": 0.2691,
+      "step": 37810
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.301731231004317e-06,
+      "loss": 0.269,
+      "step": 37811
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.30139539569795e-06,
+      "loss": 0.2122,
+      "step": 37812
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.301059560391584e-06,
+      "loss": 0.264,
+      "step": 37813
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.300723725085219e-06,
+      "loss": 0.2382,
+      "step": 37814
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.300387889778853e-06,
+      "loss": 0.2721,
+      "step": 37815
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 7.300052054472487e-06,
+      "loss": 0.2389,
+      "step": 37816
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.299716219166122e-06,
+      "loss": 0.3871,
+      "step": 37817
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.299380383859756e-06,
+      "loss": 0.1776,
+      "step": 37818
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.29904454855339e-06,
+      "loss": 0.1726,
+      "step": 37819
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.298708713247024e-06,
+      "loss": 0.3139,
+      "step": 37820
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.298372877940658e-06,
+      "loss": 0.4123,
+      "step": 37821
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.298037042634293e-06,
+      "loss": 0.4241,
+      "step": 37822
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.297701207327927e-06,
+      "loss": 0.1897,
+      "step": 37823
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.297365372021561e-06,
+      "loss": 0.3569,
+      "step": 37824
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.297029536715196e-06,
+      "loss": 0.3374,
+      "step": 37825
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2966937014088305e-06,
+      "loss": 0.3233,
+      "step": 37826
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2963578661024634e-06,
+      "loss": 0.1836,
+      "step": 37827
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.296022030796098e-06,
+      "loss": 0.4005,
+      "step": 37828
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.295686195489733e-06,
+      "loss": 0.2526,
+      "step": 37829
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2953503601833665e-06,
+      "loss": 0.3058,
+      "step": 37830
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.295014524877001e-06,
+      "loss": 0.1614,
+      "step": 37831
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.294678689570636e-06,
+      "loss": 0.5555,
+      "step": 37832
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.294342854264269e-06,
+      "loss": 0.2619,
+      "step": 37833
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2940070189579034e-06,
+      "loss": 0.3178,
+      "step": 37834
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.293671183651538e-06,
+      "loss": 0.204,
+      "step": 37835
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.293335348345172e-06,
+      "loss": 0.2298,
+      "step": 37836
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2929995130388065e-06,
+      "loss": 0.2166,
+      "step": 37837
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.292663677732441e-06,
+      "loss": 0.1245,
+      "step": 37838
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.292327842426075e-06,
+      "loss": 0.23,
+      "step": 37839
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.29199200711971e-06,
+      "loss": 0.2961,
+      "step": 37840
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.291656171813344e-06,
+      "loss": 0.376,
+      "step": 37841
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.291320336506977e-06,
+      "loss": 0.2339,
+      "step": 37842
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.290984501200612e-06,
+      "loss": 0.4981,
+      "step": 37843
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2906486658942465e-06,
+      "loss": 0.2429,
+      "step": 37844
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.29031283058788e-06,
+      "loss": 0.1848,
+      "step": 37845
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.289976995281515e-06,
+      "loss": 0.3419,
+      "step": 37846
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.28964115997515e-06,
+      "loss": 0.4877,
+      "step": 37847
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2893053246687826e-06,
+      "loss": 0.3382,
+      "step": 37848
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.288969489362417e-06,
+      "loss": 0.1707,
+      "step": 37849
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.288633654056052e-06,
+      "loss": 0.3029,
+      "step": 37850
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.288297818749686e-06,
+      "loss": 0.4833,
+      "step": 37851
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.28796198344332e-06,
+      "loss": 0.321,
+      "step": 37852
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.287626148136954e-06,
+      "loss": 0.3145,
+      "step": 37853
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.287290312830588e-06,
+      "loss": 0.2301,
+      "step": 37854
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2869544775242226e-06,
+      "loss": 0.2055,
+      "step": 37855
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.286618642217856e-06,
+      "loss": 0.1288,
+      "step": 37856
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.286282806911491e-06,
+      "loss": 0.3952,
+      "step": 37857
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.285946971605126e-06,
+      "loss": 0.2907,
+      "step": 37858
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2856111362987594e-06,
+      "loss": 0.3361,
+      "step": 37859
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.285275300992394e-06,
+      "loss": 0.2806,
+      "step": 37860
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.284939465686029e-06,
+      "loss": 0.3528,
+      "step": 37861
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.284603630379662e-06,
+      "loss": 0.3968,
+      "step": 37862
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.284267795073296e-06,
+      "loss": 0.3108,
+      "step": 37863
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.283931959766931e-06,
+      "loss": 0.2527,
+      "step": 37864
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.283596124460565e-06,
+      "loss": 0.3285,
+      "step": 37865
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2832602891541994e-06,
+      "loss": 0.2652,
+      "step": 37866
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.282924453847834e-06,
+      "loss": 0.2872,
+      "step": 37867
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.282588618541467e-06,
+      "loss": 0.2372,
+      "step": 37868
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.282252783235102e-06,
+      "loss": 0.2329,
+      "step": 37869
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.281916947928736e-06,
+      "loss": 0.2811,
+      "step": 37870
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.28158111262237e-06,
+      "loss": 0.2369,
+      "step": 37871
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.281245277316005e-06,
+      "loss": 0.3504,
+      "step": 37872
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2809094420096394e-06,
+      "loss": 0.5081,
+      "step": 37873
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.280573606703273e-06,
+      "loss": 0.2344,
+      "step": 37874
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.280237771396908e-06,
+      "loss": 0.2606,
+      "step": 37875
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.279901936090542e-06,
+      "loss": 0.2449,
+      "step": 37876
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2795661007841755e-06,
+      "loss": 0.2135,
+      "step": 37877
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.27923026547781e-06,
+      "loss": 0.2842,
+      "step": 37878
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.278894430171445e-06,
+      "loss": 0.1623,
+      "step": 37879
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2785585948650786e-06,
+      "loss": 0.2097,
+      "step": 37880
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.278222759558713e-06,
+      "loss": 0.2566,
+      "step": 37881
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.277886924252348e-06,
+      "loss": 0.2769,
+      "step": 37882
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.277551088945981e-06,
+      "loss": 0.3386,
+      "step": 37883
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2772152536396155e-06,
+      "loss": 0.3494,
+      "step": 37884
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.27687941833325e-06,
+      "loss": 0.3701,
+      "step": 37885
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.276543583026884e-06,
+      "loss": 0.2257,
+      "step": 37886
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2762077477205186e-06,
+      "loss": 0.2287,
+      "step": 37887
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.275871912414153e-06,
+      "loss": 0.2866,
+      "step": 37888
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.275536077107786e-06,
+      "loss": 0.2025,
+      "step": 37889
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.275200241801421e-06,
+      "loss": 0.3719,
+      "step": 37890
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2748644064950555e-06,
+      "loss": 0.1563,
+      "step": 37891
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.274528571188689e-06,
+      "loss": 0.2572,
+      "step": 37892
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.274192735882324e-06,
+      "loss": 0.2547,
+      "step": 37893
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2738569005759586e-06,
+      "loss": 0.3971,
+      "step": 37894
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.273521065269592e-06,
+      "loss": 0.3597,
+      "step": 37895
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.273185229963227e-06,
+      "loss": 0.2851,
+      "step": 37896
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.272849394656862e-06,
+      "loss": 0.1568,
+      "step": 37897
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.272513559350495e-06,
+      "loss": 0.2021,
+      "step": 37898
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.272177724044129e-06,
+      "loss": 0.2776,
+      "step": 37899
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.271841888737764e-06,
+      "loss": 0.3608,
+      "step": 37900
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.271506053431398e-06,
+      "loss": 0.3486,
+      "step": 37901
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.271170218125032e-06,
+      "loss": 0.3477,
+      "step": 37902
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.270834382818667e-06,
+      "loss": 0.237,
+      "step": 37903
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2704985475123e-06,
+      "loss": 0.3254,
+      "step": 37904
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.270162712205935e-06,
+      "loss": 0.2433,
+      "step": 37905
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.269826876899569e-06,
+      "loss": 0.175,
+      "step": 37906
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.269491041593203e-06,
+      "loss": 0.2797,
+      "step": 37907
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.269155206286838e-06,
+      "loss": 0.3703,
+      "step": 37908
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.268819370980472e-06,
+      "loss": 0.2256,
+      "step": 37909
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.268483535674105e-06,
+      "loss": 0.1988,
+      "step": 37910
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.26814770036774e-06,
+      "loss": 0.2916,
+      "step": 37911
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.267811865061375e-06,
+      "loss": 0.4379,
+      "step": 37912
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.267476029755008e-06,
+      "loss": 0.3183,
+      "step": 37913
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.267140194448643e-06,
+      "loss": 0.3066,
+      "step": 37914
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.266804359142278e-06,
+      "loss": 0.2125,
+      "step": 37915
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2664685238359115e-06,
+      "loss": 0.2956,
+      "step": 37916
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.266132688529546e-06,
+      "loss": 0.2383,
+      "step": 37917
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.265796853223181e-06,
+      "loss": 0.3583,
+      "step": 37918
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.265461017916814e-06,
+      "loss": 0.3135,
+      "step": 37919
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.265125182610448e-06,
+      "loss": 0.2444,
+      "step": 37920
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.264789347304083e-06,
+      "loss": 0.4316,
+      "step": 37921
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.264453511997717e-06,
+      "loss": 0.2967,
+      "step": 37922
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2641176766913515e-06,
+      "loss": 0.2803,
+      "step": 37923
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.263781841384986e-06,
+      "loss": 0.3094,
+      "step": 37924
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.263446006078619e-06,
+      "loss": 0.2541,
+      "step": 37925
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.263110170772254e-06,
+      "loss": 0.3244,
+      "step": 37926
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.262774335465888e-06,
+      "loss": 0.4036,
+      "step": 37927
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.262438500159522e-06,
+      "loss": 0.4185,
+      "step": 37928
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.262102664853157e-06,
+      "loss": 0.1832,
+      "step": 37929
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2617668295467915e-06,
+      "loss": 0.2781,
+      "step": 37930
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.261430994240425e-06,
+      "loss": 0.3034,
+      "step": 37931
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.261095158934059e-06,
+      "loss": 0.5882,
+      "step": 37932
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.260759323627694e-06,
+      "loss": 0.15,
+      "step": 37933
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2604234883213275e-06,
+      "loss": 0.3567,
+      "step": 37934
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.260087653014962e-06,
+      "loss": 0.2228,
+      "step": 37935
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.259751817708597e-06,
+      "loss": 0.3037,
+      "step": 37936
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.259415982402231e-06,
+      "loss": 0.471,
+      "step": 37937
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.259080147095865e-06,
+      "loss": 0.2544,
+      "step": 37938
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2587443117895e-06,
+      "loss": 0.432,
+      "step": 37939
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.258408476483133e-06,
+      "loss": 0.2287,
+      "step": 37940
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2580726411767675e-06,
+      "loss": 0.3656,
+      "step": 37941
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.257736805870402e-06,
+      "loss": 0.1098,
+      "step": 37942
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.257400970564036e-06,
+      "loss": 0.3482,
+      "step": 37943
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.257065135257671e-06,
+      "loss": 0.1084,
+      "step": 37944
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2567292999513035e-06,
+      "loss": 0.4019,
+      "step": 37945
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.256393464644938e-06,
+      "loss": 0.2442,
+      "step": 37946
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.256057629338573e-06,
+      "loss": 0.1761,
+      "step": 37947
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.255721794032207e-06,
+      "loss": 0.2557,
+      "step": 37948
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.255385958725841e-06,
+      "loss": 0.1415,
+      "step": 37949
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.255050123419476e-06,
+      "loss": 0.2013,
+      "step": 37950
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.25471428811311e-06,
+      "loss": 0.2153,
+      "step": 37951
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.254378452806744e-06,
+      "loss": 0.2006,
+      "step": 37952
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.254042617500379e-06,
+      "loss": 0.2124,
+      "step": 37953
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.253706782194012e-06,
+      "loss": 0.1933,
+      "step": 37954
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.253370946887647e-06,
+      "loss": 0.2618,
+      "step": 37955
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.253035111581281e-06,
+      "loss": 0.2644,
+      "step": 37956
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.252699276274915e-06,
+      "loss": 0.2056,
+      "step": 37957
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.25236344096855e-06,
+      "loss": 0.2814,
+      "step": 37958
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.252027605662184e-06,
+      "loss": 0.2644,
+      "step": 37959
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.251691770355817e-06,
+      "loss": 0.5119,
+      "step": 37960
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.251355935049452e-06,
+      "loss": 0.2489,
+      "step": 37961
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.251020099743087e-06,
+      "loss": 0.3508,
+      "step": 37962
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.25068426443672e-06,
+      "loss": 0.2768,
+      "step": 37963
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.250348429130355e-06,
+      "loss": 0.2568,
+      "step": 37964
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.25001259382399e-06,
+      "loss": 0.2431,
+      "step": 37965
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.249676758517623e-06,
+      "loss": 0.5236,
+      "step": 37966
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.249340923211257e-06,
+      "loss": 0.3498,
+      "step": 37967
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.249005087904892e-06,
+      "loss": 0.3409,
+      "step": 37968
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.248669252598526e-06,
+      "loss": 0.4445,
+      "step": 37969
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.24833341729216e-06,
+      "loss": 0.2797,
+      "step": 37970
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.247997581985795e-06,
+      "loss": 0.2898,
+      "step": 37971
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.247661746679429e-06,
+      "loss": 0.4318,
+      "step": 37972
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2473259113730635e-06,
+      "loss": 0.2554,
+      "step": 37973
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.246990076066698e-06,
+      "loss": 0.1729,
+      "step": 37974
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.246654240760331e-06,
+      "loss": 0.4624,
+      "step": 37975
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.246318405453966e-06,
+      "loss": 0.3718,
+      "step": 37976
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2459825701476e-06,
+      "loss": 0.3324,
+      "step": 37977
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.245646734841234e-06,
+      "loss": 0.3047,
+      "step": 37978
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.245310899534869e-06,
+      "loss": 0.2755,
+      "step": 37979
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2449750642285035e-06,
+      "loss": 0.1553,
+      "step": 37980
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2446392289221365e-06,
+      "loss": 0.1662,
+      "step": 37981
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.244303393615771e-06,
+      "loss": 0.3081,
+      "step": 37982
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.243967558309406e-06,
+      "loss": 0.4224,
+      "step": 37983
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2436317230030395e-06,
+      "loss": 0.3604,
+      "step": 37984
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.243295887696674e-06,
+      "loss": 0.2868,
+      "step": 37985
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.242960052390309e-06,
+      "loss": 0.3028,
+      "step": 37986
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.242624217083943e-06,
+      "loss": 0.1544,
+      "step": 37987
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.242288381777577e-06,
+      "loss": 0.3471,
+      "step": 37988
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.241952546471211e-06,
+      "loss": 0.1744,
+      "step": 37989
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.241616711164845e-06,
+      "loss": 0.4014,
+      "step": 37990
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2412808758584795e-06,
+      "loss": 0.3423,
+      "step": 37991
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.240945040552114e-06,
+      "loss": 0.2803,
+      "step": 37992
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.240609205245748e-06,
+      "loss": 0.3141,
+      "step": 37993
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.240273369939383e-06,
+      "loss": 0.3609,
+      "step": 37994
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.239937534633017e-06,
+      "loss": 0.3388,
+      "step": 37995
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.23960169932665e-06,
+      "loss": 0.2142,
+      "step": 37996
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.239265864020285e-06,
+      "loss": 0.2173,
+      "step": 37997
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.2389300287139195e-06,
+      "loss": 0.2321,
+      "step": 37998
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.238594193407553e-06,
+      "loss": 0.0949,
+      "step": 37999
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.238258358101188e-06,
+      "loss": 0.3347,
+      "step": 38000
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.237922522794823e-06,
+      "loss": 0.3066,
+      "step": 38001
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.237586687488456e-06,
+      "loss": 0.3147,
+      "step": 38002
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.23725085218209e-06,
+      "loss": 0.4152,
+      "step": 38003
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.236915016875725e-06,
+      "loss": 0.2667,
+      "step": 38004
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.236579181569359e-06,
+      "loss": 0.3025,
+      "step": 38005
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.236243346262993e-06,
+      "loss": 0.3374,
+      "step": 38006
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.235907510956628e-06,
+      "loss": 0.2563,
+      "step": 38007
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.235571675650262e-06,
+      "loss": 0.2632,
+      "step": 38008
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.235235840343896e-06,
+      "loss": 0.2014,
+      "step": 38009
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.234900005037531e-06,
+      "loss": 0.2185,
+      "step": 38010
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.234564169731164e-06,
+      "loss": 0.3997,
+      "step": 38011
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.234228334424799e-06,
+      "loss": 0.4224,
+      "step": 38012
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.233892499118433e-06,
+      "loss": 0.4474,
+      "step": 38013
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 7.233556663812067e-06,
+      "loss": 0.1302,
+      "step": 38014
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.233220828505702e-06,
+      "loss": 0.4228,
+      "step": 38015
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.232884993199336e-06,
+      "loss": 0.1894,
+      "step": 38016
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.232549157892969e-06,
+      "loss": 0.2845,
+      "step": 38017
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.232213322586604e-06,
+      "loss": 0.2742,
+      "step": 38018
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.231877487280239e-06,
+      "loss": 0.3479,
+      "step": 38019
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.2315416519738725e-06,
+      "loss": 0.3198,
+      "step": 38020
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.231205816667507e-06,
+      "loss": 0.2956,
+      "step": 38021
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.230869981361142e-06,
+      "loss": 0.1996,
+      "step": 38022
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.230534146054775e-06,
+      "loss": 0.4316,
+      "step": 38023
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.230198310748409e-06,
+      "loss": 0.3125,
+      "step": 38024
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.229862475442044e-06,
+      "loss": 0.2426,
+      "step": 38025
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.229526640135678e-06,
+      "loss": 0.2867,
+      "step": 38026
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.2291908048293124e-06,
+      "loss": 0.301,
+      "step": 38027
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.228854969522947e-06,
+      "loss": 0.3324,
+      "step": 38028
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.228519134216581e-06,
+      "loss": 0.2748,
+      "step": 38029
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.2281832989102155e-06,
+      "loss": 0.3063,
+      "step": 38030
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.22784746360385e-06,
+      "loss": 0.3524,
+      "step": 38031
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.227511628297483e-06,
+      "loss": 0.2505,
+      "step": 38032
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.227175792991118e-06,
+      "loss": 0.2877,
+      "step": 38033
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.226839957684752e-06,
+      "loss": 0.3362,
+      "step": 38034
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.226504122378386e-06,
+      "loss": 0.3047,
+      "step": 38035
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.226168287072021e-06,
+      "loss": 0.2982,
+      "step": 38036
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.225832451765654e-06,
+      "loss": 0.2377,
+      "step": 38037
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.2254966164592885e-06,
+      "loss": 0.3227,
+      "step": 38038
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.225160781152923e-06,
+      "loss": 0.2446,
+      "step": 38039
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.224824945846557e-06,
+      "loss": 0.2252,
+      "step": 38040
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.224489110540192e-06,
+      "loss": 0.343,
+      "step": 38041
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.224153275233826e-06,
+      "loss": 0.3293,
+      "step": 38042
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.22381743992746e-06,
+      "loss": 0.4283,
+      "step": 38043
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.223481604621095e-06,
+      "loss": 0.2229,
+      "step": 38044
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.2231457693147285e-06,
+      "loss": 0.3299,
+      "step": 38045
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.222809934008362e-06,
+      "loss": 0.3047,
+      "step": 38046
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.222474098701997e-06,
+      "loss": 0.2999,
+      "step": 38047
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.2221382633956316e-06,
+      "loss": 0.3558,
+      "step": 38048
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.221802428089265e-06,
+      "loss": 0.3537,
+      "step": 38049
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.2214665927829e-06,
+      "loss": 0.3422,
+      "step": 38050
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.221130757476535e-06,
+      "loss": 0.2266,
+      "step": 38051
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.220794922170168e-06,
+      "loss": 0.3535,
+      "step": 38052
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.220459086863802e-06,
+      "loss": 0.2458,
+      "step": 38053
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.220123251557437e-06,
+      "loss": 0.2066,
+      "step": 38054
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.219787416251071e-06,
+      "loss": 0.2323,
+      "step": 38055
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.219451580944705e-06,
+      "loss": 0.3212,
+      "step": 38056
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.21911574563834e-06,
+      "loss": 0.1844,
+      "step": 38057
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.218779910331973e-06,
+      "loss": 0.3209,
+      "step": 38058
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.218444075025608e-06,
+      "loss": 0.406,
+      "step": 38059
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.218108239719242e-06,
+      "loss": 0.589,
+      "step": 38060
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.217772404412876e-06,
+      "loss": 0.3515,
+      "step": 38061
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.217436569106511e-06,
+      "loss": 0.4357,
+      "step": 38062
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.217100733800145e-06,
+      "loss": 0.3202,
+      "step": 38063
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.216764898493779e-06,
+      "loss": 0.3342,
+      "step": 38064
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.216429063187414e-06,
+      "loss": 0.3914,
+      "step": 38065
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.2160932278810484e-06,
+      "loss": 0.3327,
+      "step": 38066
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.215757392574681e-06,
+      "loss": 0.3112,
+      "step": 38067
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.215421557268316e-06,
+      "loss": 0.3465,
+      "step": 38068
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.215085721961951e-06,
+      "loss": 0.2272,
+      "step": 38069
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.2147498866555845e-06,
+      "loss": 0.4924,
+      "step": 38070
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.214414051349219e-06,
+      "loss": 0.2635,
+      "step": 38071
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.214078216042854e-06,
+      "loss": 0.3087,
+      "step": 38072
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.213742380736487e-06,
+      "loss": 0.2283,
+      "step": 38073
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.213406545430121e-06,
+      "loss": 0.3037,
+      "step": 38074
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.213070710123756e-06,
+      "loss": 0.3133,
+      "step": 38075
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.21273487481739e-06,
+      "loss": 0.2456,
+      "step": 38076
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.2123990395110245e-06,
+      "loss": 0.3237,
+      "step": 38077
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.212063204204659e-06,
+      "loss": 0.322,
+      "step": 38078
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.211727368898292e-06,
+      "loss": 0.2203,
+      "step": 38079
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.211391533591927e-06,
+      "loss": 0.1947,
+      "step": 38080
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.211055698285561e-06,
+      "loss": 0.327,
+      "step": 38081
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.210719862979195e-06,
+      "loss": 0.34,
+      "step": 38082
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.21038402767283e-06,
+      "loss": 0.3468,
+      "step": 38083
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.2100481923664645e-06,
+      "loss": 0.3717,
+      "step": 38084
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.209712357060098e-06,
+      "loss": 0.2745,
+      "step": 38085
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.209376521753733e-06,
+      "loss": 0.3687,
+      "step": 38086
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.2090406864473676e-06,
+      "loss": 0.2635,
+      "step": 38087
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.2087048511410005e-06,
+      "loss": 0.2643,
+      "step": 38088
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.208369015834635e-06,
+      "loss": 0.1846,
+      "step": 38089
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.20803318052827e-06,
+      "loss": 0.2377,
+      "step": 38090
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.207697345221904e-06,
+      "loss": 0.2014,
+      "step": 38091
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.207361509915538e-06,
+      "loss": 0.3751,
+      "step": 38092
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.207025674609173e-06,
+      "loss": 0.4055,
+      "step": 38093
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.206689839302806e-06,
+      "loss": 0.3388,
+      "step": 38094
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.2063540039964405e-06,
+      "loss": 0.2711,
+      "step": 38095
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.206018168690075e-06,
+      "loss": 0.2637,
+      "step": 38096
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.205682333383709e-06,
+      "loss": 0.3757,
+      "step": 38097
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.205346498077344e-06,
+      "loss": 0.2207,
+      "step": 38098
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.205010662770978e-06,
+      "loss": 0.3473,
+      "step": 38099
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.204674827464612e-06,
+      "loss": 0.266,
+      "step": 38100
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.204338992158246e-06,
+      "loss": 0.3607,
+      "step": 38101
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.2040031568518805e-06,
+      "loss": 0.1821,
+      "step": 38102
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.203667321545514e-06,
+      "loss": 0.2911,
+      "step": 38103
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.203331486239149e-06,
+      "loss": 0.3247,
+      "step": 38104
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.202995650932784e-06,
+      "loss": 0.3522,
+      "step": 38105
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.202659815626417e-06,
+      "loss": 0.2553,
+      "step": 38106
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.202323980320052e-06,
+      "loss": 0.3016,
+      "step": 38107
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.201988145013687e-06,
+      "loss": 0.2984,
+      "step": 38108
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.20165230970732e-06,
+      "loss": 0.4309,
+      "step": 38109
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.201316474400954e-06,
+      "loss": 0.2699,
+      "step": 38110
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.200980639094589e-06,
+      "loss": 0.2774,
+      "step": 38111
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.200644803788223e-06,
+      "loss": 0.2291,
+      "step": 38112
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.200308968481857e-06,
+      "loss": 0.2331,
+      "step": 38113
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.199973133175492e-06,
+      "loss": 0.2137,
+      "step": 38114
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.199637297869125e-06,
+      "loss": 0.3959,
+      "step": 38115
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.19930146256276e-06,
+      "loss": 0.3013,
+      "step": 38116
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.198965627256394e-06,
+      "loss": 0.3289,
+      "step": 38117
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.198629791950028e-06,
+      "loss": 0.1152,
+      "step": 38118
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.198293956643663e-06,
+      "loss": 0.3884,
+      "step": 38119
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.197958121337297e-06,
+      "loss": 0.4573,
+      "step": 38120
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.197622286030931e-06,
+      "loss": 0.2059,
+      "step": 38121
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.197286450724566e-06,
+      "loss": 0.174,
+      "step": 38122
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.196950615418199e-06,
+      "loss": 0.271,
+      "step": 38123
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.1966147801118334e-06,
+      "loss": 0.2551,
+      "step": 38124
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.196278944805468e-06,
+      "loss": 0.231,
+      "step": 38125
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.195943109499102e-06,
+      "loss": 0.1461,
+      "step": 38126
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.1956072741927365e-06,
+      "loss": 0.4908,
+      "step": 38127
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.195271438886371e-06,
+      "loss": 0.2326,
+      "step": 38128
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.194935603580004e-06,
+      "loss": 0.3243,
+      "step": 38129
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.194599768273639e-06,
+      "loss": 0.1889,
+      "step": 38130
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.194263932967273e-06,
+      "loss": 0.3104,
+      "step": 38131
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.193928097660907e-06,
+      "loss": 0.3197,
+      "step": 38132
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.193592262354542e-06,
+      "loss": 0.172,
+      "step": 38133
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.1932564270481765e-06,
+      "loss": 0.3458,
+      "step": 38134
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.19292059174181e-06,
+      "loss": 0.2997,
+      "step": 38135
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.192584756435444e-06,
+      "loss": 0.3915,
+      "step": 38136
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.192248921129079e-06,
+      "loss": 0.5121,
+      "step": 38137
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.1919130858227126e-06,
+      "loss": 0.4557,
+      "step": 38138
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.191577250516347e-06,
+      "loss": 0.4475,
+      "step": 38139
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.191241415209982e-06,
+      "loss": 0.3183,
+      "step": 38140
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.190905579903616e-06,
+      "loss": 0.4225,
+      "step": 38141
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.19056974459725e-06,
+      "loss": 0.2086,
+      "step": 38142
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.190233909290885e-06,
+      "loss": 0.3572,
+      "step": 38143
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.189898073984518e-06,
+      "loss": 0.1873,
+      "step": 38144
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.1895622386781526e-06,
+      "loss": 0.2643,
+      "step": 38145
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.189226403371787e-06,
+      "loss": 0.2137,
+      "step": 38146
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.188890568065421e-06,
+      "loss": 0.2277,
+      "step": 38147
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.188554732759056e-06,
+      "loss": 0.24,
+      "step": 38148
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.18821889745269e-06,
+      "loss": 0.2153,
+      "step": 38149
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.187883062146323e-06,
+      "loss": 0.1756,
+      "step": 38150
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.187547226839958e-06,
+      "loss": 0.244,
+      "step": 38151
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.1872113915335925e-06,
+      "loss": 0.2763,
+      "step": 38152
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.186875556227226e-06,
+      "loss": 0.3336,
+      "step": 38153
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.186539720920861e-06,
+      "loss": 0.1311,
+      "step": 38154
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.186203885614496e-06,
+      "loss": 0.2052,
+      "step": 38155
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.1858680503081294e-06,
+      "loss": 0.1202,
+      "step": 38156
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.185532215001764e-06,
+      "loss": 0.2886,
+      "step": 38157
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.185196379695398e-06,
+      "loss": 0.2158,
+      "step": 38158
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.184860544389032e-06,
+      "loss": 0.3516,
+      "step": 38159
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.184524709082666e-06,
+      "loss": 0.1767,
+      "step": 38160
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.184188873776301e-06,
+      "loss": 0.3012,
+      "step": 38161
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.183853038469935e-06,
+      "loss": 0.4425,
+      "step": 38162
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.1835172031635694e-06,
+      "loss": 0.1917,
+      "step": 38163
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.183181367857204e-06,
+      "loss": 0.2092,
+      "step": 38164
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.182845532550837e-06,
+      "loss": 0.4002,
+      "step": 38165
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.182509697244472e-06,
+      "loss": 0.4196,
+      "step": 38166
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.182173861938106e-06,
+      "loss": 0.2773,
+      "step": 38167
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.18183802663174e-06,
+      "loss": 0.3771,
+      "step": 38168
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.181502191325375e-06,
+      "loss": 0.2575,
+      "step": 38169
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.181166356019009e-06,
+      "loss": 0.486,
+      "step": 38170
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.180830520712642e-06,
+      "loss": 0.4728,
+      "step": 38171
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.180494685406277e-06,
+      "loss": 0.2549,
+      "step": 38172
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.180158850099912e-06,
+      "loss": 0.1666,
+      "step": 38173
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.1798230147935455e-06,
+      "loss": 0.1209,
+      "step": 38174
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.17948717948718e-06,
+      "loss": 0.1729,
+      "step": 38175
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.179151344180815e-06,
+      "loss": 0.2197,
+      "step": 38176
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.1788155088744486e-06,
+      "loss": 0.2807,
+      "step": 38177
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.178479673568083e-06,
+      "loss": 0.2339,
+      "step": 38178
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.178143838261718e-06,
+      "loss": 0.2746,
+      "step": 38179
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.177808002955351e-06,
+      "loss": 0.1982,
+      "step": 38180
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.1774721676489855e-06,
+      "loss": 0.267,
+      "step": 38181
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.17713633234262e-06,
+      "loss": 0.2597,
+      "step": 38182
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.176800497036254e-06,
+      "loss": 0.3412,
+      "step": 38183
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.1764646617298886e-06,
+      "loss": 0.415,
+      "step": 38184
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.176128826423523e-06,
+      "loss": 0.321,
+      "step": 38185
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.175792991117156e-06,
+      "loss": 0.3661,
+      "step": 38186
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.175457155810791e-06,
+      "loss": 0.194,
+      "step": 38187
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.1751213205044255e-06,
+      "loss": 0.3216,
+      "step": 38188
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.174785485198059e-06,
+      "loss": 0.3854,
+      "step": 38189
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.174449649891694e-06,
+      "loss": 0.3116,
+      "step": 38190
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.1741138145853285e-06,
+      "loss": 0.4191,
+      "step": 38191
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.1737779792789615e-06,
+      "loss": 0.2122,
+      "step": 38192
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.173442143972596e-06,
+      "loss": 0.2838,
+      "step": 38193
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.173106308666231e-06,
+      "loss": 0.2813,
+      "step": 38194
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.172770473359865e-06,
+      "loss": 0.169,
+      "step": 38195
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.172434638053499e-06,
+      "loss": 0.2629,
+      "step": 38196
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.172098802747134e-06,
+      "loss": 0.223,
+      "step": 38197
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.171762967440768e-06,
+      "loss": 0.345,
+      "step": 38198
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.171427132134402e-06,
+      "loss": 0.2374,
+      "step": 38199
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.171091296828037e-06,
+      "loss": 0.4062,
+      "step": 38200
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.17075546152167e-06,
+      "loss": 0.2667,
+      "step": 38201
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.170419626215305e-06,
+      "loss": 0.5883,
+      "step": 38202
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.170083790908939e-06,
+      "loss": 0.2259,
+      "step": 38203
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.169747955602573e-06,
+      "loss": 0.2394,
+      "step": 38204
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.169412120296208e-06,
+      "loss": 0.288,
+      "step": 38205
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.169076284989842e-06,
+      "loss": 0.4048,
+      "step": 38206
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.168740449683475e-06,
+      "loss": 0.2173,
+      "step": 38207
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.16840461437711e-06,
+      "loss": 0.146,
+      "step": 38208
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.168068779070745e-06,
+      "loss": 0.3253,
+      "step": 38209
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.167732943764378e-06,
+      "loss": 0.2076,
+      "step": 38210
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.167397108458013e-06,
+      "loss": 0.3159,
+      "step": 38211
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.167061273151648e-06,
+      "loss": 0.439,
+      "step": 38212
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.1667254378452815e-06,
+      "loss": 0.3331,
+      "step": 38213
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.166389602538915e-06,
+      "loss": 0.3411,
+      "step": 38214
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.166053767232549e-06,
+      "loss": 0.3283,
+      "step": 38215
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.165717931926184e-06,
+      "loss": 0.2794,
+      "step": 38216
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.165382096619818e-06,
+      "loss": 0.3544,
+      "step": 38217
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.165046261313452e-06,
+      "loss": 0.4584,
+      "step": 38218
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.164710426007087e-06,
+      "loss": 0.2469,
+      "step": 38219
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1643745907007215e-06,
+      "loss": 0.223,
+      "step": 38220
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.164038755394354e-06,
+      "loss": 0.3011,
+      "step": 38221
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.163702920087989e-06,
+      "loss": 0.3288,
+      "step": 38222
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.163367084781624e-06,
+      "loss": 0.2834,
+      "step": 38223
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1630312494752575e-06,
+      "loss": 0.408,
+      "step": 38224
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.162695414168892e-06,
+      "loss": 0.3208,
+      "step": 38225
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.162359578862527e-06,
+      "loss": 0.3252,
+      "step": 38226
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.16202374355616e-06,
+      "loss": 0.2072,
+      "step": 38227
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.161687908249794e-06,
+      "loss": 0.4001,
+      "step": 38228
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.161352072943429e-06,
+      "loss": 0.1894,
+      "step": 38229
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.161016237637063e-06,
+      "loss": 0.2156,
+      "step": 38230
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1606804023306975e-06,
+      "loss": 0.2543,
+      "step": 38231
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.160344567024332e-06,
+      "loss": 0.3147,
+      "step": 38232
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.160008731717966e-06,
+      "loss": 0.3343,
+      "step": 38233
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.159672896411601e-06,
+      "loss": 0.4014,
+      "step": 38234
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.159337061105235e-06,
+      "loss": 0.3114,
+      "step": 38235
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.159001225798868e-06,
+      "loss": 0.4873,
+      "step": 38236
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.158665390492503e-06,
+      "loss": 0.296,
+      "step": 38237
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1583295551861375e-06,
+      "loss": 0.2369,
+      "step": 38238
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.157993719879771e-06,
+      "loss": 0.1383,
+      "step": 38239
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.157657884573406e-06,
+      "loss": 0.4451,
+      "step": 38240
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.157322049267041e-06,
+      "loss": 0.1677,
+      "step": 38241
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1569862139606735e-06,
+      "loss": 0.2847,
+      "step": 38242
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.156650378654308e-06,
+      "loss": 0.444,
+      "step": 38243
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.156314543347943e-06,
+      "loss": 0.3387,
+      "step": 38244
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.155978708041577e-06,
+      "loss": 0.223,
+      "step": 38245
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.155642872735211e-06,
+      "loss": 0.3756,
+      "step": 38246
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.155307037428846e-06,
+      "loss": 0.284,
+      "step": 38247
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.15497120212248e-06,
+      "loss": 0.2374,
+      "step": 38248
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1546353668161135e-06,
+      "loss": 0.4183,
+      "step": 38249
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.154299531509748e-06,
+      "loss": 0.2525,
+      "step": 38250
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.153963696203382e-06,
+      "loss": 0.2805,
+      "step": 38251
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.153627860897017e-06,
+      "loss": 0.236,
+      "step": 38252
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.153292025590651e-06,
+      "loss": 0.2551,
+      "step": 38253
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.152956190284285e-06,
+      "loss": 0.2302,
+      "step": 38254
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.15262035497792e-06,
+      "loss": 0.1679,
+      "step": 38255
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.152284519671554e-06,
+      "loss": 0.2603,
+      "step": 38256
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.151948684365187e-06,
+      "loss": 0.234,
+      "step": 38257
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.151612849058822e-06,
+      "loss": 0.3785,
+      "step": 38258
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.151277013752457e-06,
+      "loss": 0.2532,
+      "step": 38259
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.15094117844609e-06,
+      "loss": 0.3288,
+      "step": 38260
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.150605343139725e-06,
+      "loss": 0.1943,
+      "step": 38261
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.15026950783336e-06,
+      "loss": 0.2546,
+      "step": 38262
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.149933672526993e-06,
+      "loss": 0.1596,
+      "step": 38263
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.149597837220627e-06,
+      "loss": 0.5537,
+      "step": 38264
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.149262001914262e-06,
+      "loss": 0.3265,
+      "step": 38265
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.148926166607896e-06,
+      "loss": 0.551,
+      "step": 38266
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.14859033130153e-06,
+      "loss": 0.2712,
+      "step": 38267
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.148254495995165e-06,
+      "loss": 0.2437,
+      "step": 38268
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.147918660688799e-06,
+      "loss": 0.4216,
+      "step": 38269
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1475828253824335e-06,
+      "loss": 0.415,
+      "step": 38270
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.147246990076067e-06,
+      "loss": 0.2808,
+      "step": 38271
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.146911154769701e-06,
+      "loss": 0.2475,
+      "step": 38272
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.146575319463336e-06,
+      "loss": 0.288,
+      "step": 38273
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.14623948415697e-06,
+      "loss": 0.3349,
+      "step": 38274
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.145903648850604e-06,
+      "loss": 0.2197,
+      "step": 38275
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.145567813544239e-06,
+      "loss": 0.2303,
+      "step": 38276
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1452319782378735e-06,
+      "loss": 0.3032,
+      "step": 38277
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1448961429315064e-06,
+      "loss": 0.3423,
+      "step": 38278
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.144560307625141e-06,
+      "loss": 0.3069,
+      "step": 38279
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.144224472318776e-06,
+      "loss": 0.177,
+      "step": 38280
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1438886370124095e-06,
+      "loss": 0.2949,
+      "step": 38281
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.143552801706044e-06,
+      "loss": 0.35,
+      "step": 38282
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.143216966399679e-06,
+      "loss": 0.2271,
+      "step": 38283
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.142881131093312e-06,
+      "loss": 0.3975,
+      "step": 38284
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1425452957869464e-06,
+      "loss": 0.194,
+      "step": 38285
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.142209460480581e-06,
+      "loss": 0.2104,
+      "step": 38286
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.141873625174215e-06,
+      "loss": 0.3921,
+      "step": 38287
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1415377898678495e-06,
+      "loss": 0.405,
+      "step": 38288
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.141201954561484e-06,
+      "loss": 0.3759,
+      "step": 38289
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.140866119255118e-06,
+      "loss": 0.2753,
+      "step": 38290
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.140530283948753e-06,
+      "loss": 0.1487,
+      "step": 38291
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.140194448642387e-06,
+      "loss": 0.3291,
+      "step": 38292
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.13985861333602e-06,
+      "loss": 0.314,
+      "step": 38293
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.139522778029655e-06,
+      "loss": 0.2646,
+      "step": 38294
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1391869427232895e-06,
+      "loss": 0.214,
+      "step": 38295
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.138851107416923e-06,
+      "loss": 0.2324,
+      "step": 38296
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.138515272110558e-06,
+      "loss": 0.2508,
+      "step": 38297
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.138179436804193e-06,
+      "loss": 0.2344,
+      "step": 38298
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1378436014978256e-06,
+      "loss": 0.1912,
+      "step": 38299
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.13750776619146e-06,
+      "loss": 0.1979,
+      "step": 38300
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.137171930885095e-06,
+      "loss": 0.4346,
+      "step": 38301
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.136836095578729e-06,
+      "loss": 0.2785,
+      "step": 38302
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.136500260272363e-06,
+      "loss": 0.278,
+      "step": 38303
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.136164424965997e-06,
+      "loss": 0.212,
+      "step": 38304
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.135828589659631e-06,
+      "loss": 0.2026,
+      "step": 38305
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1354927543532656e-06,
+      "loss": 0.2097,
+      "step": 38306
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.135156919046899e-06,
+      "loss": 0.18,
+      "step": 38307
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.134821083740534e-06,
+      "loss": 0.2176,
+      "step": 38308
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.134485248434169e-06,
+      "loss": 0.4053,
+      "step": 38309
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1341494131278025e-06,
+      "loss": 0.3686,
+      "step": 38310
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.133813577821437e-06,
+      "loss": 0.3578,
+      "step": 38311
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.133477742515072e-06,
+      "loss": 0.3738,
+      "step": 38312
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.133141907208705e-06,
+      "loss": 0.2321,
+      "step": 38313
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.132806071902339e-06,
+      "loss": 0.3332,
+      "step": 38314
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.132470236595974e-06,
+      "loss": 0.1421,
+      "step": 38315
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.132134401289608e-06,
+      "loss": 0.3483,
+      "step": 38316
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1317985659832424e-06,
+      "loss": 0.2116,
+      "step": 38317
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.131462730676877e-06,
+      "loss": 0.257,
+      "step": 38318
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.13112689537051e-06,
+      "loss": 0.4752,
+      "step": 38319
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.130791060064145e-06,
+      "loss": 0.3997,
+      "step": 38320
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.130455224757779e-06,
+      "loss": 0.3685,
+      "step": 38321
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.130119389451413e-06,
+      "loss": 0.2114,
+      "step": 38322
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.129783554145048e-06,
+      "loss": 0.1104,
+      "step": 38323
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1294477188386824e-06,
+      "loss": 0.4198,
+      "step": 38324
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.129111883532316e-06,
+      "loss": 0.2087,
+      "step": 38325
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.128776048225951e-06,
+      "loss": 0.5069,
+      "step": 38326
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.128440212919585e-06,
+      "loss": 0.1679,
+      "step": 38327
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1281043776132185e-06,
+      "loss": 0.3966,
+      "step": 38328
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.127768542306853e-06,
+      "loss": 0.2024,
+      "step": 38329
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.127432707000488e-06,
+      "loss": 0.3514,
+      "step": 38330
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.127096871694122e-06,
+      "loss": 0.3359,
+      "step": 38331
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.126761036387756e-06,
+      "loss": 0.2241,
+      "step": 38332
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.126425201081391e-06,
+      "loss": 0.2117,
+      "step": 38333
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.126089365775024e-06,
+      "loss": 0.3587,
+      "step": 38334
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1257535304686585e-06,
+      "loss": 0.2639,
+      "step": 38335
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.125417695162293e-06,
+      "loss": 0.3197,
+      "step": 38336
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.125081859855927e-06,
+      "loss": 0.3064,
+      "step": 38337
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1247460245495616e-06,
+      "loss": 0.1751,
+      "step": 38338
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.124410189243196e-06,
+      "loss": 0.2975,
+      "step": 38339
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.124074353936829e-06,
+      "loss": 0.1996,
+      "step": 38340
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.123738518630464e-06,
+      "loss": 0.3302,
+      "step": 38341
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1234026833240985e-06,
+      "loss": 0.3933,
+      "step": 38342
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.123066848017732e-06,
+      "loss": 0.3009,
+      "step": 38343
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.122731012711367e-06,
+      "loss": 0.4847,
+      "step": 38344
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1223951774050016e-06,
+      "loss": 0.1769,
+      "step": 38345
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.122059342098635e-06,
+      "loss": 0.2193,
+      "step": 38346
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.12172350679227e-06,
+      "loss": 0.2187,
+      "step": 38347
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.121387671485905e-06,
+      "loss": 0.3148,
+      "step": 38348
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.121051836179538e-06,
+      "loss": 0.2597,
+      "step": 38349
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.120716000873172e-06,
+      "loss": 0.2867,
+      "step": 38350
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.120380165566807e-06,
+      "loss": 0.2167,
+      "step": 38351
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.120044330260441e-06,
+      "loss": 0.2549,
+      "step": 38352
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.119708494954075e-06,
+      "loss": 0.232,
+      "step": 38353
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.11937265964771e-06,
+      "loss": 0.4123,
+      "step": 38354
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.119036824341343e-06,
+      "loss": 0.1573,
+      "step": 38355
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.118700989034978e-06,
+      "loss": 0.3233,
+      "step": 38356
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.118365153728612e-06,
+      "loss": 0.2032,
+      "step": 38357
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.118029318422246e-06,
+      "loss": 0.3008,
+      "step": 38358
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.117693483115881e-06,
+      "loss": 0.5134,
+      "step": 38359
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.117357647809515e-06,
+      "loss": 0.1545,
+      "step": 38360
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.117021812503148e-06,
+      "loss": 0.2866,
+      "step": 38361
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.116685977196783e-06,
+      "loss": 0.2128,
+      "step": 38362
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.116350141890418e-06,
+      "loss": 0.2959,
+      "step": 38363
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.116014306584051e-06,
+      "loss": 0.2295,
+      "step": 38364
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.115678471277686e-06,
+      "loss": 0.2792,
+      "step": 38365
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.115342635971321e-06,
+      "loss": 0.1945,
+      "step": 38366
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1150068006649545e-06,
+      "loss": 0.2289,
+      "step": 38367
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.114670965358589e-06,
+      "loss": 0.2349,
+      "step": 38368
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.114335130052224e-06,
+      "loss": 0.2334,
+      "step": 38369
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.113999294745857e-06,
+      "loss": 0.3321,
+      "step": 38370
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.113663459439491e-06,
+      "loss": 0.2914,
+      "step": 38371
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.113327624133126e-06,
+      "loss": 0.3498,
+      "step": 38372
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.11299178882676e-06,
+      "loss": 0.214,
+      "step": 38373
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1126559535203945e-06,
+      "loss": 0.2895,
+      "step": 38374
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.112320118214029e-06,
+      "loss": 0.1652,
+      "step": 38375
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.111984282907662e-06,
+      "loss": 0.1923,
+      "step": 38376
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.111648447601297e-06,
+      "loss": 0.2845,
+      "step": 38377
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.111312612294931e-06,
+      "loss": 0.4402,
+      "step": 38378
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.110976776988565e-06,
+      "loss": 0.2649,
+      "step": 38379
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1106409416822e-06,
+      "loss": 0.3639,
+      "step": 38380
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1103051063758345e-06,
+      "loss": 0.3904,
+      "step": 38381
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.109969271069468e-06,
+      "loss": 0.1888,
+      "step": 38382
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.109633435763103e-06,
+      "loss": 0.3875,
+      "step": 38383
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.109297600456737e-06,
+      "loss": 0.3699,
+      "step": 38384
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1089617651503705e-06,
+      "loss": 0.1377,
+      "step": 38385
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.108625929844005e-06,
+      "loss": 0.2024,
+      "step": 38386
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.10829009453764e-06,
+      "loss": 0.1143,
+      "step": 38387
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.107954259231274e-06,
+      "loss": 0.1799,
+      "step": 38388
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.107618423924908e-06,
+      "loss": 0.1987,
+      "step": 38389
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.107282588618543e-06,
+      "loss": 0.1694,
+      "step": 38390
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.106946753312176e-06,
+      "loss": 0.4318,
+      "step": 38391
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1066109180058105e-06,
+      "loss": 0.1232,
+      "step": 38392
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.106275082699445e-06,
+      "loss": 0.4812,
+      "step": 38393
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.105939247393079e-06,
+      "loss": 0.2959,
+      "step": 38394
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.105603412086714e-06,
+      "loss": 0.153,
+      "step": 38395
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.1052675767803466e-06,
+      "loss": 0.2694,
+      "step": 38396
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.104931741473981e-06,
+      "loss": 0.2839,
+      "step": 38397
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.104595906167616e-06,
+      "loss": 0.3811,
+      "step": 38398
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.10426007086125e-06,
+      "loss": 0.4505,
+      "step": 38399
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.103924235554884e-06,
+      "loss": 0.3734,
+      "step": 38400
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.103588400248519e-06,
+      "loss": 0.3482,
+      "step": 38401
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.103252564942153e-06,
+      "loss": 0.2864,
+      "step": 38402
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.102916729635787e-06,
+      "loss": 0.2107,
+      "step": 38403
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.102580894329422e-06,
+      "loss": 0.257,
+      "step": 38404
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.102245059023055e-06,
+      "loss": 0.2153,
+      "step": 38405
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.10190922371669e-06,
+      "loss": 0.3638,
+      "step": 38406
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.101573388410324e-06,
+      "loss": 0.3109,
+      "step": 38407
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.101237553103958e-06,
+      "loss": 0.3408,
+      "step": 38408
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.100901717797593e-06,
+      "loss": 0.2339,
+      "step": 38409
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.100565882491227e-06,
+      "loss": 0.2383,
+      "step": 38410
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 7.10023004718486e-06,
+      "loss": 0.2208,
+      "step": 38411
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.099894211878495e-06,
+      "loss": 0.2479,
+      "step": 38412
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.09955837657213e-06,
+      "loss": 0.1678,
+      "step": 38413
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0992225412657634e-06,
+      "loss": 0.1889,
+      "step": 38414
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.098886705959398e-06,
+      "loss": 0.3044,
+      "step": 38415
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.098550870653033e-06,
+      "loss": 0.3777,
+      "step": 38416
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0982150353466665e-06,
+      "loss": 0.365,
+      "step": 38417
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0978792000403e-06,
+      "loss": 0.3849,
+      "step": 38418
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.097543364733935e-06,
+      "loss": 0.294,
+      "step": 38419
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.097207529427569e-06,
+      "loss": 0.3519,
+      "step": 38420
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0968716941212034e-06,
+      "loss": 0.2166,
+      "step": 38421
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.096535858814838e-06,
+      "loss": 0.444,
+      "step": 38422
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.096200023508472e-06,
+      "loss": 0.4623,
+      "step": 38423
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0958641882021065e-06,
+      "loss": 0.269,
+      "step": 38424
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.095528352895741e-06,
+      "loss": 0.398,
+      "step": 38425
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.095192517589374e-06,
+      "loss": 0.3057,
+      "step": 38426
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.094856682283009e-06,
+      "loss": 0.3323,
+      "step": 38427
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.094520846976643e-06,
+      "loss": 0.3081,
+      "step": 38428
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.094185011670277e-06,
+      "loss": 0.2562,
+      "step": 38429
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.093849176363912e-06,
+      "loss": 0.318,
+      "step": 38430
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0935133410575465e-06,
+      "loss": 0.3511,
+      "step": 38431
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0931775057511795e-06,
+      "loss": 0.493,
+      "step": 38432
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.092841670444814e-06,
+      "loss": 0.2588,
+      "step": 38433
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.092505835138449e-06,
+      "loss": 0.1789,
+      "step": 38434
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0921699998320826e-06,
+      "loss": 0.1968,
+      "step": 38435
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.091834164525717e-06,
+      "loss": 0.4075,
+      "step": 38436
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.091498329219352e-06,
+      "loss": 0.424,
+      "step": 38437
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.091162493912986e-06,
+      "loss": 0.229,
+      "step": 38438
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.09082665860662e-06,
+      "loss": 0.3558,
+      "step": 38439
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.090490823300254e-06,
+      "loss": 0.4161,
+      "step": 38440
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.090154987993888e-06,
+      "loss": 0.2857,
+      "step": 38441
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0898191526875225e-06,
+      "loss": 0.157,
+      "step": 38442
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.089483317381157e-06,
+      "loss": 0.2923,
+      "step": 38443
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.089147482074791e-06,
+      "loss": 0.3113,
+      "step": 38444
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.088811646768426e-06,
+      "loss": 0.2427,
+      "step": 38445
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.08847581146206e-06,
+      "loss": 0.1723,
+      "step": 38446
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.088139976155693e-06,
+      "loss": 0.2673,
+      "step": 38447
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.087804140849328e-06,
+      "loss": 0.2291,
+      "step": 38448
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0874683055429625e-06,
+      "loss": 0.2606,
+      "step": 38449
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.087132470236596e-06,
+      "loss": 0.1722,
+      "step": 38450
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.086796634930231e-06,
+      "loss": 0.1342,
+      "step": 38451
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.086460799623866e-06,
+      "loss": 0.3033,
+      "step": 38452
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.086124964317499e-06,
+      "loss": 0.2526,
+      "step": 38453
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.085789129011133e-06,
+      "loss": 0.3652,
+      "step": 38454
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.085453293704768e-06,
+      "loss": 0.2776,
+      "step": 38455
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.085117458398402e-06,
+      "loss": 0.3392,
+      "step": 38456
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.084781623092036e-06,
+      "loss": 0.1926,
+      "step": 38457
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.084445787785671e-06,
+      "loss": 0.1981,
+      "step": 38458
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.084109952479305e-06,
+      "loss": 0.2919,
+      "step": 38459
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0837741171729394e-06,
+      "loss": 0.2861,
+      "step": 38460
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.083438281866574e-06,
+      "loss": 0.1994,
+      "step": 38461
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.083102446560207e-06,
+      "loss": 0.2226,
+      "step": 38462
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.082766611253842e-06,
+      "loss": 0.2654,
+      "step": 38463
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.082430775947476e-06,
+      "loss": 0.2339,
+      "step": 38464
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.08209494064111e-06,
+      "loss": 0.2501,
+      "step": 38465
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.081759105334745e-06,
+      "loss": 0.3238,
+      "step": 38466
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.081423270028379e-06,
+      "loss": 0.335,
+      "step": 38467
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.081087434722012e-06,
+      "loss": 0.2589,
+      "step": 38468
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.080751599415647e-06,
+      "loss": 0.2879,
+      "step": 38469
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.080415764109282e-06,
+      "loss": 0.2326,
+      "step": 38470
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0800799288029155e-06,
+      "loss": 0.2319,
+      "step": 38471
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.07974409349655e-06,
+      "loss": 0.3278,
+      "step": 38472
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.079408258190185e-06,
+      "loss": 0.2732,
+      "step": 38473
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.079072422883818e-06,
+      "loss": 0.3103,
+      "step": 38474
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.078736587577452e-06,
+      "loss": 0.317,
+      "step": 38475
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.078400752271087e-06,
+      "loss": 0.2971,
+      "step": 38476
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.078064916964721e-06,
+      "loss": 0.3854,
+      "step": 38477
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0777290816583555e-06,
+      "loss": 0.2551,
+      "step": 38478
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.07739324635199e-06,
+      "loss": 0.3064,
+      "step": 38479
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.077057411045624e-06,
+      "loss": 0.2974,
+      "step": 38480
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0767215757392586e-06,
+      "loss": 0.2085,
+      "step": 38481
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.076385740432893e-06,
+      "loss": 0.362,
+      "step": 38482
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.076049905126526e-06,
+      "loss": 0.2908,
+      "step": 38483
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.075714069820161e-06,
+      "loss": 0.35,
+      "step": 38484
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.075378234513795e-06,
+      "loss": 0.1972,
+      "step": 38485
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.075042399207429e-06,
+      "loss": 0.2901,
+      "step": 38486
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.074706563901064e-06,
+      "loss": 0.2418,
+      "step": 38487
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.074370728594697e-06,
+      "loss": 0.3421,
+      "step": 38488
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0740348932883315e-06,
+      "loss": 0.2451,
+      "step": 38489
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.073699057981966e-06,
+      "loss": 0.1691,
+      "step": 38490
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0733632226756e-06,
+      "loss": 0.2992,
+      "step": 38491
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.073027387369235e-06,
+      "loss": 0.2935,
+      "step": 38492
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.072691552062869e-06,
+      "loss": 0.2815,
+      "step": 38493
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.072355716756503e-06,
+      "loss": 0.1855,
+      "step": 38494
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.072019881450138e-06,
+      "loss": 0.2913,
+      "step": 38495
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0716840461437715e-06,
+      "loss": 0.322,
+      "step": 38496
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.071348210837405e-06,
+      "loss": 0.2939,
+      "step": 38497
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.07101237553104e-06,
+      "loss": 0.5391,
+      "step": 38498
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.070676540224675e-06,
+      "loss": 0.3734,
+      "step": 38499
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.070340704918308e-06,
+      "loss": 0.3918,
+      "step": 38500
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.070004869611943e-06,
+      "loss": 0.1293,
+      "step": 38501
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.069669034305578e-06,
+      "loss": 0.2477,
+      "step": 38502
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.069333198999211e-06,
+      "loss": 0.1787,
+      "step": 38503
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.068997363692845e-06,
+      "loss": 0.2861,
+      "step": 38504
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.06866152838648e-06,
+      "loss": 0.2378,
+      "step": 38505
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.068325693080114e-06,
+      "loss": 0.2397,
+      "step": 38506
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.067989857773748e-06,
+      "loss": 0.3017,
+      "step": 38507
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.067654022467383e-06,
+      "loss": 0.2486,
+      "step": 38508
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.067318187161016e-06,
+      "loss": 0.3117,
+      "step": 38509
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.066982351854651e-06,
+      "loss": 0.4506,
+      "step": 38510
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.066646516548285e-06,
+      "loss": 0.3,
+      "step": 38511
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.066310681241919e-06,
+      "loss": 0.1807,
+      "step": 38512
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.065974845935554e-06,
+      "loss": 0.2063,
+      "step": 38513
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.065639010629188e-06,
+      "loss": 0.6546,
+      "step": 38514
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.065303175322822e-06,
+      "loss": 0.1623,
+      "step": 38515
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.064967340016457e-06,
+      "loss": 0.1722,
+      "step": 38516
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0646315047100915e-06,
+      "loss": 0.2667,
+      "step": 38517
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.064295669403724e-06,
+      "loss": 0.2396,
+      "step": 38518
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.063959834097359e-06,
+      "loss": 0.2123,
+      "step": 38519
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.063623998790994e-06,
+      "loss": 0.2439,
+      "step": 38520
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0632881634846275e-06,
+      "loss": 0.371,
+      "step": 38521
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.062952328178262e-06,
+      "loss": 0.4162,
+      "step": 38522
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.062616492871897e-06,
+      "loss": 0.3656,
+      "step": 38523
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.06228065756553e-06,
+      "loss": 0.4015,
+      "step": 38524
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.061944822259164e-06,
+      "loss": 0.2566,
+      "step": 38525
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.061608986952799e-06,
+      "loss": 0.3314,
+      "step": 38526
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.061273151646433e-06,
+      "loss": 0.3613,
+      "step": 38527
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0609373163400675e-06,
+      "loss": 0.2942,
+      "step": 38528
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.060601481033702e-06,
+      "loss": 0.2266,
+      "step": 38529
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.060265645727336e-06,
+      "loss": 0.0793,
+      "step": 38530
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.05992981042097e-06,
+      "loss": 0.2411,
+      "step": 38531
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.059593975114604e-06,
+      "loss": 0.3291,
+      "step": 38532
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.059258139808238e-06,
+      "loss": 0.2667,
+      "step": 38533
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.058922304501873e-06,
+      "loss": 0.2945,
+      "step": 38534
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0585864691955075e-06,
+      "loss": 0.2439,
+      "step": 38535
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.058250633889141e-06,
+      "loss": 0.3272,
+      "step": 38536
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.057914798582776e-06,
+      "loss": 0.2717,
+      "step": 38537
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.057578963276411e-06,
+      "loss": 0.2639,
+      "step": 38538
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0572431279700435e-06,
+      "loss": 0.263,
+      "step": 38539
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.056907292663678e-06,
+      "loss": 0.1589,
+      "step": 38540
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.056571457357313e-06,
+      "loss": 0.1135,
+      "step": 38541
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.056235622050947e-06,
+      "loss": 0.3433,
+      "step": 38542
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.055899786744581e-06,
+      "loss": 0.3419,
+      "step": 38543
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.055563951438216e-06,
+      "loss": 0.2861,
+      "step": 38544
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.055228116131849e-06,
+      "loss": 0.4822,
+      "step": 38545
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0548922808254835e-06,
+      "loss": 0.2133,
+      "step": 38546
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.054556445519118e-06,
+      "loss": 0.3149,
+      "step": 38547
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.054220610212752e-06,
+      "loss": 0.2041,
+      "step": 38548
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.053884774906387e-06,
+      "loss": 0.1824,
+      "step": 38549
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.053548939600021e-06,
+      "loss": 0.3483,
+      "step": 38550
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.053213104293655e-06,
+      "loss": 0.3056,
+      "step": 38551
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.05287726898729e-06,
+      "loss": 0.2386,
+      "step": 38552
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0525414336809235e-06,
+      "loss": 0.2627,
+      "step": 38553
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.052205598374557e-06,
+      "loss": 0.3287,
+      "step": 38554
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.051869763068192e-06,
+      "loss": 0.2971,
+      "step": 38555
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.051533927761827e-06,
+      "loss": 0.1842,
+      "step": 38556
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.05119809245546e-06,
+      "loss": 0.2668,
+      "step": 38557
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.050862257149095e-06,
+      "loss": 0.3059,
+      "step": 38558
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.05052642184273e-06,
+      "loss": 0.3777,
+      "step": 38559
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.050190586536363e-06,
+      "loss": 0.3203,
+      "step": 38560
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.049854751229997e-06,
+      "loss": 0.3146,
+      "step": 38561
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.049518915923632e-06,
+      "loss": 0.4021,
+      "step": 38562
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.049183080617266e-06,
+      "loss": 0.2749,
+      "step": 38563
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0488472453109e-06,
+      "loss": 0.4215,
+      "step": 38564
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.048511410004535e-06,
+      "loss": 0.218,
+      "step": 38565
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.048175574698168e-06,
+      "loss": 0.3139,
+      "step": 38566
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.047839739391803e-06,
+      "loss": 0.3051,
+      "step": 38567
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.047503904085437e-06,
+      "loss": 0.3288,
+      "step": 38568
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.047168068779071e-06,
+      "loss": 0.1389,
+      "step": 38569
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.046832233472706e-06,
+      "loss": 0.1851,
+      "step": 38570
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.04649639816634e-06,
+      "loss": 0.1655,
+      "step": 38571
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.046160562859974e-06,
+      "loss": 0.2748,
+      "step": 38572
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.045824727553609e-06,
+      "loss": 0.2543,
+      "step": 38573
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0454888922472435e-06,
+      "loss": 0.3992,
+      "step": 38574
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0451530569408764e-06,
+      "loss": 0.3442,
+      "step": 38575
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.044817221634511e-06,
+      "loss": 0.3327,
+      "step": 38576
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.044481386328145e-06,
+      "loss": 0.2041,
+      "step": 38577
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0441455510217795e-06,
+      "loss": 0.2947,
+      "step": 38578
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.043809715715414e-06,
+      "loss": 0.2754,
+      "step": 38579
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.043473880409047e-06,
+      "loss": 0.301,
+      "step": 38580
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.043138045102682e-06,
+      "loss": 0.3812,
+      "step": 38581
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0428022097963164e-06,
+      "loss": 0.2854,
+      "step": 38582
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.04246637448995e-06,
+      "loss": 0.3825,
+      "step": 38583
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.042130539183585e-06,
+      "loss": 0.3803,
+      "step": 38584
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0417947038772195e-06,
+      "loss": 0.2078,
+      "step": 38585
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.041458868570853e-06,
+      "loss": 0.5006,
+      "step": 38586
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.041123033264487e-06,
+      "loss": 0.3305,
+      "step": 38587
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.040787197958122e-06,
+      "loss": 0.2779,
+      "step": 38588
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0404513626517556e-06,
+      "loss": 0.4325,
+      "step": 38589
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.04011552734539e-06,
+      "loss": 0.1967,
+      "step": 38590
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.039779692039025e-06,
+      "loss": 0.2547,
+      "step": 38591
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.039443856732659e-06,
+      "loss": 0.3601,
+      "step": 38592
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.039108021426293e-06,
+      "loss": 0.2433,
+      "step": 38593
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.038772186119928e-06,
+      "loss": 0.2341,
+      "step": 38594
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.038436350813561e-06,
+      "loss": 0.2621,
+      "step": 38595
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0381005155071956e-06,
+      "loss": 0.2533,
+      "step": 38596
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.03776468020083e-06,
+      "loss": 0.1883,
+      "step": 38597
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.037428844894464e-06,
+      "loss": 0.4362,
+      "step": 38598
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.037093009588099e-06,
+      "loss": 0.3616,
+      "step": 38599
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.036757174281733e-06,
+      "loss": 0.25,
+      "step": 38600
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.036421338975366e-06,
+      "loss": 0.2249,
+      "step": 38601
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.036085503669001e-06,
+      "loss": 0.4383,
+      "step": 38602
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0357496683626356e-06,
+      "loss": 0.2465,
+      "step": 38603
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.035413833056269e-06,
+      "loss": 0.4177,
+      "step": 38604
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.035077997749904e-06,
+      "loss": 0.2893,
+      "step": 38605
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.034742162443539e-06,
+      "loss": 0.42,
+      "step": 38606
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.0344063271371725e-06,
+      "loss": 0.2453,
+      "step": 38607
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.034070491830807e-06,
+      "loss": 0.3093,
+      "step": 38608
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.033734656524441e-06,
+      "loss": 0.2071,
+      "step": 38609
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 7.033398821218075e-06,
+      "loss": 0.2692,
+      "step": 38610
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.033062985911709e-06,
+      "loss": 0.3982,
+      "step": 38611
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.032727150605344e-06,
+      "loss": 0.3015,
+      "step": 38612
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.032391315298978e-06,
+      "loss": 0.3571,
+      "step": 38613
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.0320554799926124e-06,
+      "loss": 0.2637,
+      "step": 38614
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.031719644686247e-06,
+      "loss": 0.1376,
+      "step": 38615
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.03138380937988e-06,
+      "loss": 0.12,
+      "step": 38616
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.031047974073515e-06,
+      "loss": 0.1893,
+      "step": 38617
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.030712138767149e-06,
+      "loss": 0.2675,
+      "step": 38618
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.030376303460783e-06,
+      "loss": 0.3576,
+      "step": 38619
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.030040468154418e-06,
+      "loss": 0.2276,
+      "step": 38620
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.0297046328480524e-06,
+      "loss": 0.2675,
+      "step": 38621
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.029368797541685e-06,
+      "loss": 0.3326,
+      "step": 38622
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.02903296223532e-06,
+      "loss": 0.3377,
+      "step": 38623
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.028697126928955e-06,
+      "loss": 0.2279,
+      "step": 38624
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.0283612916225885e-06,
+      "loss": 0.1837,
+      "step": 38625
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.028025456316223e-06,
+      "loss": 0.1924,
+      "step": 38626
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.027689621009858e-06,
+      "loss": 0.3005,
+      "step": 38627
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.027353785703492e-06,
+      "loss": 0.3833,
+      "step": 38628
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.027017950397126e-06,
+      "loss": 0.3385,
+      "step": 38629
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.026682115090761e-06,
+      "loss": 0.3001,
+      "step": 38630
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.026346279784394e-06,
+      "loss": 0.184,
+      "step": 38631
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.0260104444780285e-06,
+      "loss": 0.3543,
+      "step": 38632
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.025674609171663e-06,
+      "loss": 0.3101,
+      "step": 38633
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.025338773865297e-06,
+      "loss": 0.2537,
+      "step": 38634
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.0250029385589316e-06,
+      "loss": 0.25,
+      "step": 38635
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.024667103252566e-06,
+      "loss": 0.3048,
+      "step": 38636
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.024331267946199e-06,
+      "loss": 0.3385,
+      "step": 38637
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.023995432639834e-06,
+      "loss": 0.267,
+      "step": 38638
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.0236595973334685e-06,
+      "loss": 0.299,
+      "step": 38639
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.023323762027102e-06,
+      "loss": 0.1734,
+      "step": 38640
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.022987926720737e-06,
+      "loss": 0.1678,
+      "step": 38641
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.0226520914143716e-06,
+      "loss": 0.4622,
+      "step": 38642
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.022316256108005e-06,
+      "loss": 0.2678,
+      "step": 38643
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.021980420801639e-06,
+      "loss": 0.369,
+      "step": 38644
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.021644585495274e-06,
+      "loss": 0.3169,
+      "step": 38645
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.021308750188908e-06,
+      "loss": 0.1179,
+      "step": 38646
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.020972914882542e-06,
+      "loss": 0.3557,
+      "step": 38647
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.020637079576177e-06,
+      "loss": 0.3144,
+      "step": 38648
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.020301244269811e-06,
+      "loss": 0.2863,
+      "step": 38649
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.019965408963445e-06,
+      "loss": 0.3773,
+      "step": 38650
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.01962957365708e-06,
+      "loss": 0.2343,
+      "step": 38651
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.019293738350713e-06,
+      "loss": 0.2766,
+      "step": 38652
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.018957903044348e-06,
+      "loss": 0.3718,
+      "step": 38653
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.018622067737982e-06,
+      "loss": 0.2145,
+      "step": 38654
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.018286232431616e-06,
+      "loss": 0.2505,
+      "step": 38655
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.017950397125251e-06,
+      "loss": 0.4393,
+      "step": 38656
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.017614561818885e-06,
+      "loss": 0.3031,
+      "step": 38657
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.017278726512518e-06,
+      "loss": 0.3657,
+      "step": 38658
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.016942891206153e-06,
+      "loss": 0.1895,
+      "step": 38659
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.016607055899788e-06,
+      "loss": 0.4427,
+      "step": 38660
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.016271220593421e-06,
+      "loss": 0.1663,
+      "step": 38661
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.015935385287056e-06,
+      "loss": 0.2101,
+      "step": 38662
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.015599549980691e-06,
+      "loss": 0.3355,
+      "step": 38663
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.0152637146743245e-06,
+      "loss": 0.441,
+      "step": 38664
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.014927879367959e-06,
+      "loss": 0.203,
+      "step": 38665
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.014592044061592e-06,
+      "loss": 0.28,
+      "step": 38666
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.014256208755227e-06,
+      "loss": 0.1812,
+      "step": 38667
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.013920373448861e-06,
+      "loss": 0.3004,
+      "step": 38668
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.013584538142495e-06,
+      "loss": 0.378,
+      "step": 38669
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.01324870283613e-06,
+      "loss": 0.2983,
+      "step": 38670
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.0129128675297645e-06,
+      "loss": 0.2732,
+      "step": 38671
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.0125770322233974e-06,
+      "loss": 0.2993,
+      "step": 38672
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.012241196917032e-06,
+      "loss": 0.217,
+      "step": 38673
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.011905361610667e-06,
+      "loss": 0.3267,
+      "step": 38674
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.0115695263043005e-06,
+      "loss": 0.4381,
+      "step": 38675
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.011233690997935e-06,
+      "loss": 0.285,
+      "step": 38676
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.01089785569157e-06,
+      "loss": 0.5268,
+      "step": 38677
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.010562020385203e-06,
+      "loss": 0.1458,
+      "step": 38678
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.010226185078837e-06,
+      "loss": 0.3255,
+      "step": 38679
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.009890349772472e-06,
+      "loss": 0.2355,
+      "step": 38680
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.009554514466106e-06,
+      "loss": 0.2212,
+      "step": 38681
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.0092186791597405e-06,
+      "loss": 0.2911,
+      "step": 38682
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.008882843853375e-06,
+      "loss": 0.2664,
+      "step": 38683
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.008547008547009e-06,
+      "loss": 0.2941,
+      "step": 38684
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.008211173240644e-06,
+      "loss": 0.1902,
+      "step": 38685
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.007875337934278e-06,
+      "loss": 0.4234,
+      "step": 38686
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.007539502627911e-06,
+      "loss": 0.2971,
+      "step": 38687
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.007203667321546e-06,
+      "loss": 0.4283,
+      "step": 38688
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.0068678320151805e-06,
+      "loss": 0.3228,
+      "step": 38689
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.006531996708814e-06,
+      "loss": 0.3,
+      "step": 38690
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.006196161402449e-06,
+      "loss": 0.1999,
+      "step": 38691
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.005860326096084e-06,
+      "loss": 0.471,
+      "step": 38692
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.0055244907897166e-06,
+      "loss": 0.1886,
+      "step": 38693
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.005188655483351e-06,
+      "loss": 0.4552,
+      "step": 38694
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.004852820176986e-06,
+      "loss": 0.3494,
+      "step": 38695
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.00451698487062e-06,
+      "loss": 0.3104,
+      "step": 38696
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.004181149564254e-06,
+      "loss": 0.2857,
+      "step": 38697
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.003845314257889e-06,
+      "loss": 0.2152,
+      "step": 38698
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.003509478951523e-06,
+      "loss": 0.2467,
+      "step": 38699
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.0031736436451565e-06,
+      "loss": 0.313,
+      "step": 38700
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.002837808338791e-06,
+      "loss": 0.2533,
+      "step": 38701
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.002501973032425e-06,
+      "loss": 0.2659,
+      "step": 38702
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.00216613772606e-06,
+      "loss": 0.3963,
+      "step": 38703
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.001830302419694e-06,
+      "loss": 0.33,
+      "step": 38704
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.001494467113328e-06,
+      "loss": 0.5417,
+      "step": 38705
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.001158631806963e-06,
+      "loss": 0.3758,
+      "step": 38706
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.000822796500597e-06,
+      "loss": 0.2936,
+      "step": 38707
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.00048696119423e-06,
+      "loss": 0.2803,
+      "step": 38708
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 7.000151125887865e-06,
+      "loss": 0.2905,
+      "step": 38709
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.9998152905815e-06,
+      "loss": 0.2435,
+      "step": 38710
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.9994794552751334e-06,
+      "loss": 0.3761,
+      "step": 38711
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.999143619968768e-06,
+      "loss": 0.2973,
+      "step": 38712
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.998807784662403e-06,
+      "loss": 0.4149,
+      "step": 38713
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.998471949356036e-06,
+      "loss": 0.2849,
+      "step": 38714
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.99813611404967e-06,
+      "loss": 0.349,
+      "step": 38715
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.997800278743305e-06,
+      "loss": 0.3547,
+      "step": 38716
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.997464443436939e-06,
+      "loss": 0.3738,
+      "step": 38717
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.997128608130573e-06,
+      "loss": 0.4363,
+      "step": 38718
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.996792772824208e-06,
+      "loss": 0.3648,
+      "step": 38719
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.996456937517842e-06,
+      "loss": 0.1949,
+      "step": 38720
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.9961211022114765e-06,
+      "loss": 0.3346,
+      "step": 38721
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.99578526690511e-06,
+      "loss": 0.2712,
+      "step": 38722
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.995449431598744e-06,
+      "loss": 0.3333,
+      "step": 38723
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.995113596292379e-06,
+      "loss": 0.2513,
+      "step": 38724
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.994777760986013e-06,
+      "loss": 0.2696,
+      "step": 38725
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.994441925679647e-06,
+      "loss": 0.2642,
+      "step": 38726
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.994106090373282e-06,
+      "loss": 0.2961,
+      "step": 38727
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.9937702550669165e-06,
+      "loss": 0.3667,
+      "step": 38728
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.9934344197605495e-06,
+      "loss": 0.2041,
+      "step": 38729
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.993098584454184e-06,
+      "loss": 0.1794,
+      "step": 38730
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.992762749147819e-06,
+      "loss": 0.2216,
+      "step": 38731
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.9924269138414526e-06,
+      "loss": 0.3136,
+      "step": 38732
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.992091078535087e-06,
+      "loss": 0.285,
+      "step": 38733
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.991755243228722e-06,
+      "loss": 0.4029,
+      "step": 38734
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.991419407922355e-06,
+      "loss": 0.2772,
+      "step": 38735
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.9910835726159894e-06,
+      "loss": 0.3056,
+      "step": 38736
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.990747737309624e-06,
+      "loss": 0.2147,
+      "step": 38737
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.990411902003258e-06,
+      "loss": 0.2468,
+      "step": 38738
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.9900760666968925e-06,
+      "loss": 0.3052,
+      "step": 38739
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.989740231390527e-06,
+      "loss": 0.1979,
+      "step": 38740
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.989404396084161e-06,
+      "loss": 0.2398,
+      "step": 38741
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.989068560777796e-06,
+      "loss": 0.2925,
+      "step": 38742
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.98873272547143e-06,
+      "loss": 0.193,
+      "step": 38743
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.988396890165063e-06,
+      "loss": 0.327,
+      "step": 38744
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.988061054858698e-06,
+      "loss": 0.3055,
+      "step": 38745
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.9877252195523325e-06,
+      "loss": 0.2056,
+      "step": 38746
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.987389384245966e-06,
+      "loss": 0.2858,
+      "step": 38747
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.987053548939601e-06,
+      "loss": 0.4249,
+      "step": 38748
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.986717713633236e-06,
+      "loss": 0.2617,
+      "step": 38749
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.986381878326869e-06,
+      "loss": 0.1863,
+      "step": 38750
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.986046043020503e-06,
+      "loss": 0.295,
+      "step": 38751
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.985710207714138e-06,
+      "loss": 0.3016,
+      "step": 38752
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.985374372407772e-06,
+      "loss": 0.2248,
+      "step": 38753
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.985038537101406e-06,
+      "loss": 0.4066,
+      "step": 38754
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.98470270179504e-06,
+      "loss": 0.3722,
+      "step": 38755
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.984366866488674e-06,
+      "loss": 0.2186,
+      "step": 38756
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.9840310311823086e-06,
+      "loss": 0.2896,
+      "step": 38757
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.983695195875942e-06,
+      "loss": 0.2508,
+      "step": 38758
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.983359360569577e-06,
+      "loss": 0.3245,
+      "step": 38759
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.983023525263212e-06,
+      "loss": 0.3553,
+      "step": 38760
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.9826876899568455e-06,
+      "loss": 0.2997,
+      "step": 38761
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.98235185465048e-06,
+      "loss": 0.3581,
+      "step": 38762
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.982016019344115e-06,
+      "loss": 0.3149,
+      "step": 38763
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.981680184037748e-06,
+      "loss": 0.1936,
+      "step": 38764
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.981344348731382e-06,
+      "loss": 0.3618,
+      "step": 38765
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.981008513425017e-06,
+      "loss": 0.3661,
+      "step": 38766
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.980672678118651e-06,
+      "loss": 0.3345,
+      "step": 38767
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.9803368428122855e-06,
+      "loss": 0.2687,
+      "step": 38768
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.98000100750592e-06,
+      "loss": 0.239,
+      "step": 38769
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.979665172199553e-06,
+      "loss": 0.432,
+      "step": 38770
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.979329336893188e-06,
+      "loss": 0.2633,
+      "step": 38771
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.978993501586822e-06,
+      "loss": 0.1806,
+      "step": 38772
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.978657666280456e-06,
+      "loss": 0.39,
+      "step": 38773
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.978321830974091e-06,
+      "loss": 0.417,
+      "step": 38774
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.9779859956677255e-06,
+      "loss": 0.3354,
+      "step": 38775
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.977650160361359e-06,
+      "loss": 0.2752,
+      "step": 38776
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.977314325054994e-06,
+      "loss": 0.4029,
+      "step": 38777
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.9769784897486285e-06,
+      "loss": 0.388,
+      "step": 38778
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.9766426544422615e-06,
+      "loss": 0.2125,
+      "step": 38779
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.976306819135896e-06,
+      "loss": 0.2653,
+      "step": 38780
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.975970983829531e-06,
+      "loss": 0.2036,
+      "step": 38781
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.975635148523165e-06,
+      "loss": 0.2519,
+      "step": 38782
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.975299313216799e-06,
+      "loss": 0.3324,
+      "step": 38783
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.974963477910434e-06,
+      "loss": 0.4149,
+      "step": 38784
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.974627642604067e-06,
+      "loss": 0.3063,
+      "step": 38785
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.9742918072977015e-06,
+      "loss": 0.3023,
+      "step": 38786
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.973955971991336e-06,
+      "loss": 0.1535,
+      "step": 38787
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.97362013668497e-06,
+      "loss": 0.2137,
+      "step": 38788
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.973284301378605e-06,
+      "loss": 0.2676,
+      "step": 38789
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.972948466072239e-06,
+      "loss": 0.1575,
+      "step": 38790
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.972612630765872e-06,
+      "loss": 0.207,
+      "step": 38791
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.972276795459507e-06,
+      "loss": 0.4913,
+      "step": 38792
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.9719409601531415e-06,
+      "loss": 0.3204,
+      "step": 38793
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.971605124846775e-06,
+      "loss": 0.2472,
+      "step": 38794
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.97126928954041e-06,
+      "loss": 0.2483,
+      "step": 38795
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.9709334542340446e-06,
+      "loss": 0.3168,
+      "step": 38796
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.970597618927678e-06,
+      "loss": 0.1331,
+      "step": 38797
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.970261783621313e-06,
+      "loss": 0.385,
+      "step": 38798
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.969925948314948e-06,
+      "loss": 0.2373,
+      "step": 38799
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.969590113008581e-06,
+      "loss": 0.2504,
+      "step": 38800
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.969254277702215e-06,
+      "loss": 0.4633,
+      "step": 38801
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.96891844239585e-06,
+      "loss": 0.4445,
+      "step": 38802
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.968582607089484e-06,
+      "loss": 0.2942,
+      "step": 38803
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.968246771783118e-06,
+      "loss": 0.4967,
+      "step": 38804
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.967910936476753e-06,
+      "loss": 0.3726,
+      "step": 38805
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.967575101170386e-06,
+      "loss": 0.3191,
+      "step": 38806
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.967239265864021e-06,
+      "loss": 0.3585,
+      "step": 38807
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 6.966903430557655e-06,
+      "loss": 0.368,
+      "step": 38808
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.966567595251289e-06,
+      "loss": 0.2723,
+      "step": 38809
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.966231759944924e-06,
+      "loss": 0.296,
+      "step": 38810
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.965895924638558e-06,
+      "loss": 0.2897,
+      "step": 38811
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.965560089332192e-06,
+      "loss": 0.2791,
+      "step": 38812
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.965224254025826e-06,
+      "loss": 0.2691,
+      "step": 38813
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.964888418719461e-06,
+      "loss": 0.3936,
+      "step": 38814
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.964552583413094e-06,
+      "loss": 0.2964,
+      "step": 38815
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.964216748106729e-06,
+      "loss": 0.2581,
+      "step": 38816
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.963880912800364e-06,
+      "loss": 0.1483,
+      "step": 38817
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9635450774939975e-06,
+      "loss": 0.2916,
+      "step": 38818
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.963209242187632e-06,
+      "loss": 0.2799,
+      "step": 38819
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.962873406881267e-06,
+      "loss": 0.2667,
+      "step": 38820
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9625375715749e-06,
+      "loss": 0.2623,
+      "step": 38821
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.962201736268534e-06,
+      "loss": 0.3431,
+      "step": 38822
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.961865900962169e-06,
+      "loss": 0.2162,
+      "step": 38823
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.961530065655803e-06,
+      "loss": 0.2864,
+      "step": 38824
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9611942303494375e-06,
+      "loss": 0.3159,
+      "step": 38825
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.960858395043072e-06,
+      "loss": 0.3335,
+      "step": 38826
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.960522559736705e-06,
+      "loss": 0.1969,
+      "step": 38827
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.96018672443034e-06,
+      "loss": 0.289,
+      "step": 38828
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.959850889123974e-06,
+      "loss": 0.1658,
+      "step": 38829
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.959515053817608e-06,
+      "loss": 0.3383,
+      "step": 38830
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.959179218511243e-06,
+      "loss": 0.307,
+      "step": 38831
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9588433832048775e-06,
+      "loss": 0.3346,
+      "step": 38832
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.958507547898511e-06,
+      "loss": 0.2377,
+      "step": 38833
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.958171712592146e-06,
+      "loss": 0.3032,
+      "step": 38834
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.95783587728578e-06,
+      "loss": 0.171,
+      "step": 38835
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9575000419794135e-06,
+      "loss": 0.2833,
+      "step": 38836
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.957164206673048e-06,
+      "loss": 0.4197,
+      "step": 38837
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.956828371366683e-06,
+      "loss": 0.2841,
+      "step": 38838
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.956492536060317e-06,
+      "loss": 0.3665,
+      "step": 38839
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.956156700753951e-06,
+      "loss": 0.195,
+      "step": 38840
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.955820865447586e-06,
+      "loss": 0.2445,
+      "step": 38841
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.955485030141219e-06,
+      "loss": 0.3168,
+      "step": 38842
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9551491948348535e-06,
+      "loss": 0.2322,
+      "step": 38843
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.954813359528488e-06,
+      "loss": 0.2774,
+      "step": 38844
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.954477524222122e-06,
+      "loss": 0.5157,
+      "step": 38845
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.954141688915757e-06,
+      "loss": 0.3569,
+      "step": 38846
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9538058536093896e-06,
+      "loss": 0.3356,
+      "step": 38847
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.953470018303024e-06,
+      "loss": 0.4131,
+      "step": 38848
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.953134182996659e-06,
+      "loss": 0.3537,
+      "step": 38849
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.952798347690293e-06,
+      "loss": 0.2138,
+      "step": 38850
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.952462512383927e-06,
+      "loss": 0.2492,
+      "step": 38851
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.952126677077562e-06,
+      "loss": 0.2213,
+      "step": 38852
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.951790841771196e-06,
+      "loss": 0.2583,
+      "step": 38853
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.95145500646483e-06,
+      "loss": 0.2928,
+      "step": 38854
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.951119171158465e-06,
+      "loss": 0.4447,
+      "step": 38855
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.950783335852098e-06,
+      "loss": 0.3439,
+      "step": 38856
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.950447500545733e-06,
+      "loss": 0.1495,
+      "step": 38857
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.950111665239367e-06,
+      "loss": 0.2984,
+      "step": 38858
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.949775829933001e-06,
+      "loss": 0.365,
+      "step": 38859
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.949439994626636e-06,
+      "loss": 0.1911,
+      "step": 38860
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.94910415932027e-06,
+      "loss": 0.2475,
+      "step": 38861
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.948768324013903e-06,
+      "loss": 0.2275,
+      "step": 38862
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.948432488707538e-06,
+      "loss": 0.1733,
+      "step": 38863
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.948096653401173e-06,
+      "loss": 0.187,
+      "step": 38864
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9477608180948064e-06,
+      "loss": 0.3399,
+      "step": 38865
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.947424982788441e-06,
+      "loss": 0.3793,
+      "step": 38866
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.947089147482076e-06,
+      "loss": 0.2485,
+      "step": 38867
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9467533121757095e-06,
+      "loss": 0.3783,
+      "step": 38868
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.946417476869343e-06,
+      "loss": 0.2086,
+      "step": 38869
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.946081641562978e-06,
+      "loss": 0.1615,
+      "step": 38870
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.945745806256612e-06,
+      "loss": 0.3591,
+      "step": 38871
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9454099709502464e-06,
+      "loss": 0.1922,
+      "step": 38872
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.945074135643881e-06,
+      "loss": 0.2657,
+      "step": 38873
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.944738300337515e-06,
+      "loss": 0.1823,
+      "step": 38874
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9444024650311495e-06,
+      "loss": 0.2423,
+      "step": 38875
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.944066629724784e-06,
+      "loss": 0.3168,
+      "step": 38876
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.943730794418417e-06,
+      "loss": 0.2689,
+      "step": 38877
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.943394959112052e-06,
+      "loss": 0.3605,
+      "step": 38878
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9430591238056864e-06,
+      "loss": 0.317,
+      "step": 38879
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.94272328849932e-06,
+      "loss": 0.2669,
+      "step": 38880
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.942387453192955e-06,
+      "loss": 0.3215,
+      "step": 38881
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9420516178865895e-06,
+      "loss": 0.2585,
+      "step": 38882
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9417157825802225e-06,
+      "loss": 0.4137,
+      "step": 38883
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.941379947273857e-06,
+      "loss": 0.6961,
+      "step": 38884
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.941044111967492e-06,
+      "loss": 0.2311,
+      "step": 38885
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9407082766611256e-06,
+      "loss": 0.1808,
+      "step": 38886
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.94037244135476e-06,
+      "loss": 0.5389,
+      "step": 38887
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.940036606048395e-06,
+      "loss": 0.424,
+      "step": 38888
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.939700770742029e-06,
+      "loss": 0.2413,
+      "step": 38889
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.939364935435663e-06,
+      "loss": 0.244,
+      "step": 38890
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.939029100129297e-06,
+      "loss": 0.4362,
+      "step": 38891
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.938693264822931e-06,
+      "loss": 0.2404,
+      "step": 38892
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9383574295165656e-06,
+      "loss": 0.3619,
+      "step": 38893
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9380215942102e-06,
+      "loss": 0.4446,
+      "step": 38894
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.937685758903834e-06,
+      "loss": 0.4798,
+      "step": 38895
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.937349923597469e-06,
+      "loss": 0.2068,
+      "step": 38896
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.937014088291103e-06,
+      "loss": 0.3484,
+      "step": 38897
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.936678252984736e-06,
+      "loss": 0.2743,
+      "step": 38898
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.936342417678371e-06,
+      "loss": 0.2863,
+      "step": 38899
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9360065823720056e-06,
+      "loss": 0.158,
+      "step": 38900
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.935670747065639e-06,
+      "loss": 0.2304,
+      "step": 38901
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.935334911759274e-06,
+      "loss": 0.2649,
+      "step": 38902
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.934999076452909e-06,
+      "loss": 0.3577,
+      "step": 38903
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.934663241146542e-06,
+      "loss": 0.3937,
+      "step": 38904
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.934327405840176e-06,
+      "loss": 0.332,
+      "step": 38905
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.933991570533811e-06,
+      "loss": 0.1087,
+      "step": 38906
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.933655735227445e-06,
+      "loss": 0.2899,
+      "step": 38907
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.933319899921079e-06,
+      "loss": 0.2529,
+      "step": 38908
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.932984064614714e-06,
+      "loss": 0.1573,
+      "step": 38909
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.932648229308348e-06,
+      "loss": 0.1612,
+      "step": 38910
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9323123940019824e-06,
+      "loss": 0.2882,
+      "step": 38911
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.931976558695617e-06,
+      "loss": 0.2553,
+      "step": 38912
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.93164072338925e-06,
+      "loss": 0.1862,
+      "step": 38913
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.931304888082885e-06,
+      "loss": 0.2503,
+      "step": 38914
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.930969052776519e-06,
+      "loss": 0.3101,
+      "step": 38915
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.930633217470153e-06,
+      "loss": 0.2612,
+      "step": 38916
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.930297382163788e-06,
+      "loss": 0.3177,
+      "step": 38917
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9299615468574224e-06,
+      "loss": 0.2053,
+      "step": 38918
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.929625711551055e-06,
+      "loss": 0.3356,
+      "step": 38919
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.92928987624469e-06,
+      "loss": 0.3095,
+      "step": 38920
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.928954040938325e-06,
+      "loss": 0.3894,
+      "step": 38921
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9286182056319585e-06,
+      "loss": 0.1578,
+      "step": 38922
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.928282370325593e-06,
+      "loss": 0.2134,
+      "step": 38923
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.927946535019228e-06,
+      "loss": 0.2303,
+      "step": 38924
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9276106997128616e-06,
+      "loss": 0.2457,
+      "step": 38925
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.927274864406495e-06,
+      "loss": 0.2234,
+      "step": 38926
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.92693902910013e-06,
+      "loss": 0.2788,
+      "step": 38927
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.926603193793764e-06,
+      "loss": 0.2989,
+      "step": 38928
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9262673584873985e-06,
+      "loss": 0.192,
+      "step": 38929
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.925931523181033e-06,
+      "loss": 0.264,
+      "step": 38930
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.925595687874667e-06,
+      "loss": 0.1335,
+      "step": 38931
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9252598525683016e-06,
+      "loss": 0.4212,
+      "step": 38932
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.924924017261936e-06,
+      "loss": 0.2509,
+      "step": 38933
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.924588181955569e-06,
+      "loss": 0.1922,
+      "step": 38934
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.924252346649204e-06,
+      "loss": 0.2863,
+      "step": 38935
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.923916511342838e-06,
+      "loss": 0.304,
+      "step": 38936
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.923580676036472e-06,
+      "loss": 0.1508,
+      "step": 38937
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.923244840730107e-06,
+      "loss": 0.1688,
+      "step": 38938
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.92290900542374e-06,
+      "loss": 0.3195,
+      "step": 38939
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9225731701173745e-06,
+      "loss": 0.2715,
+      "step": 38940
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.922237334811009e-06,
+      "loss": 0.3172,
+      "step": 38941
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.921901499504643e-06,
+      "loss": 0.1897,
+      "step": 38942
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.921565664198278e-06,
+      "loss": 0.1564,
+      "step": 38943
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.921229828891912e-06,
+      "loss": 0.2434,
+      "step": 38944
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.920893993585546e-06,
+      "loss": 0.3867,
+      "step": 38945
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.920558158279181e-06,
+      "loss": 0.3312,
+      "step": 38946
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.920222322972815e-06,
+      "loss": 0.1683,
+      "step": 38947
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.919886487666448e-06,
+      "loss": 0.3781,
+      "step": 38948
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.919550652360083e-06,
+      "loss": 0.4364,
+      "step": 38949
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.919214817053718e-06,
+      "loss": 0.2554,
+      "step": 38950
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.918878981747351e-06,
+      "loss": 0.3752,
+      "step": 38951
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.918543146440986e-06,
+      "loss": 0.336,
+      "step": 38952
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.918207311134621e-06,
+      "loss": 0.3104,
+      "step": 38953
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.917871475828254e-06,
+      "loss": 0.4231,
+      "step": 38954
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.917535640521888e-06,
+      "loss": 0.2114,
+      "step": 38955
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.917199805215523e-06,
+      "loss": 0.2682,
+      "step": 38956
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.916863969909157e-06,
+      "loss": 0.3709,
+      "step": 38957
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.916528134602791e-06,
+      "loss": 0.1732,
+      "step": 38958
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.916192299296426e-06,
+      "loss": 0.3039,
+      "step": 38959
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.915856463990059e-06,
+      "loss": 0.2128,
+      "step": 38960
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.915520628683694e-06,
+      "loss": 0.2703,
+      "step": 38961
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.915184793377328e-06,
+      "loss": 0.1786,
+      "step": 38962
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.914848958070962e-06,
+      "loss": 0.4013,
+      "step": 38963
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.914513122764597e-06,
+      "loss": 0.4192,
+      "step": 38964
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.914177287458231e-06,
+      "loss": 0.1597,
+      "step": 38965
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.913841452151865e-06,
+      "loss": 0.3621,
+      "step": 38966
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9135056168455e-06,
+      "loss": 0.252,
+      "step": 38967
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9131697815391345e-06,
+      "loss": 0.3309,
+      "step": 38968
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.912833946232767e-06,
+      "loss": 0.3176,
+      "step": 38969
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.912498110926402e-06,
+      "loss": 0.33,
+      "step": 38970
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.912162275620037e-06,
+      "loss": 0.2892,
+      "step": 38971
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9118264403136705e-06,
+      "loss": 0.3577,
+      "step": 38972
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.911490605007305e-06,
+      "loss": 0.4461,
+      "step": 38973
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.91115476970094e-06,
+      "loss": 0.3841,
+      "step": 38974
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.910818934394573e-06,
+      "loss": 0.2308,
+      "step": 38975
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.910483099088207e-06,
+      "loss": 0.228,
+      "step": 38976
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.910147263781842e-06,
+      "loss": 0.2231,
+      "step": 38977
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.909811428475476e-06,
+      "loss": 0.1612,
+      "step": 38978
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9094755931691105e-06,
+      "loss": 0.2024,
+      "step": 38979
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.909139757862745e-06,
+      "loss": 0.4335,
+      "step": 38980
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.908803922556379e-06,
+      "loss": 0.2757,
+      "step": 38981
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.908468087250013e-06,
+      "loss": 0.3854,
+      "step": 38982
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.908132251943647e-06,
+      "loss": 0.2996,
+      "step": 38983
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.907796416637281e-06,
+      "loss": 0.4006,
+      "step": 38984
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.907460581330916e-06,
+      "loss": 0.2335,
+      "step": 38985
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9071247460245505e-06,
+      "loss": 0.3366,
+      "step": 38986
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.906788910718184e-06,
+      "loss": 0.3348,
+      "step": 38987
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.906453075411819e-06,
+      "loss": 0.1741,
+      "step": 38988
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.906117240105454e-06,
+      "loss": 0.4308,
+      "step": 38989
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9057814047990865e-06,
+      "loss": 0.2868,
+      "step": 38990
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.905445569492721e-06,
+      "loss": 0.3411,
+      "step": 38991
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.905109734186356e-06,
+      "loss": 0.3103,
+      "step": 38992
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.90477389887999e-06,
+      "loss": 0.2064,
+      "step": 38993
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.904438063573624e-06,
+      "loss": 0.3231,
+      "step": 38994
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.904102228267259e-06,
+      "loss": 0.2696,
+      "step": 38995
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.903766392960892e-06,
+      "loss": 0.3601,
+      "step": 38996
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9034305576545265e-06,
+      "loss": 0.1897,
+      "step": 38997
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.903094722348161e-06,
+      "loss": 0.279,
+      "step": 38998
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.902758887041795e-06,
+      "loss": 0.3275,
+      "step": 38999
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.90242305173543e-06,
+      "loss": 0.349,
+      "step": 39000
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.902087216429064e-06,
+      "loss": 0.3449,
+      "step": 39001
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.901751381122698e-06,
+      "loss": 0.2448,
+      "step": 39002
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.901415545816333e-06,
+      "loss": 0.2923,
+      "step": 39003
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9010797105099665e-06,
+      "loss": 0.2019,
+      "step": 39004
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.9007438752036e-06,
+      "loss": 0.2254,
+      "step": 39005
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.900408039897235e-06,
+      "loss": 0.2965,
+      "step": 39006
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 6.90007220459087e-06,
+      "loss": 0.2804,
+      "step": 39007
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8997363692845034e-06,
+      "loss": 0.2967,
+      "step": 39008
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.899400533978138e-06,
+      "loss": 0.3358,
+      "step": 39009
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.899064698671773e-06,
+      "loss": 0.3279,
+      "step": 39010
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.898728863365406e-06,
+      "loss": 0.2946,
+      "step": 39011
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.89839302805904e-06,
+      "loss": 0.313,
+      "step": 39012
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.898057192752675e-06,
+      "loss": 0.263,
+      "step": 39013
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.897721357446309e-06,
+      "loss": 0.346,
+      "step": 39014
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.897385522139943e-06,
+      "loss": 0.3753,
+      "step": 39015
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.897049686833578e-06,
+      "loss": 0.4603,
+      "step": 39016
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.896713851527211e-06,
+      "loss": 0.3665,
+      "step": 39017
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.896378016220846e-06,
+      "loss": 0.4105,
+      "step": 39018
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.89604218091448e-06,
+      "loss": 0.1464,
+      "step": 39019
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.895706345608114e-06,
+      "loss": 0.3082,
+      "step": 39020
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.895370510301749e-06,
+      "loss": 0.3314,
+      "step": 39021
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.895034674995383e-06,
+      "loss": 0.3004,
+      "step": 39022
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.894698839689017e-06,
+      "loss": 0.1935,
+      "step": 39023
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.894363004382652e-06,
+      "loss": 0.2567,
+      "step": 39024
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8940271690762865e-06,
+      "loss": 0.3245,
+      "step": 39025
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8936913337699195e-06,
+      "loss": 0.2202,
+      "step": 39026
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.893355498463554e-06,
+      "loss": 0.3002,
+      "step": 39027
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.893019663157188e-06,
+      "loss": 0.2953,
+      "step": 39028
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8926838278508225e-06,
+      "loss": 0.2627,
+      "step": 39029
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.892347992544457e-06,
+      "loss": 0.193,
+      "step": 39030
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.89201215723809e-06,
+      "loss": 0.2641,
+      "step": 39031
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.891676321931725e-06,
+      "loss": 0.3051,
+      "step": 39032
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8913404866253594e-06,
+      "loss": 0.3091,
+      "step": 39033
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.891004651318993e-06,
+      "loss": 0.2519,
+      "step": 39034
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.890668816012628e-06,
+      "loss": 0.2688,
+      "step": 39035
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8903329807062625e-06,
+      "loss": 0.1804,
+      "step": 39036
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.889997145399896e-06,
+      "loss": 0.2599,
+      "step": 39037
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.889661310093531e-06,
+      "loss": 0.2266,
+      "step": 39038
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.889325474787165e-06,
+      "loss": 0.266,
+      "step": 39039
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.888989639480799e-06,
+      "loss": 0.3426,
+      "step": 39040
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.888653804174433e-06,
+      "loss": 0.3901,
+      "step": 39041
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.888317968868068e-06,
+      "loss": 0.2562,
+      "step": 39042
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.887982133561702e-06,
+      "loss": 0.3439,
+      "step": 39043
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.887646298255336e-06,
+      "loss": 0.173,
+      "step": 39044
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.887310462948971e-06,
+      "loss": 0.3323,
+      "step": 39045
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.886974627642604e-06,
+      "loss": 0.3127,
+      "step": 39046
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.886638792336239e-06,
+      "loss": 0.3401,
+      "step": 39047
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.886302957029873e-06,
+      "loss": 0.4708,
+      "step": 39048
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.885967121723507e-06,
+      "loss": 0.482,
+      "step": 39049
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.885631286417142e-06,
+      "loss": 0.3564,
+      "step": 39050
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.885295451110776e-06,
+      "loss": 0.3298,
+      "step": 39051
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.884959615804409e-06,
+      "loss": 0.3157,
+      "step": 39052
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.884623780498044e-06,
+      "loss": 0.4704,
+      "step": 39053
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8842879451916786e-06,
+      "loss": 0.3568,
+      "step": 39054
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.883952109885312e-06,
+      "loss": 0.2334,
+      "step": 39055
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.883616274578947e-06,
+      "loss": 0.257,
+      "step": 39056
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.883280439272582e-06,
+      "loss": 0.1778,
+      "step": 39057
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8829446039662155e-06,
+      "loss": 0.1762,
+      "step": 39058
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.88260876865985e-06,
+      "loss": 0.4063,
+      "step": 39059
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.882272933353485e-06,
+      "loss": 0.1267,
+      "step": 39060
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.881937098047118e-06,
+      "loss": 0.3913,
+      "step": 39061
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.881601262740752e-06,
+      "loss": 0.3272,
+      "step": 39062
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.881265427434387e-06,
+      "loss": 0.3991,
+      "step": 39063
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.880929592128021e-06,
+      "loss": 0.4766,
+      "step": 39064
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8805937568216555e-06,
+      "loss": 0.2705,
+      "step": 39065
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.88025792151529e-06,
+      "loss": 0.2846,
+      "step": 39066
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.879922086208923e-06,
+      "loss": 0.2661,
+      "step": 39067
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.879586250902558e-06,
+      "loss": 0.2863,
+      "step": 39068
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.879250415596192e-06,
+      "loss": 0.1922,
+      "step": 39069
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.878914580289826e-06,
+      "loss": 0.3656,
+      "step": 39070
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.878578744983461e-06,
+      "loss": 0.3853,
+      "step": 39071
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8782429096770954e-06,
+      "loss": 0.3219,
+      "step": 39072
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.877907074370728e-06,
+      "loss": 0.385,
+      "step": 39073
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.877571239064363e-06,
+      "loss": 0.1915,
+      "step": 39074
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.877235403757998e-06,
+      "loss": 0.4952,
+      "step": 39075
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8768995684516315e-06,
+      "loss": 0.2728,
+      "step": 39076
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.876563733145266e-06,
+      "loss": 0.2657,
+      "step": 39077
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.876227897838901e-06,
+      "loss": 0.2228,
+      "step": 39078
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.875892062532535e-06,
+      "loss": 0.2368,
+      "step": 39079
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.875556227226169e-06,
+      "loss": 0.4991,
+      "step": 39080
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.875220391919804e-06,
+      "loss": 0.3848,
+      "step": 39081
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.874884556613437e-06,
+      "loss": 0.268,
+      "step": 39082
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8745487213070715e-06,
+      "loss": 0.4483,
+      "step": 39083
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.874212886000706e-06,
+      "loss": 0.2794,
+      "step": 39084
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.87387705069434e-06,
+      "loss": 0.353,
+      "step": 39085
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.873541215387975e-06,
+      "loss": 0.2527,
+      "step": 39086
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.873205380081609e-06,
+      "loss": 0.4032,
+      "step": 39087
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.872869544775242e-06,
+      "loss": 0.184,
+      "step": 39088
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.872533709468877e-06,
+      "loss": 0.2052,
+      "step": 39089
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8721978741625115e-06,
+      "loss": 0.2785,
+      "step": 39090
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.871862038856145e-06,
+      "loss": 0.1983,
+      "step": 39091
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.87152620354978e-06,
+      "loss": 0.334,
+      "step": 39092
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8711903682434146e-06,
+      "loss": 0.3081,
+      "step": 39093
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.870854532937048e-06,
+      "loss": 0.168,
+      "step": 39094
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.870518697630682e-06,
+      "loss": 0.2877,
+      "step": 39095
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.870182862324317e-06,
+      "loss": 0.3446,
+      "step": 39096
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.869847027017951e-06,
+      "loss": 0.2669,
+      "step": 39097
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.869511191711585e-06,
+      "loss": 0.3438,
+      "step": 39098
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.86917535640522e-06,
+      "loss": 0.4349,
+      "step": 39099
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.868839521098854e-06,
+      "loss": 0.2656,
+      "step": 39100
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.868503685792488e-06,
+      "loss": 0.2409,
+      "step": 39101
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.868167850486123e-06,
+      "loss": 0.2235,
+      "step": 39102
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.867832015179756e-06,
+      "loss": 0.4056,
+      "step": 39103
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.867496179873391e-06,
+      "loss": 0.3544,
+      "step": 39104
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.867160344567025e-06,
+      "loss": 0.1906,
+      "step": 39105
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.866824509260659e-06,
+      "loss": 0.3236,
+      "step": 39106
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.866488673954294e-06,
+      "loss": 0.1515,
+      "step": 39107
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.866152838647928e-06,
+      "loss": 0.2618,
+      "step": 39108
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.865817003341561e-06,
+      "loss": 0.2608,
+      "step": 39109
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.865481168035196e-06,
+      "loss": 0.6126,
+      "step": 39110
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.865145332728831e-06,
+      "loss": 0.297,
+      "step": 39111
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.864809497422464e-06,
+      "loss": 0.1543,
+      "step": 39112
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.864473662116099e-06,
+      "loss": 0.2709,
+      "step": 39113
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.864137826809734e-06,
+      "loss": 0.25,
+      "step": 39114
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8638019915033675e-06,
+      "loss": 0.3002,
+      "step": 39115
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.863466156197002e-06,
+      "loss": 0.2981,
+      "step": 39116
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.863130320890635e-06,
+      "loss": 0.2003,
+      "step": 39117
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.86279448558427e-06,
+      "loss": 0.1939,
+      "step": 39118
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.862458650277904e-06,
+      "loss": 0.2422,
+      "step": 39119
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.862122814971538e-06,
+      "loss": 0.2394,
+      "step": 39120
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.861786979665173e-06,
+      "loss": 0.3172,
+      "step": 39121
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8614511443588075e-06,
+      "loss": 0.3109,
+      "step": 39122
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8611153090524404e-06,
+      "loss": 0.2093,
+      "step": 39123
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.860779473746075e-06,
+      "loss": 0.3717,
+      "step": 39124
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.86044363843971e-06,
+      "loss": 0.2496,
+      "step": 39125
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8601078031333435e-06,
+      "loss": 0.1777,
+      "step": 39126
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.859771967826978e-06,
+      "loss": 0.2685,
+      "step": 39127
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.859436132520613e-06,
+      "loss": 0.3612,
+      "step": 39128
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.859100297214246e-06,
+      "loss": 0.3481,
+      "step": 39129
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8587644619078804e-06,
+      "loss": 0.3317,
+      "step": 39130
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.858428626601515e-06,
+      "loss": 0.183,
+      "step": 39131
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.858092791295149e-06,
+      "loss": 0.3372,
+      "step": 39132
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8577569559887835e-06,
+      "loss": 0.3641,
+      "step": 39133
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.857421120682418e-06,
+      "loss": 0.2358,
+      "step": 39134
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.857085285376052e-06,
+      "loss": 0.2163,
+      "step": 39135
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.856749450069687e-06,
+      "loss": 0.2619,
+      "step": 39136
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.856413614763321e-06,
+      "loss": 0.3118,
+      "step": 39137
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.856077779456954e-06,
+      "loss": 0.2937,
+      "step": 39138
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.855741944150589e-06,
+      "loss": 0.1637,
+      "step": 39139
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8554061088442235e-06,
+      "loss": 0.284,
+      "step": 39140
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.855070273537857e-06,
+      "loss": 0.2496,
+      "step": 39141
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.854734438231492e-06,
+      "loss": 0.2531,
+      "step": 39142
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.854398602925127e-06,
+      "loss": 0.2787,
+      "step": 39143
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8540627676187596e-06,
+      "loss": 0.3665,
+      "step": 39144
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.853726932312394e-06,
+      "loss": 0.206,
+      "step": 39145
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.853391097006029e-06,
+      "loss": 0.2294,
+      "step": 39146
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.853055261699663e-06,
+      "loss": 0.2756,
+      "step": 39147
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.852719426393297e-06,
+      "loss": 0.2342,
+      "step": 39148
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.852383591086932e-06,
+      "loss": 0.2902,
+      "step": 39149
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.852047755780566e-06,
+      "loss": 0.4383,
+      "step": 39150
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8517119204741996e-06,
+      "loss": 0.3671,
+      "step": 39151
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.851376085167834e-06,
+      "loss": 0.3629,
+      "step": 39152
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.851040249861468e-06,
+      "loss": 0.3313,
+      "step": 39153
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.850704414555103e-06,
+      "loss": 0.2764,
+      "step": 39154
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.850368579248737e-06,
+      "loss": 0.3168,
+      "step": 39155
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.850032743942371e-06,
+      "loss": 0.4047,
+      "step": 39156
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.849696908636006e-06,
+      "loss": 0.2342,
+      "step": 39157
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.84936107332964e-06,
+      "loss": 0.3295,
+      "step": 39158
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.849025238023273e-06,
+      "loss": 0.2852,
+      "step": 39159
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.848689402716908e-06,
+      "loss": 0.3927,
+      "step": 39160
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.848353567410543e-06,
+      "loss": 0.2303,
+      "step": 39161
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8480177321041764e-06,
+      "loss": 0.1561,
+      "step": 39162
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.847681896797811e-06,
+      "loss": 0.4125,
+      "step": 39163
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.847346061491446e-06,
+      "loss": 0.3088,
+      "step": 39164
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.847010226185079e-06,
+      "loss": 0.2269,
+      "step": 39165
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.846674390878713e-06,
+      "loss": 0.2415,
+      "step": 39166
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.846338555572348e-06,
+      "loss": 0.3459,
+      "step": 39167
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.846002720265982e-06,
+      "loss": 0.3669,
+      "step": 39168
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8456668849596164e-06,
+      "loss": 0.1877,
+      "step": 39169
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.845331049653251e-06,
+      "loss": 0.3489,
+      "step": 39170
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.844995214346885e-06,
+      "loss": 0.3574,
+      "step": 39171
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8446593790405195e-06,
+      "loss": 0.4126,
+      "step": 39172
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.844323543734154e-06,
+      "loss": 0.3859,
+      "step": 39173
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.843987708427787e-06,
+      "loss": 0.2813,
+      "step": 39174
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.843651873121422e-06,
+      "loss": 0.3213,
+      "step": 39175
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.843316037815056e-06,
+      "loss": 0.1763,
+      "step": 39176
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.84298020250869e-06,
+      "loss": 0.2501,
+      "step": 39177
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.842644367202325e-06,
+      "loss": 0.4118,
+      "step": 39178
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8423085318959595e-06,
+      "loss": 0.1885,
+      "step": 39179
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8419726965895925e-06,
+      "loss": 0.1816,
+      "step": 39180
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.841636861283227e-06,
+      "loss": 0.3299,
+      "step": 39181
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.841301025976862e-06,
+      "loss": 0.3934,
+      "step": 39182
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8409651906704956e-06,
+      "loss": 0.2607,
+      "step": 39183
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.84062935536413e-06,
+      "loss": 0.2616,
+      "step": 39184
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.840293520057765e-06,
+      "loss": 0.2957,
+      "step": 39185
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.839957684751398e-06,
+      "loss": 0.3833,
+      "step": 39186
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8396218494450325e-06,
+      "loss": 0.3626,
+      "step": 39187
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.839286014138667e-06,
+      "loss": 0.2648,
+      "step": 39188
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.838950178832301e-06,
+      "loss": 0.3238,
+      "step": 39189
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8386143435259356e-06,
+      "loss": 0.3017,
+      "step": 39190
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.83827850821957e-06,
+      "loss": 0.2226,
+      "step": 39191
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.837942672913204e-06,
+      "loss": 0.3692,
+      "step": 39192
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.837606837606839e-06,
+      "loss": 0.4276,
+      "step": 39193
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.837271002300473e-06,
+      "loss": 0.2904,
+      "step": 39194
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.836935166994106e-06,
+      "loss": 0.3151,
+      "step": 39195
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.836599331687741e-06,
+      "loss": 0.2304,
+      "step": 39196
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.8362634963813755e-06,
+      "loss": 0.31,
+      "step": 39197
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.835927661075009e-06,
+      "loss": 0.3591,
+      "step": 39198
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.835591825768644e-06,
+      "loss": 0.2524,
+      "step": 39199
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.835255990462279e-06,
+      "loss": 0.265,
+      "step": 39200
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.834920155155912e-06,
+      "loss": 0.3796,
+      "step": 39201
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.834584319849546e-06,
+      "loss": 0.3481,
+      "step": 39202
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.834248484543181e-06,
+      "loss": 0.3128,
+      "step": 39203
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.833912649236815e-06,
+      "loss": 0.2022,
+      "step": 39204
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 6.833576813930449e-06,
+      "loss": 0.3008,
+      "step": 39205
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.833240978624083e-06,
+      "loss": 0.3821,
+      "step": 39206
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.832905143317718e-06,
+      "loss": 0.1903,
+      "step": 39207
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.832569308011352e-06,
+      "loss": 0.2993,
+      "step": 39208
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.832233472704985e-06,
+      "loss": 0.3981,
+      "step": 39209
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.83189763739862e-06,
+      "loss": 0.2003,
+      "step": 39210
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.831561802092255e-06,
+      "loss": 0.3949,
+      "step": 39211
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.8312259667858885e-06,
+      "loss": 0.2577,
+      "step": 39212
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.830890131479523e-06,
+      "loss": 0.2639,
+      "step": 39213
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.830554296173158e-06,
+      "loss": 0.2951,
+      "step": 39214
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.830218460866791e-06,
+      "loss": 0.2974,
+      "step": 39215
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.829882625560425e-06,
+      "loss": 0.2984,
+      "step": 39216
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.82954679025406e-06,
+      "loss": 0.2373,
+      "step": 39217
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.829210954947694e-06,
+      "loss": 0.1905,
+      "step": 39218
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.8288751196413285e-06,
+      "loss": 0.2514,
+      "step": 39219
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.828539284334963e-06,
+      "loss": 0.1778,
+      "step": 39220
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.828203449028596e-06,
+      "loss": 0.2373,
+      "step": 39221
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.827867613722231e-06,
+      "loss": 0.268,
+      "step": 39222
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.827531778415865e-06,
+      "loss": 0.209,
+      "step": 39223
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.827195943109499e-06,
+      "loss": 0.2801,
+      "step": 39224
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.826860107803134e-06,
+      "loss": 0.3258,
+      "step": 39225
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.8265242724967685e-06,
+      "loss": 0.3626,
+      "step": 39226
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.826188437190402e-06,
+      "loss": 0.361,
+      "step": 39227
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.825852601884037e-06,
+      "loss": 0.4621,
+      "step": 39228
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.8255167665776716e-06,
+      "loss": 0.2938,
+      "step": 39229
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.8251809312713045e-06,
+      "loss": 0.2487,
+      "step": 39230
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.824845095964939e-06,
+      "loss": 0.4504,
+      "step": 39231
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.824509260658574e-06,
+      "loss": 0.2783,
+      "step": 39232
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.824173425352208e-06,
+      "loss": 0.2595,
+      "step": 39233
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.823837590045842e-06,
+      "loss": 0.3,
+      "step": 39234
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.823501754739477e-06,
+      "loss": 0.1654,
+      "step": 39235
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.82316591943311e-06,
+      "loss": 0.3499,
+      "step": 39236
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.8228300841267445e-06,
+      "loss": 0.4247,
+      "step": 39237
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.822494248820379e-06,
+      "loss": 0.2757,
+      "step": 39238
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.822158413514013e-06,
+      "loss": 0.1538,
+      "step": 39239
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.821822578207648e-06,
+      "loss": 0.2437,
+      "step": 39240
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.821486742901282e-06,
+      "loss": 0.3666,
+      "step": 39241
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.821150907594915e-06,
+      "loss": 0.359,
+      "step": 39242
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.82081507228855e-06,
+      "loss": 0.5277,
+      "step": 39243
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.8204792369821845e-06,
+      "loss": 0.2629,
+      "step": 39244
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.820143401675818e-06,
+      "loss": 0.2824,
+      "step": 39245
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.819807566369453e-06,
+      "loss": 0.3186,
+      "step": 39246
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.819471731063088e-06,
+      "loss": 0.3569,
+      "step": 39247
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.819135895756721e-06,
+      "loss": 0.1474,
+      "step": 39248
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.818800060450356e-06,
+      "loss": 0.2301,
+      "step": 39249
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.818464225143991e-06,
+      "loss": 0.3697,
+      "step": 39250
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.818128389837624e-06,
+      "loss": 0.2632,
+      "step": 39251
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.817792554531258e-06,
+      "loss": 0.3624,
+      "step": 39252
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.817456719224893e-06,
+      "loss": 0.3647,
+      "step": 39253
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.817120883918527e-06,
+      "loss": 0.2792,
+      "step": 39254
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.816785048612161e-06,
+      "loss": 0.2276,
+      "step": 39255
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.816449213305796e-06,
+      "loss": 0.2439,
+      "step": 39256
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.816113377999429e-06,
+      "loss": 0.3229,
+      "step": 39257
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.815777542693064e-06,
+      "loss": 0.3411,
+      "step": 39258
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.815441707386698e-06,
+      "loss": 0.2065,
+      "step": 39259
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.815105872080332e-06,
+      "loss": 0.5082,
+      "step": 39260
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.814770036773967e-06,
+      "loss": 0.266,
+      "step": 39261
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.814434201467601e-06,
+      "loss": 0.3719,
+      "step": 39262
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.814098366161235e-06,
+      "loss": 0.2351,
+      "step": 39263
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.813762530854869e-06,
+      "loss": 0.3494,
+      "step": 39264
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.813426695548504e-06,
+      "loss": 0.1786,
+      "step": 39265
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.813090860242137e-06,
+      "loss": 0.2818,
+      "step": 39266
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.812755024935772e-06,
+      "loss": 0.3103,
+      "step": 39267
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.812419189629407e-06,
+      "loss": 0.3302,
+      "step": 39268
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.8120833543230405e-06,
+      "loss": 0.2624,
+      "step": 39269
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.811747519016675e-06,
+      "loss": 0.1794,
+      "step": 39270
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.81141168371031e-06,
+      "loss": 0.3522,
+      "step": 39271
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.811075848403943e-06,
+      "loss": 0.2442,
+      "step": 39272
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.810740013097577e-06,
+      "loss": 0.4219,
+      "step": 39273
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.810404177791212e-06,
+      "loss": 0.2693,
+      "step": 39274
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.810068342484846e-06,
+      "loss": 0.3114,
+      "step": 39275
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.8097325071784805e-06,
+      "loss": 0.2971,
+      "step": 39276
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.809396671872115e-06,
+      "loss": 0.3819,
+      "step": 39277
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.809060836565748e-06,
+      "loss": 0.2081,
+      "step": 39278
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.808725001259383e-06,
+      "loss": 0.4062,
+      "step": 39279
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.808389165953017e-06,
+      "loss": 0.2765,
+      "step": 39280
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.808053330646651e-06,
+      "loss": 0.3226,
+      "step": 39281
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.807717495340286e-06,
+      "loss": 0.4764,
+      "step": 39282
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.8073816600339205e-06,
+      "loss": 0.2282,
+      "step": 39283
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.807045824727554e-06,
+      "loss": 0.2765,
+      "step": 39284
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.806709989421189e-06,
+      "loss": 0.2953,
+      "step": 39285
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.806374154114823e-06,
+      "loss": 0.3361,
+      "step": 39286
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.8060383188084565e-06,
+      "loss": 0.3063,
+      "step": 39287
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.805702483502091e-06,
+      "loss": 0.3314,
+      "step": 39288
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.805366648195726e-06,
+      "loss": 0.1741,
+      "step": 39289
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.80503081288936e-06,
+      "loss": 0.3005,
+      "step": 39290
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.804694977582994e-06,
+      "loss": 0.2354,
+      "step": 39291
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.804359142276629e-06,
+      "loss": 0.1396,
+      "step": 39292
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.804023306970262e-06,
+      "loss": 0.2447,
+      "step": 39293
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.8036874716638965e-06,
+      "loss": 0.2566,
+      "step": 39294
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.803351636357531e-06,
+      "loss": 0.2735,
+      "step": 39295
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.803015801051165e-06,
+      "loss": 0.3563,
+      "step": 39296
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.8026799657448e-06,
+      "loss": 0.3024,
+      "step": 39297
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.802344130438433e-06,
+      "loss": 0.1905,
+      "step": 39298
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.802008295132067e-06,
+      "loss": 0.2725,
+      "step": 39299
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.801672459825702e-06,
+      "loss": 0.2547,
+      "step": 39300
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.801336624519336e-06,
+      "loss": 0.326,
+      "step": 39301
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.80100078921297e-06,
+      "loss": 0.2603,
+      "step": 39302
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.800664953906605e-06,
+      "loss": 0.2802,
+      "step": 39303
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.800329118600239e-06,
+      "loss": 0.2336,
+      "step": 39304
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.799993283293873e-06,
+      "loss": 0.3127,
+      "step": 39305
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.799657447987508e-06,
+      "loss": 0.3061,
+      "step": 39306
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.799321612681141e-06,
+      "loss": 0.2613,
+      "step": 39307
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.798985777374776e-06,
+      "loss": 0.3223,
+      "step": 39308
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.79864994206841e-06,
+      "loss": 0.2748,
+      "step": 39309
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.798314106762044e-06,
+      "loss": 0.3179,
+      "step": 39310
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.797978271455679e-06,
+      "loss": 0.3526,
+      "step": 39311
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.797642436149313e-06,
+      "loss": 0.5357,
+      "step": 39312
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.797306600842946e-06,
+      "loss": 0.4663,
+      "step": 39313
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.796970765536581e-06,
+      "loss": 0.2236,
+      "step": 39314
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.796634930230216e-06,
+      "loss": 0.3994,
+      "step": 39315
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.7962990949238495e-06,
+      "loss": 0.3768,
+      "step": 39316
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.795963259617484e-06,
+      "loss": 0.2542,
+      "step": 39317
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.795627424311119e-06,
+      "loss": 0.3649,
+      "step": 39318
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.7952915890047526e-06,
+      "loss": 0.2983,
+      "step": 39319
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.794955753698387e-06,
+      "loss": 0.2315,
+      "step": 39320
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.794619918392021e-06,
+      "loss": 0.4486,
+      "step": 39321
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.794284083085655e-06,
+      "loss": 0.2791,
+      "step": 39322
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.7939482477792894e-06,
+      "loss": 0.2891,
+      "step": 39323
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.793612412472924e-06,
+      "loss": 0.1875,
+      "step": 39324
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.793276577166558e-06,
+      "loss": 0.1873,
+      "step": 39325
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.7929407418601925e-06,
+      "loss": 0.2389,
+      "step": 39326
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.792604906553827e-06,
+      "loss": 0.5243,
+      "step": 39327
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.79226907124746e-06,
+      "loss": 0.3676,
+      "step": 39328
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.791933235941095e-06,
+      "loss": 0.3241,
+      "step": 39329
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.7915974006347294e-06,
+      "loss": 0.3505,
+      "step": 39330
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.791261565328363e-06,
+      "loss": 0.327,
+      "step": 39331
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.790925730021998e-06,
+      "loss": 0.3363,
+      "step": 39332
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.7905898947156325e-06,
+      "loss": 0.1986,
+      "step": 39333
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.7902540594092655e-06,
+      "loss": 0.3523,
+      "step": 39334
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.7899182241029e-06,
+      "loss": 0.1469,
+      "step": 39335
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.789582388796535e-06,
+      "loss": 0.3453,
+      "step": 39336
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.789246553490169e-06,
+      "loss": 0.3539,
+      "step": 39337
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.788910718183803e-06,
+      "loss": 0.3837,
+      "step": 39338
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.788574882877438e-06,
+      "loss": 0.3546,
+      "step": 39339
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.788239047571072e-06,
+      "loss": 0.3834,
+      "step": 39340
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.787903212264706e-06,
+      "loss": 0.378,
+      "step": 39341
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.787567376958341e-06,
+      "loss": 0.3838,
+      "step": 39342
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.787231541651974e-06,
+      "loss": 0.3197,
+      "step": 39343
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.7868957063456086e-06,
+      "loss": 0.2585,
+      "step": 39344
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.786559871039243e-06,
+      "loss": 0.1966,
+      "step": 39345
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.786224035732877e-06,
+      "loss": 0.3926,
+      "step": 39346
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.785888200426512e-06,
+      "loss": 0.3146,
+      "step": 39347
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.785552365120146e-06,
+      "loss": 0.297,
+      "step": 39348
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.785216529813779e-06,
+      "loss": 0.3433,
+      "step": 39349
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.784880694507414e-06,
+      "loss": 0.2382,
+      "step": 39350
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.7845448592010486e-06,
+      "loss": 0.2859,
+      "step": 39351
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.784209023894682e-06,
+      "loss": 0.4033,
+      "step": 39352
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.783873188588317e-06,
+      "loss": 0.3333,
+      "step": 39353
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.783537353281952e-06,
+      "loss": 0.3116,
+      "step": 39354
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.783201517975585e-06,
+      "loss": 0.2756,
+      "step": 39355
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.782865682669219e-06,
+      "loss": 0.3132,
+      "step": 39356
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.782529847362854e-06,
+      "loss": 0.1775,
+      "step": 39357
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.782194012056488e-06,
+      "loss": 0.2497,
+      "step": 39358
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.781858176750122e-06,
+      "loss": 0.1879,
+      "step": 39359
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.781522341443757e-06,
+      "loss": 0.3107,
+      "step": 39360
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.781186506137391e-06,
+      "loss": 0.3083,
+      "step": 39361
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.7808506708310254e-06,
+      "loss": 0.2475,
+      "step": 39362
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.78051483552466e-06,
+      "loss": 0.2743,
+      "step": 39363
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.780179000218293e-06,
+      "loss": 0.2684,
+      "step": 39364
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.779843164911928e-06,
+      "loss": 0.2488,
+      "step": 39365
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.779507329605562e-06,
+      "loss": 0.3856,
+      "step": 39366
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.779171494299196e-06,
+      "loss": 0.2209,
+      "step": 39367
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.778835658992831e-06,
+      "loss": 0.3559,
+      "step": 39368
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.7784998236864654e-06,
+      "loss": 0.2484,
+      "step": 39369
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.778163988380098e-06,
+      "loss": 0.2333,
+      "step": 39370
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.777828153073733e-06,
+      "loss": 0.2457,
+      "step": 39371
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.777492317767368e-06,
+      "loss": 0.2554,
+      "step": 39372
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.7771564824610015e-06,
+      "loss": 0.3617,
+      "step": 39373
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.776820647154636e-06,
+      "loss": 0.3519,
+      "step": 39374
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.776484811848271e-06,
+      "loss": 0.3415,
+      "step": 39375
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.776148976541905e-06,
+      "loss": 0.3205,
+      "step": 39376
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.775813141235538e-06,
+      "loss": 0.2653,
+      "step": 39377
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.775477305929173e-06,
+      "loss": 0.2767,
+      "step": 39378
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.775141470622807e-06,
+      "loss": 0.2905,
+      "step": 39379
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.7748056353164415e-06,
+      "loss": 0.3997,
+      "step": 39380
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.774469800010076e-06,
+      "loss": 0.2495,
+      "step": 39381
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.77413396470371e-06,
+      "loss": 0.2443,
+      "step": 39382
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.7737981293973446e-06,
+      "loss": 0.1747,
+      "step": 39383
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.773462294090979e-06,
+      "loss": 0.2573,
+      "step": 39384
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.773126458784612e-06,
+      "loss": 0.2338,
+      "step": 39385
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.772790623478247e-06,
+      "loss": 0.2525,
+      "step": 39386
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.772454788171881e-06,
+      "loss": 0.3564,
+      "step": 39387
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.772118952865515e-06,
+      "loss": 0.1851,
+      "step": 39388
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.77178311755915e-06,
+      "loss": 0.211,
+      "step": 39389
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.771447282252783e-06,
+      "loss": 0.3977,
+      "step": 39390
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.7711114469464175e-06,
+      "loss": 0.2124,
+      "step": 39391
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.770775611640052e-06,
+      "loss": 0.1837,
+      "step": 39392
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.770439776333686e-06,
+      "loss": 0.2322,
+      "step": 39393
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.770103941027321e-06,
+      "loss": 0.2234,
+      "step": 39394
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.769768105720955e-06,
+      "loss": 0.1945,
+      "step": 39395
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.769432270414589e-06,
+      "loss": 0.1195,
+      "step": 39396
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.769096435108224e-06,
+      "loss": 0.269,
+      "step": 39397
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.768760599801858e-06,
+      "loss": 0.2688,
+      "step": 39398
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.768424764495491e-06,
+      "loss": 0.2438,
+      "step": 39399
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.768088929189126e-06,
+      "loss": 0.235,
+      "step": 39400
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.767753093882761e-06,
+      "loss": 0.3288,
+      "step": 39401
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.767417258576394e-06,
+      "loss": 0.2336,
+      "step": 39402
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.767081423270029e-06,
+      "loss": 0.3511,
+      "step": 39403
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 6.766745587963664e-06,
+      "loss": 0.4101,
+      "step": 39404
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.766409752657297e-06,
+      "loss": 0.2367,
+      "step": 39405
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.766073917350931e-06,
+      "loss": 0.3304,
+      "step": 39406
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.765738082044566e-06,
+      "loss": 0.3429,
+      "step": 39407
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7654022467382e-06,
+      "loss": 0.2343,
+      "step": 39408
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.765066411431834e-06,
+      "loss": 0.255,
+      "step": 39409
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.764730576125469e-06,
+      "loss": 0.2997,
+      "step": 39410
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.764394740819102e-06,
+      "loss": 0.2833,
+      "step": 39411
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.764058905512737e-06,
+      "loss": 0.4009,
+      "step": 39412
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.763723070206371e-06,
+      "loss": 0.4872,
+      "step": 39413
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.763387234900005e-06,
+      "loss": 0.3355,
+      "step": 39414
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.76305139959364e-06,
+      "loss": 0.2255,
+      "step": 39415
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.762715564287274e-06,
+      "loss": 0.3373,
+      "step": 39416
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.762379728980908e-06,
+      "loss": 0.2003,
+      "step": 39417
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.762043893674543e-06,
+      "loss": 0.2922,
+      "step": 39418
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7617080583681775e-06,
+      "loss": 0.3525,
+      "step": 39419
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7613722230618104e-06,
+      "loss": 0.1955,
+      "step": 39420
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.761036387755445e-06,
+      "loss": 0.3608,
+      "step": 39421
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.76070055244908e-06,
+      "loss": 0.4526,
+      "step": 39422
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7603647171427135e-06,
+      "loss": 0.4012,
+      "step": 39423
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.760028881836348e-06,
+      "loss": 0.2903,
+      "step": 39424
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.759693046529983e-06,
+      "loss": 0.2547,
+      "step": 39425
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.759357211223616e-06,
+      "loss": 0.1725,
+      "step": 39426
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7590213759172504e-06,
+      "loss": 0.3206,
+      "step": 39427
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.758685540610885e-06,
+      "loss": 0.4391,
+      "step": 39428
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.758349705304519e-06,
+      "loss": 0.2196,
+      "step": 39429
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7580138699981535e-06,
+      "loss": 0.3533,
+      "step": 39430
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.757678034691788e-06,
+      "loss": 0.2206,
+      "step": 39431
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.757342199385422e-06,
+      "loss": 0.2295,
+      "step": 39432
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.757006364079056e-06,
+      "loss": 0.341,
+      "step": 39433
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.75667052877269e-06,
+      "loss": 0.2087,
+      "step": 39434
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.756334693466324e-06,
+      "loss": 0.4759,
+      "step": 39435
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.755998858159959e-06,
+      "loss": 0.289,
+      "step": 39436
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7556630228535935e-06,
+      "loss": 0.3958,
+      "step": 39437
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.755327187547227e-06,
+      "loss": 0.2494,
+      "step": 39438
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.754991352240862e-06,
+      "loss": 0.2585,
+      "step": 39439
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.754655516934497e-06,
+      "loss": 0.1515,
+      "step": 39440
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7543196816281296e-06,
+      "loss": 0.397,
+      "step": 39441
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.753983846321764e-06,
+      "loss": 0.2595,
+      "step": 39442
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.753648011015399e-06,
+      "loss": 0.3211,
+      "step": 39443
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.753312175709033e-06,
+      "loss": 0.5018,
+      "step": 39444
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.752976340402667e-06,
+      "loss": 0.356,
+      "step": 39445
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.752640505096302e-06,
+      "loss": 0.2824,
+      "step": 39446
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.752304669789935e-06,
+      "loss": 0.181,
+      "step": 39447
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7519688344835695e-06,
+      "loss": 0.267,
+      "step": 39448
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.751632999177204e-06,
+      "loss": 0.4243,
+      "step": 39449
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.751297163870838e-06,
+      "loss": 0.3199,
+      "step": 39450
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.750961328564473e-06,
+      "loss": 0.1774,
+      "step": 39451
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.750625493258107e-06,
+      "loss": 0.26,
+      "step": 39452
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.750289657951741e-06,
+      "loss": 0.2742,
+      "step": 39453
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.749953822645376e-06,
+      "loss": 0.2884,
+      "step": 39454
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.74961798733901e-06,
+      "loss": 0.2953,
+      "step": 39455
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.749282152032643e-06,
+      "loss": 0.4122,
+      "step": 39456
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.748946316726278e-06,
+      "loss": 0.2778,
+      "step": 39457
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.748610481419913e-06,
+      "loss": 0.1834,
+      "step": 39458
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7482746461135464e-06,
+      "loss": 0.2364,
+      "step": 39459
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.747938810807181e-06,
+      "loss": 0.3128,
+      "step": 39460
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.747602975500816e-06,
+      "loss": 0.3139,
+      "step": 39461
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.747267140194449e-06,
+      "loss": 0.1814,
+      "step": 39462
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.746931304888083e-06,
+      "loss": 0.2379,
+      "step": 39463
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.746595469581718e-06,
+      "loss": 0.5019,
+      "step": 39464
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.746259634275352e-06,
+      "loss": 0.399,
+      "step": 39465
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7459237989689864e-06,
+      "loss": 0.2144,
+      "step": 39466
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.745587963662621e-06,
+      "loss": 0.2254,
+      "step": 39467
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.745252128356254e-06,
+      "loss": 0.2485,
+      "step": 39468
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.744916293049889e-06,
+      "loss": 0.2709,
+      "step": 39469
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.744580457743523e-06,
+      "loss": 0.4093,
+      "step": 39470
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.744244622437157e-06,
+      "loss": 0.3822,
+      "step": 39471
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.743908787130792e-06,
+      "loss": 0.3656,
+      "step": 39472
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.743572951824426e-06,
+      "loss": 0.2426,
+      "step": 39473
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.74323711651806e-06,
+      "loss": 0.3747,
+      "step": 39474
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.742901281211695e-06,
+      "loss": 0.284,
+      "step": 39475
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7425654459053295e-06,
+      "loss": 0.5603,
+      "step": 39476
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7422296105989625e-06,
+      "loss": 0.3328,
+      "step": 39477
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.741893775292597e-06,
+      "loss": 0.3011,
+      "step": 39478
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.741557939986231e-06,
+      "loss": 0.3217,
+      "step": 39479
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7412221046798656e-06,
+      "loss": 0.2048,
+      "step": 39480
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7408862693735e-06,
+      "loss": 0.3377,
+      "step": 39481
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.740550434067133e-06,
+      "loss": 0.3385,
+      "step": 39482
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.740214598760768e-06,
+      "loss": 0.1207,
+      "step": 39483
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7398787634544025e-06,
+      "loss": 0.2902,
+      "step": 39484
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.739542928148036e-06,
+      "loss": 0.2739,
+      "step": 39485
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.739207092841671e-06,
+      "loss": 0.218,
+      "step": 39486
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7388712575353056e-06,
+      "loss": 0.3504,
+      "step": 39487
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.738535422228939e-06,
+      "loss": 0.2082,
+      "step": 39488
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.738199586922574e-06,
+      "loss": 0.3532,
+      "step": 39489
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.737863751616208e-06,
+      "loss": 0.254,
+      "step": 39490
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.737527916309842e-06,
+      "loss": 0.2347,
+      "step": 39491
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.737192081003476e-06,
+      "loss": 0.126,
+      "step": 39492
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.736856245697111e-06,
+      "loss": 0.2311,
+      "step": 39493
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.736520410390745e-06,
+      "loss": 0.3944,
+      "step": 39494
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.736184575084379e-06,
+      "loss": 0.306,
+      "step": 39495
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.735848739778014e-06,
+      "loss": 0.3553,
+      "step": 39496
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.735512904471647e-06,
+      "loss": 0.2496,
+      "step": 39497
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.735177069165282e-06,
+      "loss": 0.3419,
+      "step": 39498
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.734841233858916e-06,
+      "loss": 0.2176,
+      "step": 39499
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.73450539855255e-06,
+      "loss": 0.1763,
+      "step": 39500
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.734169563246185e-06,
+      "loss": 0.2497,
+      "step": 39501
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.733833727939819e-06,
+      "loss": 0.38,
+      "step": 39502
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.733497892633452e-06,
+      "loss": 0.3592,
+      "step": 39503
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.733162057327087e-06,
+      "loss": 0.3084,
+      "step": 39504
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.732826222020722e-06,
+      "loss": 0.2504,
+      "step": 39505
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.732490386714355e-06,
+      "loss": 0.5231,
+      "step": 39506
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.73215455140799e-06,
+      "loss": 0.3409,
+      "step": 39507
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.731818716101625e-06,
+      "loss": 0.1844,
+      "step": 39508
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7314828807952585e-06,
+      "loss": 0.2701,
+      "step": 39509
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.731147045488893e-06,
+      "loss": 0.394,
+      "step": 39510
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.730811210182528e-06,
+      "loss": 0.286,
+      "step": 39511
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.730475374876161e-06,
+      "loss": 0.2609,
+      "step": 39512
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.730139539569795e-06,
+      "loss": 0.4095,
+      "step": 39513
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.72980370426343e-06,
+      "loss": 0.2267,
+      "step": 39514
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.729467868957064e-06,
+      "loss": 0.3097,
+      "step": 39515
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7291320336506985e-06,
+      "loss": 0.2782,
+      "step": 39516
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.728796198344333e-06,
+      "loss": 0.3398,
+      "step": 39517
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.728460363037966e-06,
+      "loss": 0.3013,
+      "step": 39518
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.728124527731601e-06,
+      "loss": 0.3411,
+      "step": 39519
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.727788692425235e-06,
+      "loss": 0.3264,
+      "step": 39520
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.727452857118869e-06,
+      "loss": 0.2655,
+      "step": 39521
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.727117021812504e-06,
+      "loss": 0.145,
+      "step": 39522
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7267811865061385e-06,
+      "loss": 0.2658,
+      "step": 39523
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.726445351199771e-06,
+      "loss": 0.2993,
+      "step": 39524
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.726109515893406e-06,
+      "loss": 0.1642,
+      "step": 39525
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.725773680587041e-06,
+      "loss": 0.2387,
+      "step": 39526
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7254378452806745e-06,
+      "loss": 0.295,
+      "step": 39527
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.725102009974309e-06,
+      "loss": 0.2724,
+      "step": 39528
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.724766174667944e-06,
+      "loss": 0.3252,
+      "step": 39529
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.724430339361578e-06,
+      "loss": 0.3457,
+      "step": 39530
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.724094504055212e-06,
+      "loss": 0.3285,
+      "step": 39531
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.723758668748847e-06,
+      "loss": 0.3999,
+      "step": 39532
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.72342283344248e-06,
+      "loss": 0.2246,
+      "step": 39533
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7230869981361145e-06,
+      "loss": 0.3075,
+      "step": 39534
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.722751162829749e-06,
+      "loss": 0.3755,
+      "step": 39535
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.722415327523383e-06,
+      "loss": 0.2871,
+      "step": 39536
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.722079492217018e-06,
+      "loss": 0.3454,
+      "step": 39537
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.721743656910652e-06,
+      "loss": 0.231,
+      "step": 39538
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.721407821604285e-06,
+      "loss": 0.375,
+      "step": 39539
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.72107198629792e-06,
+      "loss": 0.3196,
+      "step": 39540
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7207361509915545e-06,
+      "loss": 0.2943,
+      "step": 39541
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.720400315685188e-06,
+      "loss": 0.4115,
+      "step": 39542
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.720064480378823e-06,
+      "loss": 0.2381,
+      "step": 39543
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.719728645072458e-06,
+      "loss": 0.2678,
+      "step": 39544
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.719392809766091e-06,
+      "loss": 0.2827,
+      "step": 39545
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.719056974459725e-06,
+      "loss": 0.3155,
+      "step": 39546
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.71872113915336e-06,
+      "loss": 0.3458,
+      "step": 39547
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.718385303846994e-06,
+      "loss": 0.2673,
+      "step": 39548
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.718049468540628e-06,
+      "loss": 0.3675,
+      "step": 39549
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.717713633234263e-06,
+      "loss": 0.4274,
+      "step": 39550
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.717377797927897e-06,
+      "loss": 0.262,
+      "step": 39551
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.717041962621531e-06,
+      "loss": 0.2436,
+      "step": 39552
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.716706127315166e-06,
+      "loss": 0.3743,
+      "step": 39553
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.716370292008799e-06,
+      "loss": 0.2301,
+      "step": 39554
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.716034456702434e-06,
+      "loss": 0.3972,
+      "step": 39555
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.715698621396068e-06,
+      "loss": 0.2059,
+      "step": 39556
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.715362786089702e-06,
+      "loss": 0.2268,
+      "step": 39557
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.715026950783337e-06,
+      "loss": 0.3045,
+      "step": 39558
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.714691115476971e-06,
+      "loss": 0.4794,
+      "step": 39559
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.714355280170604e-06,
+      "loss": 0.2575,
+      "step": 39560
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.714019444864239e-06,
+      "loss": 0.2632,
+      "step": 39561
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.713683609557874e-06,
+      "loss": 0.3539,
+      "step": 39562
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.713347774251507e-06,
+      "loss": 0.3069,
+      "step": 39563
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.713011938945142e-06,
+      "loss": 0.2483,
+      "step": 39564
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.712676103638777e-06,
+      "loss": 0.3739,
+      "step": 39565
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7123402683324105e-06,
+      "loss": 0.2155,
+      "step": 39566
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.712004433026045e-06,
+      "loss": 0.3244,
+      "step": 39567
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.711668597719678e-06,
+      "loss": 0.3272,
+      "step": 39568
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.711332762413313e-06,
+      "loss": 0.3314,
+      "step": 39569
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.710996927106947e-06,
+      "loss": 0.2666,
+      "step": 39570
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.710661091800581e-06,
+      "loss": 0.2455,
+      "step": 39571
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.710325256494216e-06,
+      "loss": 0.2684,
+      "step": 39572
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7099894211878505e-06,
+      "loss": 0.1924,
+      "step": 39573
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7096535858814835e-06,
+      "loss": 0.4142,
+      "step": 39574
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.709317750575118e-06,
+      "loss": 0.2112,
+      "step": 39575
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.708981915268753e-06,
+      "loss": 0.3429,
+      "step": 39576
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7086460799623865e-06,
+      "loss": 0.1648,
+      "step": 39577
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.708310244656021e-06,
+      "loss": 0.2516,
+      "step": 39578
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.707974409349656e-06,
+      "loss": 0.2573,
+      "step": 39579
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.70763857404329e-06,
+      "loss": 0.225,
+      "step": 39580
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7073027387369234e-06,
+      "loss": 0.2772,
+      "step": 39581
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.706966903430558e-06,
+      "loss": 0.3528,
+      "step": 39582
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.706631068124192e-06,
+      "loss": 0.159,
+      "step": 39583
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7062952328178265e-06,
+      "loss": 0.4286,
+      "step": 39584
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.705959397511461e-06,
+      "loss": 0.3054,
+      "step": 39585
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.705623562205095e-06,
+      "loss": 0.2237,
+      "step": 39586
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.70528772689873e-06,
+      "loss": 0.2067,
+      "step": 39587
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.704951891592364e-06,
+      "loss": 0.2423,
+      "step": 39588
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.704616056285997e-06,
+      "loss": 0.2384,
+      "step": 39589
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.704280220979632e-06,
+      "loss": 0.3459,
+      "step": 39590
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7039443856732665e-06,
+      "loss": 0.57,
+      "step": 39591
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7036085503669e-06,
+      "loss": 0.2466,
+      "step": 39592
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.703272715060535e-06,
+      "loss": 0.3737,
+      "step": 39593
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.70293687975417e-06,
+      "loss": 0.4012,
+      "step": 39594
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.7026010444478026e-06,
+      "loss": 0.2635,
+      "step": 39595
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.702265209141437e-06,
+      "loss": 0.2842,
+      "step": 39596
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.701929373835072e-06,
+      "loss": 0.2988,
+      "step": 39597
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.701593538528706e-06,
+      "loss": 0.2734,
+      "step": 39598
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.70125770322234e-06,
+      "loss": 0.2215,
+      "step": 39599
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.700921867915975e-06,
+      "loss": 0.3917,
+      "step": 39600
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.700586032609609e-06,
+      "loss": 0.2518,
+      "step": 39601
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 6.700250197303243e-06,
+      "loss": 0.3051,
+      "step": 39602
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.699914361996877e-06,
+      "loss": 0.3892,
+      "step": 39603
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.699578526690511e-06,
+      "loss": 0.274,
+      "step": 39604
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.699242691384146e-06,
+      "loss": 0.3364,
+      "step": 39605
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.69890685607778e-06,
+      "loss": 0.2551,
+      "step": 39606
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.698571020771414e-06,
+      "loss": 0.2921,
+      "step": 39607
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.698235185465049e-06,
+      "loss": 0.4294,
+      "step": 39608
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.697899350158683e-06,
+      "loss": 0.4223,
+      "step": 39609
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.697563514852316e-06,
+      "loss": 0.4177,
+      "step": 39610
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.697227679545951e-06,
+      "loss": 0.173,
+      "step": 39611
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.696891844239586e-06,
+      "loss": 0.3685,
+      "step": 39612
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6965560089332195e-06,
+      "loss": 0.2326,
+      "step": 39613
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.696220173626854e-06,
+      "loss": 0.2298,
+      "step": 39614
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.695884338320489e-06,
+      "loss": 0.4599,
+      "step": 39615
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.695548503014122e-06,
+      "loss": 0.3393,
+      "step": 39616
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.695212667707756e-06,
+      "loss": 0.2934,
+      "step": 39617
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.694876832401391e-06,
+      "loss": 0.4617,
+      "step": 39618
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.694540997095025e-06,
+      "loss": 0.37,
+      "step": 39619
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6942051617886594e-06,
+      "loss": 0.3327,
+      "step": 39620
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.693869326482294e-06,
+      "loss": 0.3418,
+      "step": 39621
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.693533491175928e-06,
+      "loss": 0.2023,
+      "step": 39622
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6931976558695625e-06,
+      "loss": 0.3843,
+      "step": 39623
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.692861820563197e-06,
+      "loss": 0.139,
+      "step": 39624
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.69252598525683e-06,
+      "loss": 0.478,
+      "step": 39625
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.692190149950465e-06,
+      "loss": 0.29,
+      "step": 39626
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6918543146440994e-06,
+      "loss": 0.2974,
+      "step": 39627
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.691518479337733e-06,
+      "loss": 0.4563,
+      "step": 39628
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.691182644031368e-06,
+      "loss": 0.3287,
+      "step": 39629
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6908468087250025e-06,
+      "loss": 0.3094,
+      "step": 39630
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6905109734186355e-06,
+      "loss": 0.3001,
+      "step": 39631
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.69017513811227e-06,
+      "loss": 0.2274,
+      "step": 39632
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.689839302805905e-06,
+      "loss": 0.2404,
+      "step": 39633
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.689503467499539e-06,
+      "loss": 0.3157,
+      "step": 39634
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.689167632193173e-06,
+      "loss": 0.3056,
+      "step": 39635
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.688831796886808e-06,
+      "loss": 0.3103,
+      "step": 39636
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.688495961580441e-06,
+      "loss": 0.4309,
+      "step": 39637
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6881601262740755e-06,
+      "loss": 0.4081,
+      "step": 39638
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.68782429096771e-06,
+      "loss": 0.2159,
+      "step": 39639
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.687488455661344e-06,
+      "loss": 0.2327,
+      "step": 39640
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6871526203549786e-06,
+      "loss": 0.1489,
+      "step": 39641
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.686816785048613e-06,
+      "loss": 0.272,
+      "step": 39642
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.686480949742247e-06,
+      "loss": 0.1543,
+      "step": 39643
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.686145114435882e-06,
+      "loss": 0.2447,
+      "step": 39644
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.685809279129516e-06,
+      "loss": 0.4257,
+      "step": 39645
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.685473443823149e-06,
+      "loss": 0.2172,
+      "step": 39646
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.685137608516784e-06,
+      "loss": 0.3029,
+      "step": 39647
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6848017732104186e-06,
+      "loss": 0.2419,
+      "step": 39648
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.684465937904052e-06,
+      "loss": 0.3098,
+      "step": 39649
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.684130102597687e-06,
+      "loss": 0.3108,
+      "step": 39650
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.683794267291322e-06,
+      "loss": 0.2972,
+      "step": 39651
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.683458431984955e-06,
+      "loss": 0.2386,
+      "step": 39652
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.683122596678589e-06,
+      "loss": 0.4549,
+      "step": 39653
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.682786761372224e-06,
+      "loss": 0.2394,
+      "step": 39654
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.682450926065858e-06,
+      "loss": 0.1913,
+      "step": 39655
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.682115090759492e-06,
+      "loss": 0.268,
+      "step": 39656
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.681779255453126e-06,
+      "loss": 0.2718,
+      "step": 39657
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.681443420146761e-06,
+      "loss": 0.3588,
+      "step": 39658
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.681107584840395e-06,
+      "loss": 0.3441,
+      "step": 39659
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.680771749534028e-06,
+      "loss": 0.4021,
+      "step": 39660
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.680435914227663e-06,
+      "loss": 0.2982,
+      "step": 39661
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.680100078921298e-06,
+      "loss": 0.3001,
+      "step": 39662
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6797642436149315e-06,
+      "loss": 0.277,
+      "step": 39663
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.679428408308566e-06,
+      "loss": 0.3475,
+      "step": 39664
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.679092573002201e-06,
+      "loss": 0.2608,
+      "step": 39665
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.678756737695834e-06,
+      "loss": 0.2071,
+      "step": 39666
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.678420902389468e-06,
+      "loss": 0.3536,
+      "step": 39667
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.678085067083103e-06,
+      "loss": 0.2533,
+      "step": 39668
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.677749231776737e-06,
+      "loss": 0.2393,
+      "step": 39669
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6774133964703715e-06,
+      "loss": 0.2008,
+      "step": 39670
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.677077561164006e-06,
+      "loss": 0.1712,
+      "step": 39671
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.676741725857639e-06,
+      "loss": 0.1658,
+      "step": 39672
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.676405890551274e-06,
+      "loss": 0.3665,
+      "step": 39673
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.676070055244908e-06,
+      "loss": 0.2426,
+      "step": 39674
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.675734219938542e-06,
+      "loss": 0.1507,
+      "step": 39675
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.675398384632177e-06,
+      "loss": 0.2457,
+      "step": 39676
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6750625493258115e-06,
+      "loss": 0.3282,
+      "step": 39677
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.674726714019445e-06,
+      "loss": 0.2477,
+      "step": 39678
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.67439087871308e-06,
+      "loss": 0.2585,
+      "step": 39679
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6740550434067146e-06,
+      "loss": 0.3086,
+      "step": 39680
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6737192081003475e-06,
+      "loss": 0.2638,
+      "step": 39681
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.673383372793982e-06,
+      "loss": 0.2778,
+      "step": 39682
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.673047537487617e-06,
+      "loss": 0.1385,
+      "step": 39683
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.672711702181251e-06,
+      "loss": 0.3111,
+      "step": 39684
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.672375866874885e-06,
+      "loss": 0.2824,
+      "step": 39685
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.67204003156852e-06,
+      "loss": 0.159,
+      "step": 39686
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.671704196262153e-06,
+      "loss": 0.2466,
+      "step": 39687
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6713683609557875e-06,
+      "loss": 0.1445,
+      "step": 39688
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.671032525649422e-06,
+      "loss": 0.3519,
+      "step": 39689
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.670696690343056e-06,
+      "loss": 0.3602,
+      "step": 39690
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.670360855036691e-06,
+      "loss": 0.2147,
+      "step": 39691
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.670025019730325e-06,
+      "loss": 0.3416,
+      "step": 39692
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.669689184423958e-06,
+      "loss": 0.2437,
+      "step": 39693
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.669353349117593e-06,
+      "loss": 0.1859,
+      "step": 39694
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6690175138112275e-06,
+      "loss": 0.3238,
+      "step": 39695
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.668681678504861e-06,
+      "loss": 0.2013,
+      "step": 39696
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.668345843198496e-06,
+      "loss": 0.4093,
+      "step": 39697
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.668010007892131e-06,
+      "loss": 0.2563,
+      "step": 39698
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.667674172585764e-06,
+      "loss": 0.4885,
+      "step": 39699
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.667338337279399e-06,
+      "loss": 0.2157,
+      "step": 39700
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.667002501973034e-06,
+      "loss": 0.5044,
+      "step": 39701
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 0.2837,
+      "step": 39702
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.666330831360301e-06,
+      "loss": 0.2233,
+      "step": 39703
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.665994996053936e-06,
+      "loss": 0.2747,
+      "step": 39704
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.66565916074757e-06,
+      "loss": 0.2305,
+      "step": 39705
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.665323325441204e-06,
+      "loss": 0.2847,
+      "step": 39706
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.664987490134839e-06,
+      "loss": 0.3086,
+      "step": 39707
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.664651654828472e-06,
+      "loss": 0.1317,
+      "step": 39708
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.664315819522107e-06,
+      "loss": 0.1597,
+      "step": 39709
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.663979984215741e-06,
+      "loss": 0.1629,
+      "step": 39710
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.663644148909375e-06,
+      "loss": 0.2905,
+      "step": 39711
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.66330831360301e-06,
+      "loss": 0.192,
+      "step": 39712
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.662972478296644e-06,
+      "loss": 0.5546,
+      "step": 39713
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.662636642990278e-06,
+      "loss": 0.3703,
+      "step": 39714
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.662300807683913e-06,
+      "loss": 0.2732,
+      "step": 39715
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.661964972377547e-06,
+      "loss": 0.2307,
+      "step": 39716
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6616291370711804e-06,
+      "loss": 0.2589,
+      "step": 39717
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.661293301764815e-06,
+      "loss": 0.1604,
+      "step": 39718
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.66095746645845e-06,
+      "loss": 0.3612,
+      "step": 39719
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6606216311520835e-06,
+      "loss": 0.3866,
+      "step": 39720
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.660285795845718e-06,
+      "loss": 0.2236,
+      "step": 39721
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.659949960539353e-06,
+      "loss": 0.3572,
+      "step": 39722
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.659614125232986e-06,
+      "loss": 0.2289,
+      "step": 39723
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.65927828992662e-06,
+      "loss": 0.3646,
+      "step": 39724
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.658942454620255e-06,
+      "loss": 0.2342,
+      "step": 39725
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.658606619313889e-06,
+      "loss": 0.1703,
+      "step": 39726
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6582707840075235e-06,
+      "loss": 0.2808,
+      "step": 39727
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.657934948701158e-06,
+      "loss": 0.33,
+      "step": 39728
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.657599113394791e-06,
+      "loss": 0.2939,
+      "step": 39729
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.657263278088426e-06,
+      "loss": 0.4007,
+      "step": 39730
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.65692744278206e-06,
+      "loss": 0.2327,
+      "step": 39731
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.656591607475694e-06,
+      "loss": 0.1827,
+      "step": 39732
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.656255772169329e-06,
+      "loss": 0.2089,
+      "step": 39733
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6559199368629635e-06,
+      "loss": 0.24,
+      "step": 39734
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.655584101556597e-06,
+      "loss": 0.2112,
+      "step": 39735
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.655248266250232e-06,
+      "loss": 0.1991,
+      "step": 39736
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.654912430943867e-06,
+      "loss": 0.1548,
+      "step": 39737
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6545765956374996e-06,
+      "loss": 0.2137,
+      "step": 39738
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.654240760331134e-06,
+      "loss": 0.3287,
+      "step": 39739
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.653904925024769e-06,
+      "loss": 0.3305,
+      "step": 39740
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.653569089718403e-06,
+      "loss": 0.1903,
+      "step": 39741
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.653233254412037e-06,
+      "loss": 0.1797,
+      "step": 39742
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.652897419105672e-06,
+      "loss": 0.2475,
+      "step": 39743
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.652561583799305e-06,
+      "loss": 0.3877,
+      "step": 39744
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6522257484929395e-06,
+      "loss": 0.2533,
+      "step": 39745
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.651889913186574e-06,
+      "loss": 0.2891,
+      "step": 39746
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.651554077880208e-06,
+      "loss": 0.3666,
+      "step": 39747
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.651218242573843e-06,
+      "loss": 0.182,
+      "step": 39748
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6508824072674764e-06,
+      "loss": 0.2141,
+      "step": 39749
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.65054657196111e-06,
+      "loss": 0.2249,
+      "step": 39750
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.650210736654745e-06,
+      "loss": 0.3262,
+      "step": 39751
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.649874901348379e-06,
+      "loss": 0.2634,
+      "step": 39752
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.649539066042013e-06,
+      "loss": 0.3317,
+      "step": 39753
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.649203230735648e-06,
+      "loss": 0.1135,
+      "step": 39754
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.648867395429282e-06,
+      "loss": 0.238,
+      "step": 39755
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6485315601229164e-06,
+      "loss": 0.1677,
+      "step": 39756
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.648195724816551e-06,
+      "loss": 0.3324,
+      "step": 39757
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.647859889510184e-06,
+      "loss": 0.2341,
+      "step": 39758
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.647524054203819e-06,
+      "loss": 0.258,
+      "step": 39759
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.647188218897453e-06,
+      "loss": 0.177,
+      "step": 39760
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.646852383591087e-06,
+      "loss": 0.45,
+      "step": 39761
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.646516548284722e-06,
+      "loss": 0.247,
+      "step": 39762
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.646180712978356e-06,
+      "loss": 0.3596,
+      "step": 39763
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.645844877671989e-06,
+      "loss": 0.1826,
+      "step": 39764
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.645509042365624e-06,
+      "loss": 0.3017,
+      "step": 39765
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.645173207059259e-06,
+      "loss": 0.255,
+      "step": 39766
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6448373717528925e-06,
+      "loss": 0.206,
+      "step": 39767
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.644501536446527e-06,
+      "loss": 0.1791,
+      "step": 39768
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.644165701140162e-06,
+      "loss": 0.1451,
+      "step": 39769
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6438298658337956e-06,
+      "loss": 0.3563,
+      "step": 39770
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.64349403052743e-06,
+      "loss": 0.2518,
+      "step": 39771
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.643158195221064e-06,
+      "loss": 0.2596,
+      "step": 39772
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.642822359914698e-06,
+      "loss": 0.277,
+      "step": 39773
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6424865246083325e-06,
+      "loss": 0.3065,
+      "step": 39774
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.642150689301967e-06,
+      "loss": 0.2413,
+      "step": 39775
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.641814853995601e-06,
+      "loss": 0.2148,
+      "step": 39776
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6414790186892356e-06,
+      "loss": 0.3036,
+      "step": 39777
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.64114318338287e-06,
+      "loss": 0.3735,
+      "step": 39778
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.640807348076503e-06,
+      "loss": 0.3583,
+      "step": 39779
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.640471512770138e-06,
+      "loss": 0.2476,
+      "step": 39780
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6401356774637725e-06,
+      "loss": 0.3098,
+      "step": 39781
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.639799842157406e-06,
+      "loss": 0.4894,
+      "step": 39782
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.639464006851041e-06,
+      "loss": 0.2159,
+      "step": 39783
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6391281715446755e-06,
+      "loss": 0.5658,
+      "step": 39784
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.6387923362383085e-06,
+      "loss": 0.2757,
+      "step": 39785
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.638456500931943e-06,
+      "loss": 0.3764,
+      "step": 39786
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.638120665625578e-06,
+      "loss": 0.4092,
+      "step": 39787
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.637784830319212e-06,
+      "loss": 0.2626,
+      "step": 39788
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.637448995012846e-06,
+      "loss": 0.1329,
+      "step": 39789
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.637113159706481e-06,
+      "loss": 0.2938,
+      "step": 39790
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.636777324400115e-06,
+      "loss": 0.1832,
+      "step": 39791
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.636441489093749e-06,
+      "loss": 0.3207,
+      "step": 39792
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.636105653787384e-06,
+      "loss": 0.3011,
+      "step": 39793
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.635769818481017e-06,
+      "loss": 0.2858,
+      "step": 39794
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.635433983174652e-06,
+      "loss": 0.1703,
+      "step": 39795
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.635098147868286e-06,
+      "loss": 0.2959,
+      "step": 39796
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.63476231256192e-06,
+      "loss": 0.3079,
+      "step": 39797
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.634426477255555e-06,
+      "loss": 0.2604,
+      "step": 39798
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.634090641949189e-06,
+      "loss": 0.3686,
+      "step": 39799
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.633754806642822e-06,
+      "loss": 0.1905,
+      "step": 39800
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 6.633418971336457e-06,
+      "loss": 0.1921,
+      "step": 39801
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.6330831360300916e-06,
+      "loss": 0.2721,
+      "step": 39802
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.632747300723725e-06,
+      "loss": 0.2534,
+      "step": 39803
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.63241146541736e-06,
+      "loss": 0.1694,
+      "step": 39804
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.632075630110995e-06,
+      "loss": 0.294,
+      "step": 39805
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.631739794804628e-06,
+      "loss": 0.3023,
+      "step": 39806
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.631403959498262e-06,
+      "loss": 0.3012,
+      "step": 39807
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.631068124191897e-06,
+      "loss": 0.1999,
+      "step": 39808
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.630732288885531e-06,
+      "loss": 0.3931,
+      "step": 39809
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.630396453579165e-06,
+      "loss": 0.3176,
+      "step": 39810
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.6300606182728e-06,
+      "loss": 0.3375,
+      "step": 39811
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.629724782966434e-06,
+      "loss": 0.2898,
+      "step": 39812
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.6293889476600685e-06,
+      "loss": 0.2532,
+      "step": 39813
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.629053112353703e-06,
+      "loss": 0.1603,
+      "step": 39814
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.628717277047336e-06,
+      "loss": 0.2217,
+      "step": 39815
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.628381441740971e-06,
+      "loss": 0.2002,
+      "step": 39816
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.628045606434605e-06,
+      "loss": 0.2742,
+      "step": 39817
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.627709771128239e-06,
+      "loss": 0.3967,
+      "step": 39818
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.627373935821874e-06,
+      "loss": 0.2431,
+      "step": 39819
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.6270381005155085e-06,
+      "loss": 0.2021,
+      "step": 39820
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.626702265209141e-06,
+      "loss": 0.263,
+      "step": 39821
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.626366429902776e-06,
+      "loss": 0.2033,
+      "step": 39822
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.626030594596411e-06,
+      "loss": 0.1663,
+      "step": 39823
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.6256947592900445e-06,
+      "loss": 0.2317,
+      "step": 39824
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.625358923983679e-06,
+      "loss": 0.3131,
+      "step": 39825
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.625023088677314e-06,
+      "loss": 0.2401,
+      "step": 39826
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.624687253370948e-06,
+      "loss": 0.1623,
+      "step": 39827
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.624351418064581e-06,
+      "loss": 0.3852,
+      "step": 39828
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.624015582758216e-06,
+      "loss": 0.3547,
+      "step": 39829
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.62367974745185e-06,
+      "loss": 0.1473,
+      "step": 39830
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.6233439121454845e-06,
+      "loss": 0.427,
+      "step": 39831
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.623008076839119e-06,
+      "loss": 0.3043,
+      "step": 39832
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.622672241532753e-06,
+      "loss": 0.159,
+      "step": 39833
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.622336406226388e-06,
+      "loss": 0.2027,
+      "step": 39834
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.622000570920022e-06,
+      "loss": 0.2011,
+      "step": 39835
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.621664735613655e-06,
+      "loss": 0.2,
+      "step": 39836
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.62132890030729e-06,
+      "loss": 0.3807,
+      "step": 39837
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.620993065000924e-06,
+      "loss": 0.2886,
+      "step": 39838
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.620657229694558e-06,
+      "loss": 0.1611,
+      "step": 39839
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.620321394388193e-06,
+      "loss": 0.2797,
+      "step": 39840
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.619985559081826e-06,
+      "loss": 0.1798,
+      "step": 39841
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.6196497237754605e-06,
+      "loss": 0.2013,
+      "step": 39842
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.619313888469095e-06,
+      "loss": 0.2908,
+      "step": 39843
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.618978053162729e-06,
+      "loss": 0.2534,
+      "step": 39844
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.618642217856364e-06,
+      "loss": 0.161,
+      "step": 39845
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.618306382549998e-06,
+      "loss": 0.2248,
+      "step": 39846
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.617970547243632e-06,
+      "loss": 0.2206,
+      "step": 39847
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.617634711937267e-06,
+      "loss": 0.2302,
+      "step": 39848
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.617298876630901e-06,
+      "loss": 0.1788,
+      "step": 39849
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.616963041324534e-06,
+      "loss": 0.2745,
+      "step": 39850
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.616627206018169e-06,
+      "loss": 0.2477,
+      "step": 39851
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.616291370711804e-06,
+      "loss": 0.1838,
+      "step": 39852
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.615955535405437e-06,
+      "loss": 0.4669,
+      "step": 39853
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.615619700099072e-06,
+      "loss": 0.2563,
+      "step": 39854
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.615283864792707e-06,
+      "loss": 0.2121,
+      "step": 39855
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.61494802948634e-06,
+      "loss": 0.2446,
+      "step": 39856
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.614612194179974e-06,
+      "loss": 0.1944,
+      "step": 39857
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.614276358873609e-06,
+      "loss": 0.2309,
+      "step": 39858
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.613940523567243e-06,
+      "loss": 0.2173,
+      "step": 39859
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.613604688260877e-06,
+      "loss": 0.2738,
+      "step": 39860
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.613268852954512e-06,
+      "loss": 0.2937,
+      "step": 39861
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.612933017648146e-06,
+      "loss": 0.3021,
+      "step": 39862
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.61259718234178e-06,
+      "loss": 0.3448,
+      "step": 39863
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.612261347035414e-06,
+      "loss": 0.1522,
+      "step": 39864
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.611925511729048e-06,
+      "loss": 0.3069,
+      "step": 39865
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.611589676422683e-06,
+      "loss": 0.4316,
+      "step": 39866
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.611253841116317e-06,
+      "loss": 0.2106,
+      "step": 39867
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.610918005809951e-06,
+      "loss": 0.4145,
+      "step": 39868
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.610582170503586e-06,
+      "loss": 0.2948,
+      "step": 39869
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.6102463351972205e-06,
+      "loss": 0.4244,
+      "step": 39870
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.6099104998908534e-06,
+      "loss": 0.4105,
+      "step": 39871
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.609574664584488e-06,
+      "loss": 0.1761,
+      "step": 39872
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.609238829278123e-06,
+      "loss": 0.3871,
+      "step": 39873
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.6089029939717565e-06,
+      "loss": 0.3355,
+      "step": 39874
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.608567158665391e-06,
+      "loss": 0.217,
+      "step": 39875
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.608231323359026e-06,
+      "loss": 0.2332,
+      "step": 39876
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.607895488052659e-06,
+      "loss": 0.3656,
+      "step": 39877
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.6075596527462934e-06,
+      "loss": 0.3059,
+      "step": 39878
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.607223817439928e-06,
+      "loss": 0.3053,
+      "step": 39879
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.606887982133562e-06,
+      "loss": 0.3464,
+      "step": 39880
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.6065521468271965e-06,
+      "loss": 0.4834,
+      "step": 39881
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.606216311520831e-06,
+      "loss": 0.1974,
+      "step": 39882
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.605880476214465e-06,
+      "loss": 0.2499,
+      "step": 39883
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.6055446409081e-06,
+      "loss": 0.2882,
+      "step": 39884
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.6052088056017334e-06,
+      "loss": 0.3001,
+      "step": 39885
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.604872970295367e-06,
+      "loss": 0.2054,
+      "step": 39886
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.604537134989002e-06,
+      "loss": 0.3071,
+      "step": 39887
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.6042012996826365e-06,
+      "loss": 0.2381,
+      "step": 39888
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.60386546437627e-06,
+      "loss": 0.2616,
+      "step": 39889
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.603529629069905e-06,
+      "loss": 0.2289,
+      "step": 39890
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.60319379376354e-06,
+      "loss": 0.264,
+      "step": 39891
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.6028579584571726e-06,
+      "loss": 0.3386,
+      "step": 39892
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.602522123150807e-06,
+      "loss": 0.3635,
+      "step": 39893
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.602186287844442e-06,
+      "loss": 0.3449,
+      "step": 39894
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.601850452538076e-06,
+      "loss": 0.2497,
+      "step": 39895
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.60151461723171e-06,
+      "loss": 0.2828,
+      "step": 39896
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.601178781925345e-06,
+      "loss": 0.1048,
+      "step": 39897
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.600842946618978e-06,
+      "loss": 0.2028,
+      "step": 39898
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.6005071113126126e-06,
+      "loss": 0.3719,
+      "step": 39899
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.600171276006247e-06,
+      "loss": 0.4175,
+      "step": 39900
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.599835440699881e-06,
+      "loss": 0.3535,
+      "step": 39901
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.599499605393516e-06,
+      "loss": 0.3515,
+      "step": 39902
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.59916377008715e-06,
+      "loss": 0.2351,
+      "step": 39903
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.598827934780784e-06,
+      "loss": 0.1443,
+      "step": 39904
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.598492099474419e-06,
+      "loss": 0.4183,
+      "step": 39905
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.598156264168053e-06,
+      "loss": 0.3626,
+      "step": 39906
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.597820428861686e-06,
+      "loss": 0.3468,
+      "step": 39907
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.597484593555321e-06,
+      "loss": 0.1863,
+      "step": 39908
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.597148758248956e-06,
+      "loss": 0.2395,
+      "step": 39909
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.5968129229425894e-06,
+      "loss": 0.305,
+      "step": 39910
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.596477087636224e-06,
+      "loss": 0.1774,
+      "step": 39911
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.596141252329859e-06,
+      "loss": 0.3776,
+      "step": 39912
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.595805417023492e-06,
+      "loss": 0.2542,
+      "step": 39913
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.595469581717126e-06,
+      "loss": 0.3067,
+      "step": 39914
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.595133746410761e-06,
+      "loss": 0.2711,
+      "step": 39915
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.594797911104395e-06,
+      "loss": 0.2986,
+      "step": 39916
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.5944620757980294e-06,
+      "loss": 0.1445,
+      "step": 39917
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.594126240491664e-06,
+      "loss": 0.2403,
+      "step": 39918
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.593790405185297e-06,
+      "loss": 0.346,
+      "step": 39919
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.593454569878932e-06,
+      "loss": 0.2252,
+      "step": 39920
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.593118734572566e-06,
+      "loss": 0.137,
+      "step": 39921
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.5927828992662e-06,
+      "loss": 0.4134,
+      "step": 39922
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.592447063959835e-06,
+      "loss": 0.1955,
+      "step": 39923
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.5921112286534694e-06,
+      "loss": 0.1804,
+      "step": 39924
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.591775393347103e-06,
+      "loss": 0.2736,
+      "step": 39925
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.591439558040738e-06,
+      "loss": 0.1701,
+      "step": 39926
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.5911037227343725e-06,
+      "loss": 0.2991,
+      "step": 39927
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.5907678874280055e-06,
+      "loss": 0.4296,
+      "step": 39928
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.59043205212164e-06,
+      "loss": 0.2859,
+      "step": 39929
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.590096216815274e-06,
+      "loss": 0.2177,
+      "step": 39930
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.5897603815089086e-06,
+      "loss": 0.2267,
+      "step": 39931
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.589424546202543e-06,
+      "loss": 0.3173,
+      "step": 39932
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.589088710896176e-06,
+      "loss": 0.2844,
+      "step": 39933
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.588752875589811e-06,
+      "loss": 0.255,
+      "step": 39934
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.5884170402834455e-06,
+      "loss": 0.2794,
+      "step": 39935
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.588081204977079e-06,
+      "loss": 0.2472,
+      "step": 39936
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.587745369670714e-06,
+      "loss": 0.1628,
+      "step": 39937
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.5874095343643486e-06,
+      "loss": 0.2547,
+      "step": 39938
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.587073699057982e-06,
+      "loss": 0.1203,
+      "step": 39939
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.586737863751617e-06,
+      "loss": 0.2314,
+      "step": 39940
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.586402028445251e-06,
+      "loss": 0.225,
+      "step": 39941
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.586066193138885e-06,
+      "loss": 0.3725,
+      "step": 39942
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.585730357832519e-06,
+      "loss": 0.1657,
+      "step": 39943
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.585394522526154e-06,
+      "loss": 0.2553,
+      "step": 39944
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.585058687219788e-06,
+      "loss": 0.338,
+      "step": 39945
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.584722851913422e-06,
+      "loss": 0.3484,
+      "step": 39946
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.584387016607057e-06,
+      "loss": 0.3176,
+      "step": 39947
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.58405118130069e-06,
+      "loss": 0.312,
+      "step": 39948
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.583715345994325e-06,
+      "loss": 0.4032,
+      "step": 39949
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.583379510687959e-06,
+      "loss": 0.4317,
+      "step": 39950
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.583043675381593e-06,
+      "loss": 0.2563,
+      "step": 39951
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.582707840075228e-06,
+      "loss": 0.2758,
+      "step": 39952
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.582372004768862e-06,
+      "loss": 0.1274,
+      "step": 39953
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.582036169462495e-06,
+      "loss": 0.2693,
+      "step": 39954
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.58170033415613e-06,
+      "loss": 0.2579,
+      "step": 39955
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.581364498849765e-06,
+      "loss": 0.435,
+      "step": 39956
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.581028663543398e-06,
+      "loss": 0.3695,
+      "step": 39957
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.580692828237033e-06,
+      "loss": 0.2807,
+      "step": 39958
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.580356992930668e-06,
+      "loss": 0.3361,
+      "step": 39959
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.5800211576243015e-06,
+      "loss": 0.1649,
+      "step": 39960
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.579685322317936e-06,
+      "loss": 0.2296,
+      "step": 39961
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.579349487011571e-06,
+      "loss": 0.3252,
+      "step": 39962
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.579013651705204e-06,
+      "loss": 0.248,
+      "step": 39963
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.578677816398838e-06,
+      "loss": 0.4027,
+      "step": 39964
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.578341981092473e-06,
+      "loss": 0.3021,
+      "step": 39965
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.578006145786107e-06,
+      "loss": 0.3629,
+      "step": 39966
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.5776703104797415e-06,
+      "loss": 0.2917,
+      "step": 39967
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.577334475173376e-06,
+      "loss": 0.2311,
+      "step": 39968
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.576998639867009e-06,
+      "loss": 0.2799,
+      "step": 39969
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.576662804560644e-06,
+      "loss": 0.2525,
+      "step": 39970
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.576326969254278e-06,
+      "loss": 0.3271,
+      "step": 39971
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.575991133947912e-06,
+      "loss": 0.2327,
+      "step": 39972
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.575655298641547e-06,
+      "loss": 0.2742,
+      "step": 39973
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.5753194633351815e-06,
+      "loss": 0.2272,
+      "step": 39974
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.574983628028815e-06,
+      "loss": 0.2532,
+      "step": 39975
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.574647792722449e-06,
+      "loss": 0.1745,
+      "step": 39976
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.574311957416084e-06,
+      "loss": 0.2028,
+      "step": 39977
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.5739761221097175e-06,
+      "loss": 0.1789,
+      "step": 39978
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.573640286803352e-06,
+      "loss": 0.2493,
+      "step": 39979
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.573304451496987e-06,
+      "loss": 0.379,
+      "step": 39980
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.572968616190621e-06,
+      "loss": 0.2253,
+      "step": 39981
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.572632780884255e-06,
+      "loss": 0.4005,
+      "step": 39982
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.57229694557789e-06,
+      "loss": 0.2977,
+      "step": 39983
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.571961110271523e-06,
+      "loss": 0.3928,
+      "step": 39984
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.5716252749651575e-06,
+      "loss": 0.2032,
+      "step": 39985
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.571289439658792e-06,
+      "loss": 0.2642,
+      "step": 39986
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.570953604352426e-06,
+      "loss": 0.2129,
+      "step": 39987
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.570617769046061e-06,
+      "loss": 0.3176,
+      "step": 39988
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.570281933739695e-06,
+      "loss": 0.3494,
+      "step": 39989
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.569946098433328e-06,
+      "loss": 0.1605,
+      "step": 39990
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.569610263126963e-06,
+      "loss": 0.31,
+      "step": 39991
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.5692744278205975e-06,
+      "loss": 0.1743,
+      "step": 39992
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.568938592514231e-06,
+      "loss": 0.3265,
+      "step": 39993
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.568602757207866e-06,
+      "loss": 0.1077,
+      "step": 39994
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.568266921901501e-06,
+      "loss": 0.2825,
+      "step": 39995
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.567931086595134e-06,
+      "loss": 0.2033,
+      "step": 39996
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.567595251288769e-06,
+      "loss": 0.2896,
+      "step": 39997
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.567259415982403e-06,
+      "loss": 0.1363,
+      "step": 39998
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 6.566923580676037e-06,
+      "loss": 0.2638,
+      "step": 39999
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.566587745369671e-06,
+      "loss": 0.2705,
+      "step": 40000
+    },
+    {
+      "epoch": 2.02,
+      "eval_accuracy": 0.8832161564513403,
+      "eval_f1_microaverage": 0.8832161564513403,
+      "eval_loss": 0.3103916645050049,
+      "eval_precision_macroaverage": 0.7537318952544214,
+      "eval_precision_microaverage": 0.8832161564513403,
+      "eval_recall_macroaverage": 0.6910740506304066,
+      "eval_recall_microaverage": 0.8832161564513403,
+      "eval_runtime": 1173.6017,
+      "eval_samples_per_second": 231.968,
+      "eval_steps_per_second": 3.625,
+      "step": 40000
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.566251910063306e-06,
+      "loss": 0.3796,
+      "step": 40001
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.56591607475694e-06,
+      "loss": 0.4793,
+      "step": 40002
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.565580239450574e-06,
+      "loss": 0.3217,
+      "step": 40003
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.565244404144209e-06,
+      "loss": 0.1604,
+      "step": 40004
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.564908568837842e-06,
+      "loss": 0.1087,
+      "step": 40005
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.564572733531477e-06,
+      "loss": 0.2764,
+      "step": 40006
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.564236898225111e-06,
+      "loss": 0.5381,
+      "step": 40007
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.563901062918745e-06,
+      "loss": 0.3162,
+      "step": 40008
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.56356522761238e-06,
+      "loss": 0.3001,
+      "step": 40009
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.563229392306014e-06,
+      "loss": 0.2626,
+      "step": 40010
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.562893556999647e-06,
+      "loss": 0.1698,
+      "step": 40011
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.562557721693282e-06,
+      "loss": 0.4259,
+      "step": 40012
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.562221886386917e-06,
+      "loss": 0.2056,
+      "step": 40013
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5618860510805504e-06,
+      "loss": 0.2156,
+      "step": 40014
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.561550215774185e-06,
+      "loss": 0.2618,
+      "step": 40015
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.56121438046782e-06,
+      "loss": 0.4648,
+      "step": 40016
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5608785451614535e-06,
+      "loss": 0.35,
+      "step": 40017
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.560542709855088e-06,
+      "loss": 0.2471,
+      "step": 40018
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.560206874548721e-06,
+      "loss": 0.2323,
+      "step": 40019
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.559871039242356e-06,
+      "loss": 0.2476,
+      "step": 40020
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.55953520393599e-06,
+      "loss": 0.3651,
+      "step": 40021
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.559199368629624e-06,
+      "loss": 0.2858,
+      "step": 40022
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.558863533323259e-06,
+      "loss": 0.1977,
+      "step": 40023
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5585276980168935e-06,
+      "loss": 0.2761,
+      "step": 40024
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5581918627105265e-06,
+      "loss": 0.2314,
+      "step": 40025
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.557856027404161e-06,
+      "loss": 0.2281,
+      "step": 40026
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.557520192097796e-06,
+      "loss": 0.2358,
+      "step": 40027
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5571843567914296e-06,
+      "loss": 0.2456,
+      "step": 40028
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.556848521485064e-06,
+      "loss": 0.1807,
+      "step": 40029
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.556512686178699e-06,
+      "loss": 0.2631,
+      "step": 40030
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.556176850872333e-06,
+      "loss": 0.1878,
+      "step": 40031
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5558410155659665e-06,
+      "loss": 0.3087,
+      "step": 40032
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.555505180259601e-06,
+      "loss": 0.3516,
+      "step": 40033
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.555169344953235e-06,
+      "loss": 0.3243,
+      "step": 40034
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5548335096468695e-06,
+      "loss": 0.2984,
+      "step": 40035
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.554497674340504e-06,
+      "loss": 0.2651,
+      "step": 40036
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.554161839034138e-06,
+      "loss": 0.2378,
+      "step": 40037
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.553826003727773e-06,
+      "loss": 0.2851,
+      "step": 40038
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.553490168421407e-06,
+      "loss": 0.3138,
+      "step": 40039
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.55315433311504e-06,
+      "loss": 0.2536,
+      "step": 40040
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.552818497808675e-06,
+      "loss": 0.3196,
+      "step": 40041
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5524826625023095e-06,
+      "loss": 0.4007,
+      "step": 40042
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.552146827195943e-06,
+      "loss": 0.3386,
+      "step": 40043
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.551810991889578e-06,
+      "loss": 0.2932,
+      "step": 40044
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.551475156583213e-06,
+      "loss": 0.1833,
+      "step": 40045
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.551139321276846e-06,
+      "loss": 0.3624,
+      "step": 40046
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.55080348597048e-06,
+      "loss": 0.2944,
+      "step": 40047
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.550467650664115e-06,
+      "loss": 0.1749,
+      "step": 40048
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.550131815357749e-06,
+      "loss": 0.3128,
+      "step": 40049
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.549795980051383e-06,
+      "loss": 0.4731,
+      "step": 40050
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.549460144745018e-06,
+      "loss": 0.2054,
+      "step": 40051
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.549124309438652e-06,
+      "loss": 0.2354,
+      "step": 40052
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5487884741322864e-06,
+      "loss": 0.1375,
+      "step": 40053
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.54845263882592e-06,
+      "loss": 0.2132,
+      "step": 40054
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.548116803519554e-06,
+      "loss": 0.285,
+      "step": 40055
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.547780968213189e-06,
+      "loss": 0.1555,
+      "step": 40056
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.547445132906823e-06,
+      "loss": 0.1156,
+      "step": 40057
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.547109297600457e-06,
+      "loss": 0.256,
+      "step": 40058
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.546773462294092e-06,
+      "loss": 0.2067,
+      "step": 40059
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.546437626987726e-06,
+      "loss": 0.2483,
+      "step": 40060
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.546101791681359e-06,
+      "loss": 0.3369,
+      "step": 40061
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.545765956374994e-06,
+      "loss": 0.5269,
+      "step": 40062
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.545430121068629e-06,
+      "loss": 0.2702,
+      "step": 40063
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5450942857622625e-06,
+      "loss": 0.3006,
+      "step": 40064
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.544758450455897e-06,
+      "loss": 0.4423,
+      "step": 40065
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.544422615149532e-06,
+      "loss": 0.1378,
+      "step": 40066
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.544086779843165e-06,
+      "loss": 0.5605,
+      "step": 40067
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.543750944536799e-06,
+      "loss": 0.4433,
+      "step": 40068
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.543415109230434e-06,
+      "loss": 0.2215,
+      "step": 40069
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.543079273924068e-06,
+      "loss": 0.1169,
+      "step": 40070
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5427434386177025e-06,
+      "loss": 0.2767,
+      "step": 40071
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.542407603311337e-06,
+      "loss": 0.3593,
+      "step": 40072
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.542071768004971e-06,
+      "loss": 0.3213,
+      "step": 40073
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5417359326986055e-06,
+      "loss": 0.3064,
+      "step": 40074
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.54140009739224e-06,
+      "loss": 0.2397,
+      "step": 40075
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.541064262085873e-06,
+      "loss": 0.4441,
+      "step": 40076
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.540728426779508e-06,
+      "loss": 0.2815,
+      "step": 40077
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5403925914731424e-06,
+      "loss": 0.2097,
+      "step": 40078
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.540056756166776e-06,
+      "loss": 0.302,
+      "step": 40079
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.539720920860411e-06,
+      "loss": 0.442,
+      "step": 40080
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5393850855540455e-06,
+      "loss": 0.3072,
+      "step": 40081
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5390492502476785e-06,
+      "loss": 0.2547,
+      "step": 40082
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.538713414941313e-06,
+      "loss": 0.3005,
+      "step": 40083
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.538377579634948e-06,
+      "loss": 0.1975,
+      "step": 40084
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.538041744328582e-06,
+      "loss": 0.2063,
+      "step": 40085
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.537705909022216e-06,
+      "loss": 0.2716,
+      "step": 40086
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.537370073715851e-06,
+      "loss": 0.5106,
+      "step": 40087
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.537034238409484e-06,
+      "loss": 0.3099,
+      "step": 40088
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5366984031031185e-06,
+      "loss": 0.2105,
+      "step": 40089
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.536362567796753e-06,
+      "loss": 0.2279,
+      "step": 40090
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.536026732490387e-06,
+      "loss": 0.2564,
+      "step": 40091
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.535690897184022e-06,
+      "loss": 0.2154,
+      "step": 40092
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.535355061877656e-06,
+      "loss": 0.2178,
+      "step": 40093
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.53501922657129e-06,
+      "loss": 0.2333,
+      "step": 40094
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.534683391264925e-06,
+      "loss": 0.2521,
+      "step": 40095
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.534347555958559e-06,
+      "loss": 0.2657,
+      "step": 40096
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.534011720652192e-06,
+      "loss": 0.1177,
+      "step": 40097
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.533675885345827e-06,
+      "loss": 0.2654,
+      "step": 40098
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5333400500394616e-06,
+      "loss": 0.2857,
+      "step": 40099
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.533004214733095e-06,
+      "loss": 0.2593,
+      "step": 40100
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.53266837942673e-06,
+      "loss": 0.2058,
+      "step": 40101
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.532332544120365e-06,
+      "loss": 0.2223,
+      "step": 40102
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.531996708813998e-06,
+      "loss": 0.2465,
+      "step": 40103
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.531660873507632e-06,
+      "loss": 0.2803,
+      "step": 40104
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.531325038201267e-06,
+      "loss": 0.1612,
+      "step": 40105
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.530989202894901e-06,
+      "loss": 0.4519,
+      "step": 40106
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.530653367588535e-06,
+      "loss": 0.3004,
+      "step": 40107
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.53031753228217e-06,
+      "loss": 0.2867,
+      "step": 40108
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.529981696975804e-06,
+      "loss": 0.1288,
+      "step": 40109
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5296458616694385e-06,
+      "loss": 0.3521,
+      "step": 40110
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.529310026363071e-06,
+      "loss": 0.4596,
+      "step": 40111
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.528974191056706e-06,
+      "loss": 0.2447,
+      "step": 40112
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.528638355750341e-06,
+      "loss": 0.3265,
+      "step": 40113
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5283025204439745e-06,
+      "loss": 0.2055,
+      "step": 40114
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.527966685137609e-06,
+      "loss": 0.3452,
+      "step": 40115
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.527630849831244e-06,
+      "loss": 0.372,
+      "step": 40116
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.527295014524877e-06,
+      "loss": 0.2124,
+      "step": 40117
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.526959179218511e-06,
+      "loss": 0.2762,
+      "step": 40118
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.526623343912146e-06,
+      "loss": 0.2587,
+      "step": 40119
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.52628750860578e-06,
+      "loss": 0.3188,
+      "step": 40120
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5259516732994145e-06,
+      "loss": 0.1521,
+      "step": 40121
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.525615837993049e-06,
+      "loss": 0.4999,
+      "step": 40122
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.525280002686682e-06,
+      "loss": 0.3788,
+      "step": 40123
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.524944167380317e-06,
+      "loss": 0.2149,
+      "step": 40124
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.524608332073951e-06,
+      "loss": 0.2491,
+      "step": 40125
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.524272496767585e-06,
+      "loss": 0.2287,
+      "step": 40126
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.52393666146122e-06,
+      "loss": 0.237,
+      "step": 40127
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5236008261548545e-06,
+      "loss": 0.3247,
+      "step": 40128
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.523264990848488e-06,
+      "loss": 0.2584,
+      "step": 40129
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.522929155542123e-06,
+      "loss": 0.1946,
+      "step": 40130
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.522593320235758e-06,
+      "loss": 0.1604,
+      "step": 40131
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5222574849293905e-06,
+      "loss": 0.362,
+      "step": 40132
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.521921649623025e-06,
+      "loss": 0.1651,
+      "step": 40133
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.52158581431666e-06,
+      "loss": 0.2601,
+      "step": 40134
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.521249979010294e-06,
+      "loss": 0.2987,
+      "step": 40135
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.520914143703928e-06,
+      "loss": 0.275,
+      "step": 40136
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.520578308397563e-06,
+      "loss": 0.2481,
+      "step": 40137
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.520242473091196e-06,
+      "loss": 0.2405,
+      "step": 40138
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5199066377848305e-06,
+      "loss": 0.3601,
+      "step": 40139
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.519570802478465e-06,
+      "loss": 0.2412,
+      "step": 40140
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.519234967172099e-06,
+      "loss": 0.222,
+      "step": 40141
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.518899131865734e-06,
+      "loss": 0.1783,
+      "step": 40142
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.518563296559368e-06,
+      "loss": 0.1683,
+      "step": 40143
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.518227461253002e-06,
+      "loss": 0.2311,
+      "step": 40144
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.517891625946636e-06,
+      "loss": 0.1601,
+      "step": 40145
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5175557906402705e-06,
+      "loss": 0.267,
+      "step": 40146
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.517219955333904e-06,
+      "loss": 0.1309,
+      "step": 40147
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.516884120027539e-06,
+      "loss": 0.2069,
+      "step": 40148
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.516548284721174e-06,
+      "loss": 0.4437,
+      "step": 40149
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.516212449414807e-06,
+      "loss": 0.4718,
+      "step": 40150
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.515876614108442e-06,
+      "loss": 0.2821,
+      "step": 40151
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.515540778802077e-06,
+      "loss": 0.2473,
+      "step": 40152
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.51520494349571e-06,
+      "loss": 0.2676,
+      "step": 40153
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.514869108189344e-06,
+      "loss": 0.2591,
+      "step": 40154
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.514533272882979e-06,
+      "loss": 0.1898,
+      "step": 40155
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.514197437576613e-06,
+      "loss": 0.2876,
+      "step": 40156
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.513861602270247e-06,
+      "loss": 0.2138,
+      "step": 40157
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.513525766963882e-06,
+      "loss": 0.3118,
+      "step": 40158
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.513189931657515e-06,
+      "loss": 0.2159,
+      "step": 40159
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.51285409635115e-06,
+      "loss": 0.251,
+      "step": 40160
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.512518261044784e-06,
+      "loss": 0.1417,
+      "step": 40161
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.512182425738418e-06,
+      "loss": 0.1547,
+      "step": 40162
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.511846590432053e-06,
+      "loss": 0.2271,
+      "step": 40163
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.511510755125687e-06,
+      "loss": 0.3169,
+      "step": 40164
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.511174919819321e-06,
+      "loss": 0.2469,
+      "step": 40165
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.510839084512956e-06,
+      "loss": 0.4839,
+      "step": 40166
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.51050324920659e-06,
+      "loss": 0.2503,
+      "step": 40167
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5101674139002234e-06,
+      "loss": 0.2961,
+      "step": 40168
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.509831578593858e-06,
+      "loss": 0.3817,
+      "step": 40169
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.509495743287493e-06,
+      "loss": 0.1544,
+      "step": 40170
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5091599079811265e-06,
+      "loss": 0.4304,
+      "step": 40171
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.508824072674761e-06,
+      "loss": 0.2158,
+      "step": 40172
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.508488237368396e-06,
+      "loss": 0.2253,
+      "step": 40173
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.508152402062029e-06,
+      "loss": 0.3069,
+      "step": 40174
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5078165667556634e-06,
+      "loss": 0.2377,
+      "step": 40175
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.507480731449298e-06,
+      "loss": 0.4033,
+      "step": 40176
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.507144896142932e-06,
+      "loss": 0.4695,
+      "step": 40177
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5068090608365665e-06,
+      "loss": 0.2971,
+      "step": 40178
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.506473225530201e-06,
+      "loss": 0.2007,
+      "step": 40179
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.506137390223834e-06,
+      "loss": 0.2404,
+      "step": 40180
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.505801554917469e-06,
+      "loss": 0.2181,
+      "step": 40181
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.505465719611103e-06,
+      "loss": 0.2771,
+      "step": 40182
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.505129884304737e-06,
+      "loss": 0.2763,
+      "step": 40183
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.504794048998372e-06,
+      "loss": 0.2315,
+      "step": 40184
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5044582136920065e-06,
+      "loss": 0.3785,
+      "step": 40185
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.50412237838564e-06,
+      "loss": 0.4311,
+      "step": 40186
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.503786543079275e-06,
+      "loss": 0.1887,
+      "step": 40187
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.50345070777291e-06,
+      "loss": 0.2762,
+      "step": 40188
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5031148724665426e-06,
+      "loss": 0.2819,
+      "step": 40189
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.502779037160177e-06,
+      "loss": 0.3053,
+      "step": 40190
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.502443201853812e-06,
+      "loss": 0.1744,
+      "step": 40191
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.502107366547446e-06,
+      "loss": 0.4834,
+      "step": 40192
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.50177153124108e-06,
+      "loss": 0.2103,
+      "step": 40193
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.501435695934715e-06,
+      "loss": 0.1807,
+      "step": 40194
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.501099860628348e-06,
+      "loss": 0.3664,
+      "step": 40195
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.5007640253219826e-06,
+      "loss": 0.3146,
+      "step": 40196
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.500428190015617e-06,
+      "loss": 0.2217,
+      "step": 40197
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 6.500092354709251e-06,
+      "loss": 0.2575,
+      "step": 40198
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.499756519402886e-06,
+      "loss": 0.2282,
+      "step": 40199
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4994206840965195e-06,
+      "loss": 0.2741,
+      "step": 40200
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.499084848790153e-06,
+      "loss": 0.3582,
+      "step": 40201
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.498749013483788e-06,
+      "loss": 0.209,
+      "step": 40202
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.498413178177422e-06,
+      "loss": 0.3763,
+      "step": 40203
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.498077342871056e-06,
+      "loss": 0.3307,
+      "step": 40204
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.497741507564691e-06,
+      "loss": 0.2554,
+      "step": 40205
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.497405672258325e-06,
+      "loss": 0.233,
+      "step": 40206
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4970698369519594e-06,
+      "loss": 0.1894,
+      "step": 40207
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.496734001645594e-06,
+      "loss": 0.3146,
+      "step": 40208
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.496398166339227e-06,
+      "loss": 0.2977,
+      "step": 40209
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.496062331032862e-06,
+      "loss": 0.1661,
+      "step": 40210
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.495726495726496e-06,
+      "loss": 0.4539,
+      "step": 40211
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.49539066042013e-06,
+      "loss": 0.3831,
+      "step": 40212
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.495054825113765e-06,
+      "loss": 0.1107,
+      "step": 40213
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4947189898073994e-06,
+      "loss": 0.226,
+      "step": 40214
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.494383154501032e-06,
+      "loss": 0.4306,
+      "step": 40215
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.494047319194667e-06,
+      "loss": 0.1853,
+      "step": 40216
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.493711483888302e-06,
+      "loss": 0.1747,
+      "step": 40217
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4933756485819355e-06,
+      "loss": 0.2508,
+      "step": 40218
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.49303981327557e-06,
+      "loss": 0.3189,
+      "step": 40219
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.492703977969205e-06,
+      "loss": 0.2183,
+      "step": 40220
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4923681426628386e-06,
+      "loss": 0.4083,
+      "step": 40221
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.492032307356473e-06,
+      "loss": 0.2431,
+      "step": 40222
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.491696472050107e-06,
+      "loss": 0.2884,
+      "step": 40223
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.491360636743741e-06,
+      "loss": 0.1961,
+      "step": 40224
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4910248014373755e-06,
+      "loss": 0.2204,
+      "step": 40225
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.49068896613101e-06,
+      "loss": 0.2511,
+      "step": 40226
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.490353130824644e-06,
+      "loss": 0.2684,
+      "step": 40227
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4900172955182786e-06,
+      "loss": 0.2091,
+      "step": 40228
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.489681460211913e-06,
+      "loss": 0.2952,
+      "step": 40229
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.489345624905546e-06,
+      "loss": 0.3919,
+      "step": 40230
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.489009789599181e-06,
+      "loss": 0.3779,
+      "step": 40231
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4886739542928155e-06,
+      "loss": 0.3497,
+      "step": 40232
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.488338118986449e-06,
+      "loss": 0.3046,
+      "step": 40233
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.488002283680084e-06,
+      "loss": 0.3817,
+      "step": 40234
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4876664483737186e-06,
+      "loss": 0.1408,
+      "step": 40235
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4873306130673515e-06,
+      "loss": 0.3713,
+      "step": 40236
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.486994777760986e-06,
+      "loss": 0.1859,
+      "step": 40237
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.486658942454621e-06,
+      "loss": 0.3247,
+      "step": 40238
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.486323107148255e-06,
+      "loss": 0.2586,
+      "step": 40239
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.485987271841889e-06,
+      "loss": 0.2435,
+      "step": 40240
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.485651436535524e-06,
+      "loss": 0.2746,
+      "step": 40241
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.485315601229158e-06,
+      "loss": 0.3394,
+      "step": 40242
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.484979765922792e-06,
+      "loss": 0.3806,
+      "step": 40243
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.484643930616427e-06,
+      "loss": 0.1552,
+      "step": 40244
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.48430809531006e-06,
+      "loss": 0.2064,
+      "step": 40245
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.483972260003695e-06,
+      "loss": 0.2833,
+      "step": 40246
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.483636424697329e-06,
+      "loss": 0.3325,
+      "step": 40247
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.483300589390963e-06,
+      "loss": 0.2028,
+      "step": 40248
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.482964754084598e-06,
+      "loss": 0.3038,
+      "step": 40249
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.482628918778232e-06,
+      "loss": 0.3641,
+      "step": 40250
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.482293083471865e-06,
+      "loss": 0.2908,
+      "step": 40251
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4819572481655e-06,
+      "loss": 0.2136,
+      "step": 40252
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.481621412859135e-06,
+      "loss": 0.2369,
+      "step": 40253
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.481285577552768e-06,
+      "loss": 0.317,
+      "step": 40254
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.480949742246403e-06,
+      "loss": 0.1824,
+      "step": 40255
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.480613906940038e-06,
+      "loss": 0.2318,
+      "step": 40256
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4802780716336715e-06,
+      "loss": 0.404,
+      "step": 40257
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.479942236327305e-06,
+      "loss": 0.155,
+      "step": 40258
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.47960640102094e-06,
+      "loss": 0.1021,
+      "step": 40259
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.479270565714574e-06,
+      "loss": 0.3283,
+      "step": 40260
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.478934730408208e-06,
+      "loss": 0.2174,
+      "step": 40261
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.478598895101843e-06,
+      "loss": 0.2733,
+      "step": 40262
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.478263059795477e-06,
+      "loss": 0.1615,
+      "step": 40263
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4779272244891115e-06,
+      "loss": 0.2372,
+      "step": 40264
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.477591389182746e-06,
+      "loss": 0.2951,
+      "step": 40265
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.477255553876379e-06,
+      "loss": 0.2599,
+      "step": 40266
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.476919718570014e-06,
+      "loss": 0.1916,
+      "step": 40267
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.476583883263648e-06,
+      "loss": 0.3136,
+      "step": 40268
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.476248047957282e-06,
+      "loss": 0.332,
+      "step": 40269
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.475912212650917e-06,
+      "loss": 0.3717,
+      "step": 40270
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4755763773445515e-06,
+      "loss": 0.1999,
+      "step": 40271
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.475240542038184e-06,
+      "loss": 0.2349,
+      "step": 40272
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.474904706731819e-06,
+      "loss": 0.3112,
+      "step": 40273
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.474568871425454e-06,
+      "loss": 0.1705,
+      "step": 40274
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4742330361190875e-06,
+      "loss": 0.1434,
+      "step": 40275
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.473897200812722e-06,
+      "loss": 0.3116,
+      "step": 40276
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.473561365506357e-06,
+      "loss": 0.3579,
+      "step": 40277
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.473225530199991e-06,
+      "loss": 0.2128,
+      "step": 40278
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.472889694893625e-06,
+      "loss": 0.2551,
+      "step": 40279
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.472553859587259e-06,
+      "loss": 0.422,
+      "step": 40280
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.472218024280893e-06,
+      "loss": 0.2804,
+      "step": 40281
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4718821889745275e-06,
+      "loss": 0.234,
+      "step": 40282
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.471546353668162e-06,
+      "loss": 0.1801,
+      "step": 40283
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.471210518361796e-06,
+      "loss": 0.4195,
+      "step": 40284
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.470874683055431e-06,
+      "loss": 0.1712,
+      "step": 40285
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.470538847749065e-06,
+      "loss": 0.2385,
+      "step": 40286
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.470203012442698e-06,
+      "loss": 0.1867,
+      "step": 40287
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.469867177136333e-06,
+      "loss": 0.2442,
+      "step": 40288
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.469531341829967e-06,
+      "loss": 0.4152,
+      "step": 40289
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.469195506523601e-06,
+      "loss": 0.2265,
+      "step": 40290
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.468859671217236e-06,
+      "loss": 0.2583,
+      "step": 40291
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.468523835910869e-06,
+      "loss": 0.2803,
+      "step": 40292
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4681880006045035e-06,
+      "loss": 0.3332,
+      "step": 40293
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.467852165298138e-06,
+      "loss": 0.2126,
+      "step": 40294
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.467516329991772e-06,
+      "loss": 0.3734,
+      "step": 40295
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.467180494685407e-06,
+      "loss": 0.1965,
+      "step": 40296
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.466844659379041e-06,
+      "loss": 0.3109,
+      "step": 40297
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.466508824072675e-06,
+      "loss": 0.2748,
+      "step": 40298
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.46617298876631e-06,
+      "loss": 0.3192,
+      "step": 40299
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.465837153459944e-06,
+      "loss": 0.3256,
+      "step": 40300
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.465501318153577e-06,
+      "loss": 0.3372,
+      "step": 40301
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.465165482847212e-06,
+      "loss": 0.4369,
+      "step": 40302
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.464829647540847e-06,
+      "loss": 0.2097,
+      "step": 40303
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4644938122344804e-06,
+      "loss": 0.3452,
+      "step": 40304
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.464157976928115e-06,
+      "loss": 0.3292,
+      "step": 40305
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.46382214162175e-06,
+      "loss": 0.3565,
+      "step": 40306
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.463486306315383e-06,
+      "loss": 0.2227,
+      "step": 40307
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.463150471009017e-06,
+      "loss": 0.3817,
+      "step": 40308
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.462814635702652e-06,
+      "loss": 0.1322,
+      "step": 40309
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.462478800396286e-06,
+      "loss": 0.1979,
+      "step": 40310
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.46214296508992e-06,
+      "loss": 0.3179,
+      "step": 40311
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.461807129783555e-06,
+      "loss": 0.1891,
+      "step": 40312
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.461471294477189e-06,
+      "loss": 0.374,
+      "step": 40313
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.461135459170823e-06,
+      "loss": 0.2449,
+      "step": 40314
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.460799623864457e-06,
+      "loss": 0.2538,
+      "step": 40315
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.460463788558091e-06,
+      "loss": 0.2516,
+      "step": 40316
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.460127953251726e-06,
+      "loss": 0.2367,
+      "step": 40317
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.45979211794536e-06,
+      "loss": 0.2696,
+      "step": 40318
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.459456282638994e-06,
+      "loss": 0.1962,
+      "step": 40319
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.459120447332629e-06,
+      "loss": 0.3542,
+      "step": 40320
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4587846120262635e-06,
+      "loss": 0.3466,
+      "step": 40321
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4584487767198965e-06,
+      "loss": 0.292,
+      "step": 40322
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.458112941413531e-06,
+      "loss": 0.2309,
+      "step": 40323
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.457777106107166e-06,
+      "loss": 0.3363,
+      "step": 40324
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4574412708007996e-06,
+      "loss": 0.2371,
+      "step": 40325
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.457105435494434e-06,
+      "loss": 0.2631,
+      "step": 40326
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.456769600188069e-06,
+      "loss": 0.3536,
+      "step": 40327
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.456433764881702e-06,
+      "loss": 0.2767,
+      "step": 40328
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4560979295753364e-06,
+      "loss": 0.2203,
+      "step": 40329
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.455762094268971e-06,
+      "loss": 0.3469,
+      "step": 40330
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.455426258962605e-06,
+      "loss": 0.1564,
+      "step": 40331
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4550904236562395e-06,
+      "loss": 0.3483,
+      "step": 40332
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.454754588349874e-06,
+      "loss": 0.4337,
+      "step": 40333
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.454418753043508e-06,
+      "loss": 0.2328,
+      "step": 40334
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.454082917737143e-06,
+      "loss": 0.3269,
+      "step": 40335
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4537470824307764e-06,
+      "loss": 0.235,
+      "step": 40336
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.45341124712441e-06,
+      "loss": 0.1982,
+      "step": 40337
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.453075411818045e-06,
+      "loss": 0.3176,
+      "step": 40338
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4527395765116795e-06,
+      "loss": 0.185,
+      "step": 40339
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.452403741205313e-06,
+      "loss": 0.3069,
+      "step": 40340
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.452067905898948e-06,
+      "loss": 0.3836,
+      "step": 40341
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.451732070592583e-06,
+      "loss": 0.333,
+      "step": 40342
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.451396235286216e-06,
+      "loss": 0.2896,
+      "step": 40343
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.45106039997985e-06,
+      "loss": 0.2282,
+      "step": 40344
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.450724564673485e-06,
+      "loss": 0.2338,
+      "step": 40345
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.450388729367119e-06,
+      "loss": 0.362,
+      "step": 40346
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.450052894060753e-06,
+      "loss": 0.2086,
+      "step": 40347
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.449717058754388e-06,
+      "loss": 0.5398,
+      "step": 40348
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.449381223448021e-06,
+      "loss": 0.2883,
+      "step": 40349
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4490453881416556e-06,
+      "loss": 0.1935,
+      "step": 40350
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.44870955283529e-06,
+      "loss": 0.2269,
+      "step": 40351
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.448373717528924e-06,
+      "loss": 0.2594,
+      "step": 40352
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.448037882222559e-06,
+      "loss": 0.1394,
+      "step": 40353
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.447702046916193e-06,
+      "loss": 0.3227,
+      "step": 40354
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.447366211609827e-06,
+      "loss": 0.2681,
+      "step": 40355
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.447030376303462e-06,
+      "loss": 0.2054,
+      "step": 40356
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.446694540997096e-06,
+      "loss": 0.2098,
+      "step": 40357
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.446358705690729e-06,
+      "loss": 0.2896,
+      "step": 40358
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.446022870384364e-06,
+      "loss": 0.2061,
+      "step": 40359
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.445687035077999e-06,
+      "loss": 0.3868,
+      "step": 40360
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4453511997716325e-06,
+      "loss": 0.1858,
+      "step": 40361
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.445015364465267e-06,
+      "loss": 0.2049,
+      "step": 40362
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.444679529158902e-06,
+      "loss": 0.3378,
+      "step": 40363
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.444343693852535e-06,
+      "loss": 0.2663,
+      "step": 40364
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.444007858546169e-06,
+      "loss": 0.1673,
+      "step": 40365
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.443672023239804e-06,
+      "loss": 0.198,
+      "step": 40366
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.443336187933438e-06,
+      "loss": 0.2062,
+      "step": 40367
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4430003526270724e-06,
+      "loss": 0.1748,
+      "step": 40368
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.442664517320707e-06,
+      "loss": 0.1161,
+      "step": 40369
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.44232868201434e-06,
+      "loss": 0.1627,
+      "step": 40370
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.441992846707975e-06,
+      "loss": 0.3174,
+      "step": 40371
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.441657011401609e-06,
+      "loss": 0.2571,
+      "step": 40372
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.441321176095243e-06,
+      "loss": 0.3242,
+      "step": 40373
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.440985340788878e-06,
+      "loss": 0.1984,
+      "step": 40374
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4406495054825124e-06,
+      "loss": 0.3981,
+      "step": 40375
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.440313670176146e-06,
+      "loss": 0.1582,
+      "step": 40376
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.439977834869781e-06,
+      "loss": 0.1578,
+      "step": 40377
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4396419995634155e-06,
+      "loss": 0.1778,
+      "step": 40378
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4393061642570485e-06,
+      "loss": 0.2047,
+      "step": 40379
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.438970328950683e-06,
+      "loss": 0.2085,
+      "step": 40380
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.438634493644317e-06,
+      "loss": 0.1569,
+      "step": 40381
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.438298658337952e-06,
+      "loss": 0.1551,
+      "step": 40382
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.437962823031586e-06,
+      "loss": 0.2753,
+      "step": 40383
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.437626987725219e-06,
+      "loss": 0.3687,
+      "step": 40384
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.437291152418854e-06,
+      "loss": 0.2741,
+      "step": 40385
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4369553171124885e-06,
+      "loss": 0.4763,
+      "step": 40386
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.436619481806122e-06,
+      "loss": 0.3054,
+      "step": 40387
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.436283646499757e-06,
+      "loss": 0.2491,
+      "step": 40388
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.4359478111933916e-06,
+      "loss": 0.3648,
+      "step": 40389
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.435611975887025e-06,
+      "loss": 0.2681,
+      "step": 40390
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.43527614058066e-06,
+      "loss": 0.1484,
+      "step": 40391
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.434940305274295e-06,
+      "loss": 0.22,
+      "step": 40392
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.434604469967928e-06,
+      "loss": 0.2418,
+      "step": 40393
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.434268634661562e-06,
+      "loss": 0.364,
+      "step": 40394
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.433932799355197e-06,
+      "loss": 0.2003,
+      "step": 40395
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 6.433596964048831e-06,
+      "loss": 0.2549,
+      "step": 40396
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.433261128742465e-06,
+      "loss": 0.2106,
+      "step": 40397
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.4329252934361e-06,
+      "loss": 0.3325,
+      "step": 40398
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.432589458129733e-06,
+      "loss": 0.1993,
+      "step": 40399
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.432253622823368e-06,
+      "loss": 0.2253,
+      "step": 40400
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.431917787517002e-06,
+      "loss": 0.2613,
+      "step": 40401
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.431581952210636e-06,
+      "loss": 0.5158,
+      "step": 40402
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.431246116904271e-06,
+      "loss": 0.2116,
+      "step": 40403
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.430910281597905e-06,
+      "loss": 0.3039,
+      "step": 40404
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.430574446291538e-06,
+      "loss": 0.5159,
+      "step": 40405
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.430238610985173e-06,
+      "loss": 0.3376,
+      "step": 40406
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.429902775678808e-06,
+      "loss": 0.1773,
+      "step": 40407
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.429566940372441e-06,
+      "loss": 0.291,
+      "step": 40408
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.429231105066076e-06,
+      "loss": 0.3192,
+      "step": 40409
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.428895269759711e-06,
+      "loss": 0.5335,
+      "step": 40410
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.4285594344533445e-06,
+      "loss": 0.2681,
+      "step": 40411
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.428223599146979e-06,
+      "loss": 0.2093,
+      "step": 40412
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.427887763840614e-06,
+      "loss": 0.3919,
+      "step": 40413
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.427551928534247e-06,
+      "loss": 0.1832,
+      "step": 40414
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.427216093227881e-06,
+      "loss": 0.3155,
+      "step": 40415
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.426880257921516e-06,
+      "loss": 0.2315,
+      "step": 40416
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.42654442261515e-06,
+      "loss": 0.3111,
+      "step": 40417
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.4262085873087845e-06,
+      "loss": 0.2253,
+      "step": 40418
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.425872752002419e-06,
+      "loss": 0.346,
+      "step": 40419
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.425536916696052e-06,
+      "loss": 0.3263,
+      "step": 40420
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.425201081389687e-06,
+      "loss": 0.4171,
+      "step": 40421
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.424865246083321e-06,
+      "loss": 0.1721,
+      "step": 40422
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.424529410776955e-06,
+      "loss": 0.3922,
+      "step": 40423
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.42419357547059e-06,
+      "loss": 0.2277,
+      "step": 40424
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.4238577401642245e-06,
+      "loss": 0.1708,
+      "step": 40425
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.423521904857858e-06,
+      "loss": 0.1409,
+      "step": 40426
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.423186069551492e-06,
+      "loss": 0.2624,
+      "step": 40427
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.422850234245127e-06,
+      "loss": 0.1827,
+      "step": 40428
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.4225143989387605e-06,
+      "loss": 0.2206,
+      "step": 40429
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.422178563632395e-06,
+      "loss": 0.2905,
+      "step": 40430
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.42184272832603e-06,
+      "loss": 0.2233,
+      "step": 40431
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.421506893019664e-06,
+      "loss": 0.2096,
+      "step": 40432
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.421171057713298e-06,
+      "loss": 0.1447,
+      "step": 40433
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.420835222406933e-06,
+      "loss": 0.4052,
+      "step": 40434
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.420499387100566e-06,
+      "loss": 0.2036,
+      "step": 40435
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.4201635517942005e-06,
+      "loss": 0.3311,
+      "step": 40436
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.419827716487835e-06,
+      "loss": 0.2288,
+      "step": 40437
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.419491881181469e-06,
+      "loss": 0.1387,
+      "step": 40438
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.419156045875104e-06,
+      "loss": 0.2545,
+      "step": 40439
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.418820210568738e-06,
+      "loss": 0.1779,
+      "step": 40440
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.418484375262371e-06,
+      "loss": 0.2813,
+      "step": 40441
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.418148539956006e-06,
+      "loss": 0.3072,
+      "step": 40442
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.4178127046496405e-06,
+      "loss": 0.2814,
+      "step": 40443
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.417476869343274e-06,
+      "loss": 0.3142,
+      "step": 40444
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.417141034036909e-06,
+      "loss": 0.297,
+      "step": 40445
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.416805198730544e-06,
+      "loss": 0.0813,
+      "step": 40446
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.416469363424177e-06,
+      "loss": 0.4415,
+      "step": 40447
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.416133528117812e-06,
+      "loss": 0.2594,
+      "step": 40448
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.415797692811446e-06,
+      "loss": 0.4466,
+      "step": 40449
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.41546185750508e-06,
+      "loss": 0.2104,
+      "step": 40450
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.415126022198714e-06,
+      "loss": 0.4713,
+      "step": 40451
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.414790186892349e-06,
+      "loss": 0.3016,
+      "step": 40452
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.414454351585983e-06,
+      "loss": 0.3024,
+      "step": 40453
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.414118516279617e-06,
+      "loss": 0.2544,
+      "step": 40454
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.413782680973252e-06,
+      "loss": 0.2852,
+      "step": 40455
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.413446845666885e-06,
+      "loss": 0.1465,
+      "step": 40456
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.41311101036052e-06,
+      "loss": 0.2524,
+      "step": 40457
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.412775175054154e-06,
+      "loss": 0.1969,
+      "step": 40458
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.412439339747788e-06,
+      "loss": 0.279,
+      "step": 40459
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.412103504441423e-06,
+      "loss": 0.2796,
+      "step": 40460
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.411767669135057e-06,
+      "loss": 0.2329,
+      "step": 40461
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.41143183382869e-06,
+      "loss": 0.1666,
+      "step": 40462
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.411095998522325e-06,
+      "loss": 0.2682,
+      "step": 40463
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.41076016321596e-06,
+      "loss": 0.2996,
+      "step": 40464
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.4104243279095934e-06,
+      "loss": 0.1901,
+      "step": 40465
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.410088492603228e-06,
+      "loss": 0.3538,
+      "step": 40466
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.409752657296863e-06,
+      "loss": 0.2547,
+      "step": 40467
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.4094168219904965e-06,
+      "loss": 0.2258,
+      "step": 40468
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.409080986684131e-06,
+      "loss": 0.3537,
+      "step": 40469
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.408745151377764e-06,
+      "loss": 0.2067,
+      "step": 40470
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.408409316071399e-06,
+      "loss": 0.2538,
+      "step": 40471
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.4080734807650334e-06,
+      "loss": 0.2423,
+      "step": 40472
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.407737645458667e-06,
+      "loss": 0.2729,
+      "step": 40473
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.407401810152302e-06,
+      "loss": 0.3547,
+      "step": 40474
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.4070659748459365e-06,
+      "loss": 0.2174,
+      "step": 40475
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.4067301395395695e-06,
+      "loss": 0.1852,
+      "step": 40476
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.406394304233204e-06,
+      "loss": 0.2471,
+      "step": 40477
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.406058468926839e-06,
+      "loss": 0.1551,
+      "step": 40478
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.4057226336204726e-06,
+      "loss": 0.2631,
+      "step": 40479
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.405386798314107e-06,
+      "loss": 0.2492,
+      "step": 40480
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.405050963007742e-06,
+      "loss": 0.1325,
+      "step": 40481
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.404715127701376e-06,
+      "loss": 0.3185,
+      "step": 40482
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.4043792923950095e-06,
+      "loss": 0.2244,
+      "step": 40483
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.404043457088644e-06,
+      "loss": 0.3488,
+      "step": 40484
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.403707621782278e-06,
+      "loss": 0.1195,
+      "step": 40485
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.4033717864759126e-06,
+      "loss": 0.2713,
+      "step": 40486
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.403035951169547e-06,
+      "loss": 0.1607,
+      "step": 40487
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.402700115863181e-06,
+      "loss": 0.4618,
+      "step": 40488
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.402364280556816e-06,
+      "loss": 0.283,
+      "step": 40489
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.40202844525045e-06,
+      "loss": 0.2184,
+      "step": 40490
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.401692609944083e-06,
+      "loss": 0.1143,
+      "step": 40491
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.401356774637718e-06,
+      "loss": 0.2549,
+      "step": 40492
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.4010209393313526e-06,
+      "loss": 0.3166,
+      "step": 40493
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.400685104024986e-06,
+      "loss": 0.1992,
+      "step": 40494
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.400349268718621e-06,
+      "loss": 0.4362,
+      "step": 40495
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.400013433412256e-06,
+      "loss": 0.2188,
+      "step": 40496
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.399677598105889e-06,
+      "loss": 0.2031,
+      "step": 40497
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.399341762799523e-06,
+      "loss": 0.4523,
+      "step": 40498
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.399005927493158e-06,
+      "loss": 0.3817,
+      "step": 40499
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.398670092186792e-06,
+      "loss": 0.3052,
+      "step": 40500
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.398334256880426e-06,
+      "loss": 0.1621,
+      "step": 40501
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.397998421574061e-06,
+      "loss": 0.1502,
+      "step": 40502
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.397662586267695e-06,
+      "loss": 0.182,
+      "step": 40503
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.3973267509613294e-06,
+      "loss": 0.295,
+      "step": 40504
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.396990915654964e-06,
+      "loss": 0.343,
+      "step": 40505
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.396655080348597e-06,
+      "loss": 0.3897,
+      "step": 40506
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.396319245042232e-06,
+      "loss": 0.1872,
+      "step": 40507
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.395983409735866e-06,
+      "loss": 0.4257,
+      "step": 40508
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.3956475744295e-06,
+      "loss": 0.3602,
+      "step": 40509
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.395311739123135e-06,
+      "loss": 0.2792,
+      "step": 40510
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.3949759038167694e-06,
+      "loss": 0.1443,
+      "step": 40511
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.394640068510402e-06,
+      "loss": 0.3543,
+      "step": 40512
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.394304233204037e-06,
+      "loss": 0.2037,
+      "step": 40513
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.393968397897672e-06,
+      "loss": 0.1697,
+      "step": 40514
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.3936325625913055e-06,
+      "loss": 0.242,
+      "step": 40515
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.39329672728494e-06,
+      "loss": 0.4421,
+      "step": 40516
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.392960891978575e-06,
+      "loss": 0.2005,
+      "step": 40517
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.392625056672208e-06,
+      "loss": 0.2284,
+      "step": 40518
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.392289221365842e-06,
+      "loss": 0.4007,
+      "step": 40519
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.391953386059477e-06,
+      "loss": 0.2338,
+      "step": 40520
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.391617550753111e-06,
+      "loss": 0.2831,
+      "step": 40521
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.3912817154467455e-06,
+      "loss": 0.3546,
+      "step": 40522
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.39094588014038e-06,
+      "loss": 0.2854,
+      "step": 40523
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.390610044834014e-06,
+      "loss": 0.2819,
+      "step": 40524
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.3902742095276486e-06,
+      "loss": 0.3646,
+      "step": 40525
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.389938374221283e-06,
+      "loss": 0.3734,
+      "step": 40526
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.389602538914916e-06,
+      "loss": 0.1985,
+      "step": 40527
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.389266703608551e-06,
+      "loss": 0.1251,
+      "step": 40528
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.3889308683021855e-06,
+      "loss": 0.3181,
+      "step": 40529
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.388595032995819e-06,
+      "loss": 0.3581,
+      "step": 40530
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.388259197689454e-06,
+      "loss": 0.256,
+      "step": 40531
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.3879233623830886e-06,
+      "loss": 0.4234,
+      "step": 40532
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.3875875270767215e-06,
+      "loss": 0.2921,
+      "step": 40533
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.387251691770356e-06,
+      "loss": 0.2445,
+      "step": 40534
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.386915856463991e-06,
+      "loss": 0.273,
+      "step": 40535
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.386580021157625e-06,
+      "loss": 0.2691,
+      "step": 40536
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.386244185851259e-06,
+      "loss": 0.3218,
+      "step": 40537
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.385908350544894e-06,
+      "loss": 0.3143,
+      "step": 40538
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.385572515238528e-06,
+      "loss": 0.2129,
+      "step": 40539
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.3852366799321615e-06,
+      "loss": 0.1347,
+      "step": 40540
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.384900844625796e-06,
+      "loss": 0.2176,
+      "step": 40541
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.38456500931943e-06,
+      "loss": 0.3164,
+      "step": 40542
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.384229174013065e-06,
+      "loss": 0.3123,
+      "step": 40543
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.383893338706699e-06,
+      "loss": 0.4167,
+      "step": 40544
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.383557503400333e-06,
+      "loss": 0.1575,
+      "step": 40545
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.383221668093968e-06,
+      "loss": 0.2642,
+      "step": 40546
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.382885832787602e-06,
+      "loss": 0.2845,
+      "step": 40547
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.382549997481235e-06,
+      "loss": 0.175,
+      "step": 40548
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.38221416217487e-06,
+      "loss": 0.2152,
+      "step": 40549
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.381878326868505e-06,
+      "loss": 0.2518,
+      "step": 40550
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.381542491562138e-06,
+      "loss": 0.181,
+      "step": 40551
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.381206656255773e-06,
+      "loss": 0.2564,
+      "step": 40552
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.380870820949408e-06,
+      "loss": 0.2394,
+      "step": 40553
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.380534985643041e-06,
+      "loss": 0.2461,
+      "step": 40554
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.380199150336675e-06,
+      "loss": 0.1468,
+      "step": 40555
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.37986331503031e-06,
+      "loss": 0.2763,
+      "step": 40556
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.379527479723944e-06,
+      "loss": 0.2826,
+      "step": 40557
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.379191644417578e-06,
+      "loss": 0.1396,
+      "step": 40558
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.378855809111213e-06,
+      "loss": 0.1568,
+      "step": 40559
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.378519973804847e-06,
+      "loss": 0.1823,
+      "step": 40560
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.3781841384984815e-06,
+      "loss": 0.3293,
+      "step": 40561
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.377848303192114e-06,
+      "loss": 0.1857,
+      "step": 40562
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.377512467885749e-06,
+      "loss": 0.365,
+      "step": 40563
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.377176632579384e-06,
+      "loss": 0.1132,
+      "step": 40564
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.3768407972730175e-06,
+      "loss": 0.399,
+      "step": 40565
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.376504961966652e-06,
+      "loss": 0.2662,
+      "step": 40566
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.376169126660287e-06,
+      "loss": 0.349,
+      "step": 40567
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.37583329135392e-06,
+      "loss": 0.28,
+      "step": 40568
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.375497456047554e-06,
+      "loss": 0.2741,
+      "step": 40569
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.375161620741189e-06,
+      "loss": 0.2416,
+      "step": 40570
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.374825785434823e-06,
+      "loss": 0.1901,
+      "step": 40571
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.3744899501284575e-06,
+      "loss": 0.2582,
+      "step": 40572
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.374154114822092e-06,
+      "loss": 0.1796,
+      "step": 40573
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.373818279515725e-06,
+      "loss": 0.3629,
+      "step": 40574
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.37348244420936e-06,
+      "loss": 0.2655,
+      "step": 40575
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.373146608902994e-06,
+      "loss": 0.271,
+      "step": 40576
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.372810773596628e-06,
+      "loss": 0.2059,
+      "step": 40577
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.372474938290263e-06,
+      "loss": 0.363,
+      "step": 40578
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.3721391029838975e-06,
+      "loss": 0.4401,
+      "step": 40579
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.371803267677531e-06,
+      "loss": 0.3004,
+      "step": 40580
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.371467432371166e-06,
+      "loss": 0.2399,
+      "step": 40581
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.371131597064801e-06,
+      "loss": 0.1875,
+      "step": 40582
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.3707957617584335e-06,
+      "loss": 0.2149,
+      "step": 40583
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.370459926452068e-06,
+      "loss": 0.2752,
+      "step": 40584
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.370124091145703e-06,
+      "loss": 0.2525,
+      "step": 40585
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.369788255839337e-06,
+      "loss": 0.243,
+      "step": 40586
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.369452420532971e-06,
+      "loss": 0.3521,
+      "step": 40587
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.369116585226606e-06,
+      "loss": 0.173,
+      "step": 40588
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.368780749920239e-06,
+      "loss": 0.4404,
+      "step": 40589
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.3684449146138735e-06,
+      "loss": 0.33,
+      "step": 40590
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.368109079307508e-06,
+      "loss": 0.2297,
+      "step": 40591
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.367773244001142e-06,
+      "loss": 0.3451,
+      "step": 40592
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.367437408694777e-06,
+      "loss": 0.2841,
+      "step": 40593
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.367101573388411e-06,
+      "loss": 0.2683,
+      "step": 40594
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 6.366765738082045e-06,
+      "loss": 0.4419,
+      "step": 40595
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.366429902775679e-06,
+      "loss": 0.3063,
+      "step": 40596
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3660940674693135e-06,
+      "loss": 0.2541,
+      "step": 40597
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.365758232162947e-06,
+      "loss": 0.1315,
+      "step": 40598
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.365422396856582e-06,
+      "loss": 0.2061,
+      "step": 40599
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.365086561550217e-06,
+      "loss": 0.2569,
+      "step": 40600
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.36475072624385e-06,
+      "loss": 0.2724,
+      "step": 40601
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.364414890937485e-06,
+      "loss": 0.4005,
+      "step": 40602
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.36407905563112e-06,
+      "loss": 0.3952,
+      "step": 40603
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.363743220324753e-06,
+      "loss": 0.2933,
+      "step": 40604
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.363407385018387e-06,
+      "loss": 0.3175,
+      "step": 40605
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.363071549712022e-06,
+      "loss": 0.2004,
+      "step": 40606
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.362735714405656e-06,
+      "loss": 0.2122,
+      "step": 40607
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.36239987909929e-06,
+      "loss": 0.2806,
+      "step": 40608
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.362064043792925e-06,
+      "loss": 0.1633,
+      "step": 40609
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.361728208486558e-06,
+      "loss": 0.3075,
+      "step": 40610
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.361392373180193e-06,
+      "loss": 0.345,
+      "step": 40611
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.361056537873827e-06,
+      "loss": 0.1513,
+      "step": 40612
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.360720702567461e-06,
+      "loss": 0.3025,
+      "step": 40613
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.360384867261096e-06,
+      "loss": 0.1803,
+      "step": 40614
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.36004903195473e-06,
+      "loss": 0.3031,
+      "step": 40615
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.359713196648364e-06,
+      "loss": 0.269,
+      "step": 40616
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.359377361341999e-06,
+      "loss": 0.313,
+      "step": 40617
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.359041526035633e-06,
+      "loss": 0.3397,
+      "step": 40618
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3587056907292665e-06,
+      "loss": 0.3234,
+      "step": 40619
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.358369855422901e-06,
+      "loss": 0.2259,
+      "step": 40620
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.358034020116536e-06,
+      "loss": 0.226,
+      "step": 40621
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3576981848101695e-06,
+      "loss": 0.2915,
+      "step": 40622
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.357362349503804e-06,
+      "loss": 0.1577,
+      "step": 40623
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.357026514197439e-06,
+      "loss": 0.2315,
+      "step": 40624
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.356690678891072e-06,
+      "loss": 0.2647,
+      "step": 40625
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3563548435847064e-06,
+      "loss": 0.3901,
+      "step": 40626
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.356019008278341e-06,
+      "loss": 0.3596,
+      "step": 40627
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.355683172971975e-06,
+      "loss": 0.2937,
+      "step": 40628
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3553473376656095e-06,
+      "loss": 0.202,
+      "step": 40629
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.355011502359244e-06,
+      "loss": 0.2397,
+      "step": 40630
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.354675667052877e-06,
+      "loss": 0.1783,
+      "step": 40631
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.354339831746512e-06,
+      "loss": 0.2568,
+      "step": 40632
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3540039964401464e-06,
+      "loss": 0.3599,
+      "step": 40633
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.35366816113378e-06,
+      "loss": 0.2656,
+      "step": 40634
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.353332325827415e-06,
+      "loss": 0.3376,
+      "step": 40635
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3529964905210495e-06,
+      "loss": 0.1636,
+      "step": 40636
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.352660655214683e-06,
+      "loss": 0.3789,
+      "step": 40637
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.352324819908318e-06,
+      "loss": 0.1404,
+      "step": 40638
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.351988984601953e-06,
+      "loss": 0.3032,
+      "step": 40639
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.351653149295586e-06,
+      "loss": 0.24,
+      "step": 40640
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.35131731398922e-06,
+      "loss": 0.34,
+      "step": 40641
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.350981478682855e-06,
+      "loss": 0.2902,
+      "step": 40642
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.350645643376489e-06,
+      "loss": 0.1251,
+      "step": 40643
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.350309808070123e-06,
+      "loss": 0.2547,
+      "step": 40644
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.349973972763758e-06,
+      "loss": 0.1616,
+      "step": 40645
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.349638137457391e-06,
+      "loss": 0.2045,
+      "step": 40646
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3493023021510256e-06,
+      "loss": 0.3757,
+      "step": 40647
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.34896646684466e-06,
+      "loss": 0.2375,
+      "step": 40648
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.348630631538294e-06,
+      "loss": 0.3018,
+      "step": 40649
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.348294796231929e-06,
+      "loss": 0.2687,
+      "step": 40650
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3479589609255625e-06,
+      "loss": 0.2963,
+      "step": 40651
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.347623125619197e-06,
+      "loss": 0.1765,
+      "step": 40652
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.347287290312831e-06,
+      "loss": 0.3871,
+      "step": 40653
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.346951455006465e-06,
+      "loss": 0.2833,
+      "step": 40654
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.346615619700099e-06,
+      "loss": 0.4045,
+      "step": 40655
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.346279784393734e-06,
+      "loss": 0.2681,
+      "step": 40656
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.345943949087368e-06,
+      "loss": 0.2615,
+      "step": 40657
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3456081137810025e-06,
+      "loss": 0.2347,
+      "step": 40658
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.345272278474637e-06,
+      "loss": 0.2468,
+      "step": 40659
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.34493644316827e-06,
+      "loss": 0.2538,
+      "step": 40660
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.344600607861905e-06,
+      "loss": 0.1925,
+      "step": 40661
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.344264772555539e-06,
+      "loss": 0.2131,
+      "step": 40662
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.343928937249173e-06,
+      "loss": 0.2229,
+      "step": 40663
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.343593101942808e-06,
+      "loss": 0.3837,
+      "step": 40664
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3432572666364424e-06,
+      "loss": 0.3855,
+      "step": 40665
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.342921431330075e-06,
+      "loss": 0.0977,
+      "step": 40666
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.34258559602371e-06,
+      "loss": 0.2523,
+      "step": 40667
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.342249760717345e-06,
+      "loss": 0.1673,
+      "step": 40668
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3419139254109785e-06,
+      "loss": 0.3049,
+      "step": 40669
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.341578090104613e-06,
+      "loss": 0.2502,
+      "step": 40670
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.341242254798248e-06,
+      "loss": 0.225,
+      "step": 40671
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.340906419491882e-06,
+      "loss": 0.3509,
+      "step": 40672
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.340570584185516e-06,
+      "loss": 0.1739,
+      "step": 40673
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.340234748879151e-06,
+      "loss": 0.33,
+      "step": 40674
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.339898913572784e-06,
+      "loss": 0.2258,
+      "step": 40675
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3395630782664185e-06,
+      "loss": 0.3029,
+      "step": 40676
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.339227242960053e-06,
+      "loss": 0.2945,
+      "step": 40677
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.338891407653687e-06,
+      "loss": 0.296,
+      "step": 40678
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.338555572347322e-06,
+      "loss": 0.2354,
+      "step": 40679
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.338219737040956e-06,
+      "loss": 0.1959,
+      "step": 40680
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.337883901734589e-06,
+      "loss": 0.1593,
+      "step": 40681
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.337548066428224e-06,
+      "loss": 0.5479,
+      "step": 40682
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3372122311218585e-06,
+      "loss": 0.4364,
+      "step": 40683
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.336876395815492e-06,
+      "loss": 0.3695,
+      "step": 40684
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.336540560509127e-06,
+      "loss": 0.3158,
+      "step": 40685
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3362047252027616e-06,
+      "loss": 0.2541,
+      "step": 40686
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3358688898963945e-06,
+      "loss": 0.3594,
+      "step": 40687
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.335533054590029e-06,
+      "loss": 0.2181,
+      "step": 40688
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.335197219283664e-06,
+      "loss": 0.2531,
+      "step": 40689
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.334861383977298e-06,
+      "loss": 0.153,
+      "step": 40690
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.334525548670932e-06,
+      "loss": 0.2514,
+      "step": 40691
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.334189713364567e-06,
+      "loss": 0.382,
+      "step": 40692
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.333853878058201e-06,
+      "loss": 0.2465,
+      "step": 40693
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.333518042751835e-06,
+      "loss": 0.1749,
+      "step": 40694
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.33318220744547e-06,
+      "loss": 0.2843,
+      "step": 40695
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.332846372139103e-06,
+      "loss": 0.1162,
+      "step": 40696
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.332510536832738e-06,
+      "loss": 0.146,
+      "step": 40697
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.332174701526372e-06,
+      "loss": 0.2145,
+      "step": 40698
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.331838866220006e-06,
+      "loss": 0.2219,
+      "step": 40699
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.331503030913641e-06,
+      "loss": 0.2289,
+      "step": 40700
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.331167195607275e-06,
+      "loss": 0.3013,
+      "step": 40701
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.330831360300908e-06,
+      "loss": 0.1491,
+      "step": 40702
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.330495524994543e-06,
+      "loss": 0.2541,
+      "step": 40703
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.330159689688178e-06,
+      "loss": 0.2259,
+      "step": 40704
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.329823854381811e-06,
+      "loss": 0.1107,
+      "step": 40705
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.329488019075446e-06,
+      "loss": 0.1972,
+      "step": 40706
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.329152183769081e-06,
+      "loss": 0.2471,
+      "step": 40707
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3288163484627145e-06,
+      "loss": 0.1751,
+      "step": 40708
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.328480513156348e-06,
+      "loss": 0.4819,
+      "step": 40709
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.328144677849983e-06,
+      "loss": 0.0842,
+      "step": 40710
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.327808842543617e-06,
+      "loss": 0.4117,
+      "step": 40711
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.327473007237251e-06,
+      "loss": 0.3622,
+      "step": 40712
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.327137171930886e-06,
+      "loss": 0.3322,
+      "step": 40713
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.32680133662452e-06,
+      "loss": 0.2772,
+      "step": 40714
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3264655013181545e-06,
+      "loss": 0.3223,
+      "step": 40715
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.326129666011789e-06,
+      "loss": 0.1876,
+      "step": 40716
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.325793830705422e-06,
+      "loss": 0.2074,
+      "step": 40717
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.325457995399057e-06,
+      "loss": 0.342,
+      "step": 40718
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.325122160092691e-06,
+      "loss": 0.2903,
+      "step": 40719
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.324786324786325e-06,
+      "loss": 0.3836,
+      "step": 40720
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.32445048947996e-06,
+      "loss": 0.2749,
+      "step": 40721
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3241146541735945e-06,
+      "loss": 0.3235,
+      "step": 40722
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3237788188672274e-06,
+      "loss": 0.2661,
+      "step": 40723
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.323442983560862e-06,
+      "loss": 0.2372,
+      "step": 40724
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.323107148254497e-06,
+      "loss": 0.202,
+      "step": 40725
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3227713129481305e-06,
+      "loss": 0.3278,
+      "step": 40726
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.322435477641765e-06,
+      "loss": 0.5917,
+      "step": 40727
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3220996423354e-06,
+      "loss": 0.2876,
+      "step": 40728
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.321763807029034e-06,
+      "loss": 0.2013,
+      "step": 40729
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.321427971722668e-06,
+      "loss": 0.3485,
+      "step": 40730
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.321092136416302e-06,
+      "loss": 0.1299,
+      "step": 40731
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.320756301109936e-06,
+      "loss": 0.1528,
+      "step": 40732
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3204204658035705e-06,
+      "loss": 0.3274,
+      "step": 40733
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.320084630497205e-06,
+      "loss": 0.3367,
+      "step": 40734
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.319748795190839e-06,
+      "loss": 0.3075,
+      "step": 40735
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.319412959884474e-06,
+      "loss": 0.3456,
+      "step": 40736
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.319077124578108e-06,
+      "loss": 0.3397,
+      "step": 40737
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.318741289271741e-06,
+      "loss": 0.1938,
+      "step": 40738
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.318405453965376e-06,
+      "loss": 0.3278,
+      "step": 40739
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.31806961865901e-06,
+      "loss": 0.2665,
+      "step": 40740
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.317733783352644e-06,
+      "loss": 0.3098,
+      "step": 40741
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.317397948046279e-06,
+      "loss": 0.4481,
+      "step": 40742
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.317062112739912e-06,
+      "loss": 0.2752,
+      "step": 40743
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3167262774335466e-06,
+      "loss": 0.2166,
+      "step": 40744
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.316390442127181e-06,
+      "loss": 0.2137,
+      "step": 40745
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.316054606820815e-06,
+      "loss": 0.1896,
+      "step": 40746
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.31571877151445e-06,
+      "loss": 0.2106,
+      "step": 40747
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.315382936208084e-06,
+      "loss": 0.3234,
+      "step": 40748
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.315047100901718e-06,
+      "loss": 0.2118,
+      "step": 40749
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.314711265595353e-06,
+      "loss": 0.1409,
+      "step": 40750
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.314375430288987e-06,
+      "loss": 0.2074,
+      "step": 40751
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.31403959498262e-06,
+      "loss": 0.3307,
+      "step": 40752
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.313703759676255e-06,
+      "loss": 0.2292,
+      "step": 40753
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.31336792436989e-06,
+      "loss": 0.1815,
+      "step": 40754
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3130320890635234e-06,
+      "loss": 0.2309,
+      "step": 40755
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.312696253757158e-06,
+      "loss": 0.4295,
+      "step": 40756
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.312360418450793e-06,
+      "loss": 0.1734,
+      "step": 40757
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.312024583144426e-06,
+      "loss": 0.2757,
+      "step": 40758
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.31168874783806e-06,
+      "loss": 0.2792,
+      "step": 40759
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.311352912531695e-06,
+      "loss": 0.3653,
+      "step": 40760
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.311017077225329e-06,
+      "loss": 0.2737,
+      "step": 40761
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3106812419189634e-06,
+      "loss": 0.2015,
+      "step": 40762
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.310345406612598e-06,
+      "loss": 0.2244,
+      "step": 40763
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.310009571306232e-06,
+      "loss": 0.1699,
+      "step": 40764
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.309673735999866e-06,
+      "loss": 0.3067,
+      "step": 40765
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3093379006935e-06,
+      "loss": 0.1737,
+      "step": 40766
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.309002065387134e-06,
+      "loss": 0.303,
+      "step": 40767
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.308666230080769e-06,
+      "loss": 0.3197,
+      "step": 40768
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.308330394774403e-06,
+      "loss": 0.4525,
+      "step": 40769
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.307994559468037e-06,
+      "loss": 0.3145,
+      "step": 40770
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.307658724161672e-06,
+      "loss": 0.244,
+      "step": 40771
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3073228888553065e-06,
+      "loss": 0.1817,
+      "step": 40772
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3069870535489395e-06,
+      "loss": 0.2218,
+      "step": 40773
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.306651218242574e-06,
+      "loss": 0.2222,
+      "step": 40774
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.306315382936209e-06,
+      "loss": 0.3103,
+      "step": 40775
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3059795476298426e-06,
+      "loss": 0.1664,
+      "step": 40776
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.305643712323477e-06,
+      "loss": 0.2299,
+      "step": 40777
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.305307877017112e-06,
+      "loss": 0.3115,
+      "step": 40778
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.304972041710745e-06,
+      "loss": 0.422,
+      "step": 40779
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3046362064043795e-06,
+      "loss": 0.2119,
+      "step": 40780
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.304300371098014e-06,
+      "loss": 0.3766,
+      "step": 40781
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.303964535791648e-06,
+      "loss": 0.2514,
+      "step": 40782
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3036287004852826e-06,
+      "loss": 0.198,
+      "step": 40783
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.303292865178917e-06,
+      "loss": 0.3824,
+      "step": 40784
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.302957029872551e-06,
+      "loss": 0.2426,
+      "step": 40785
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.302621194566186e-06,
+      "loss": 0.3864,
+      "step": 40786
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.30228535925982e-06,
+      "loss": 0.227,
+      "step": 40787
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.301949523953453e-06,
+      "loss": 0.2751,
+      "step": 40788
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.301613688647088e-06,
+      "loss": 0.2514,
+      "step": 40789
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.3012778533407225e-06,
+      "loss": 0.2786,
+      "step": 40790
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.300942018034356e-06,
+      "loss": 0.3584,
+      "step": 40791
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.300606182727991e-06,
+      "loss": 0.3325,
+      "step": 40792
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 6.300270347421626e-06,
+      "loss": 0.2612,
+      "step": 40793
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.299934512115259e-06,
+      "loss": 0.179,
+      "step": 40794
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.299598676808893e-06,
+      "loss": 0.333,
+      "step": 40795
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.299262841502528e-06,
+      "loss": 0.227,
+      "step": 40796
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.298927006196162e-06,
+      "loss": 0.1954,
+      "step": 40797
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.298591170889796e-06,
+      "loss": 0.4151,
+      "step": 40798
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.298255335583431e-06,
+      "loss": 0.2707,
+      "step": 40799
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.297919500277064e-06,
+      "loss": 0.3232,
+      "step": 40800
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.297583664970699e-06,
+      "loss": 0.2603,
+      "step": 40801
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.297247829664333e-06,
+      "loss": 0.2705,
+      "step": 40802
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.296911994357967e-06,
+      "loss": 0.3783,
+      "step": 40803
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.296576159051602e-06,
+      "loss": 0.2916,
+      "step": 40804
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.296240323745236e-06,
+      "loss": 0.2229,
+      "step": 40805
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.29590448843887e-06,
+      "loss": 0.2209,
+      "step": 40806
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.295568653132505e-06,
+      "loss": 0.3278,
+      "step": 40807
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.295232817826139e-06,
+      "loss": 0.4244,
+      "step": 40808
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.294896982519772e-06,
+      "loss": 0.221,
+      "step": 40809
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.294561147213407e-06,
+      "loss": 0.1652,
+      "step": 40810
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.294225311907042e-06,
+      "loss": 0.2374,
+      "step": 40811
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2938894766006755e-06,
+      "loss": 0.3026,
+      "step": 40812
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.29355364129431e-06,
+      "loss": 0.2226,
+      "step": 40813
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.293217805987945e-06,
+      "loss": 0.3322,
+      "step": 40814
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.292881970681578e-06,
+      "loss": 0.2484,
+      "step": 40815
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.292546135375212e-06,
+      "loss": 0.2653,
+      "step": 40816
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.292210300068847e-06,
+      "loss": 0.2989,
+      "step": 40817
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.291874464762481e-06,
+      "loss": 0.2329,
+      "step": 40818
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2915386294561155e-06,
+      "loss": 0.134,
+      "step": 40819
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.29120279414975e-06,
+      "loss": 0.3143,
+      "step": 40820
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.290866958843384e-06,
+      "loss": 0.3009,
+      "step": 40821
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.290531123537018e-06,
+      "loss": 0.3212,
+      "step": 40822
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.290195288230652e-06,
+      "loss": 0.2781,
+      "step": 40823
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.289859452924286e-06,
+      "loss": 0.3921,
+      "step": 40824
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.289523617617921e-06,
+      "loss": 0.1887,
+      "step": 40825
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2891877823115555e-06,
+      "loss": 0.3129,
+      "step": 40826
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.288851947005189e-06,
+      "loss": 0.322,
+      "step": 40827
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.288516111698824e-06,
+      "loss": 0.3426,
+      "step": 40828
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2881802763924585e-06,
+      "loss": 0.1109,
+      "step": 40829
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2878444410860915e-06,
+      "loss": 0.1621,
+      "step": 40830
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.287508605779726e-06,
+      "loss": 0.2631,
+      "step": 40831
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.28717277047336e-06,
+      "loss": 0.2748,
+      "step": 40832
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.286836935166995e-06,
+      "loss": 0.2585,
+      "step": 40833
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.286501099860629e-06,
+      "loss": 0.2978,
+      "step": 40834
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.286165264554262e-06,
+      "loss": 0.408,
+      "step": 40835
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.285829429247897e-06,
+      "loss": 0.4292,
+      "step": 40836
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2854935939415315e-06,
+      "loss": 0.2241,
+      "step": 40837
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.285157758635165e-06,
+      "loss": 0.4645,
+      "step": 40838
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2848219233288e-06,
+      "loss": 0.1898,
+      "step": 40839
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.284486088022435e-06,
+      "loss": 0.1527,
+      "step": 40840
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.284150252716068e-06,
+      "loss": 0.2256,
+      "step": 40841
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.283814417409703e-06,
+      "loss": 0.3186,
+      "step": 40842
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.283478582103338e-06,
+      "loss": 0.4414,
+      "step": 40843
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.283142746796971e-06,
+      "loss": 0.1981,
+      "step": 40844
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.282806911490605e-06,
+      "loss": 0.217,
+      "step": 40845
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.28247107618424e-06,
+      "loss": 0.302,
+      "step": 40846
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.282135240877874e-06,
+      "loss": 0.2003,
+      "step": 40847
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.281799405571508e-06,
+      "loss": 0.305,
+      "step": 40848
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.281463570265143e-06,
+      "loss": 0.2014,
+      "step": 40849
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.281127734958776e-06,
+      "loss": 0.2855,
+      "step": 40850
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.280791899652411e-06,
+      "loss": 0.2886,
+      "step": 40851
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.280456064346045e-06,
+      "loss": 0.3876,
+      "step": 40852
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.280120229039679e-06,
+      "loss": 0.3159,
+      "step": 40853
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.279784393733314e-06,
+      "loss": 0.2241,
+      "step": 40854
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.279448558426948e-06,
+      "loss": 0.2186,
+      "step": 40855
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.279112723120581e-06,
+      "loss": 0.2146,
+      "step": 40856
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.278776887814216e-06,
+      "loss": 0.25,
+      "step": 40857
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.278441052507851e-06,
+      "loss": 0.2456,
+      "step": 40858
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.278105217201484e-06,
+      "loss": 0.2324,
+      "step": 40859
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.277769381895119e-06,
+      "loss": 0.3175,
+      "step": 40860
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.277433546588754e-06,
+      "loss": 0.2613,
+      "step": 40861
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2770977112823875e-06,
+      "loss": 0.3971,
+      "step": 40862
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.276761875976022e-06,
+      "loss": 0.1721,
+      "step": 40863
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.276426040669657e-06,
+      "loss": 0.1237,
+      "step": 40864
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.27609020536329e-06,
+      "loss": 0.2263,
+      "step": 40865
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.275754370056924e-06,
+      "loss": 0.3266,
+      "step": 40866
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.275418534750559e-06,
+      "loss": 0.1466,
+      "step": 40867
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.275082699444193e-06,
+      "loss": 0.2342,
+      "step": 40868
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2747468641378275e-06,
+      "loss": 0.2497,
+      "step": 40869
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.274411028831462e-06,
+      "loss": 0.2909,
+      "step": 40870
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.274075193525095e-06,
+      "loss": 0.3068,
+      "step": 40871
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.27373935821873e-06,
+      "loss": 0.2349,
+      "step": 40872
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.273403522912364e-06,
+      "loss": 0.1583,
+      "step": 40873
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.273067687605998e-06,
+      "loss": 0.1855,
+      "step": 40874
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.272731852299633e-06,
+      "loss": 0.1542,
+      "step": 40875
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2723960169932675e-06,
+      "loss": 0.1749,
+      "step": 40876
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.272060181686901e-06,
+      "loss": 0.1857,
+      "step": 40877
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.271724346380535e-06,
+      "loss": 0.3787,
+      "step": 40878
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.27138851107417e-06,
+      "loss": 0.4057,
+      "step": 40879
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2710526757678035e-06,
+      "loss": 0.3401,
+      "step": 40880
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.270716840461438e-06,
+      "loss": 0.3833,
+      "step": 40881
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.270381005155073e-06,
+      "loss": 0.1729,
+      "step": 40882
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.270045169848707e-06,
+      "loss": 0.411,
+      "step": 40883
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.269709334542341e-06,
+      "loss": 0.2714,
+      "step": 40884
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.269373499235976e-06,
+      "loss": 0.2622,
+      "step": 40885
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.269037663929609e-06,
+      "loss": 0.3084,
+      "step": 40886
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2687018286232435e-06,
+      "loss": 0.3838,
+      "step": 40887
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.268365993316878e-06,
+      "loss": 0.2419,
+      "step": 40888
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.268030158010512e-06,
+      "loss": 0.3636,
+      "step": 40889
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.267694322704147e-06,
+      "loss": 0.2851,
+      "step": 40890
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.267358487397781e-06,
+      "loss": 0.2918,
+      "step": 40891
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.267022652091414e-06,
+      "loss": 0.2692,
+      "step": 40892
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.266686816785049e-06,
+      "loss": 0.1583,
+      "step": 40893
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2663509814786835e-06,
+      "loss": 0.214,
+      "step": 40894
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.266015146172317e-06,
+      "loss": 0.2821,
+      "step": 40895
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.265679310865952e-06,
+      "loss": 0.3815,
+      "step": 40896
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.265343475559587e-06,
+      "loss": 0.2111,
+      "step": 40897
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.26500764025322e-06,
+      "loss": 0.2194,
+      "step": 40898
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.264671804946855e-06,
+      "loss": 0.2002,
+      "step": 40899
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.26433596964049e-06,
+      "loss": 0.2289,
+      "step": 40900
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.264000134334123e-06,
+      "loss": 0.2374,
+      "step": 40901
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.263664299027757e-06,
+      "loss": 0.3792,
+      "step": 40902
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.263328463721392e-06,
+      "loss": 0.1639,
+      "step": 40903
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.262992628415026e-06,
+      "loss": 0.3332,
+      "step": 40904
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.26265679310866e-06,
+      "loss": 0.3234,
+      "step": 40905
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.262320957802295e-06,
+      "loss": 0.2409,
+      "step": 40906
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.261985122495928e-06,
+      "loss": 0.262,
+      "step": 40907
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.261649287189563e-06,
+      "loss": 0.2072,
+      "step": 40908
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.261313451883197e-06,
+      "loss": 0.2646,
+      "step": 40909
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.260977616576831e-06,
+      "loss": 0.2574,
+      "step": 40910
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.260641781270466e-06,
+      "loss": 0.1732,
+      "step": 40911
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2603059459641e-06,
+      "loss": 0.1216,
+      "step": 40912
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.259970110657733e-06,
+      "loss": 0.3909,
+      "step": 40913
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.259634275351368e-06,
+      "loss": 0.1705,
+      "step": 40914
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.259298440045003e-06,
+      "loss": 0.2781,
+      "step": 40915
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2589626047386364e-06,
+      "loss": 0.2335,
+      "step": 40916
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.258626769432271e-06,
+      "loss": 0.2745,
+      "step": 40917
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.258290934125906e-06,
+      "loss": 0.3849,
+      "step": 40918
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2579550988195395e-06,
+      "loss": 0.1455,
+      "step": 40919
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.257619263513174e-06,
+      "loss": 0.1719,
+      "step": 40920
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.257283428206807e-06,
+      "loss": 0.2335,
+      "step": 40921
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.256947592900442e-06,
+      "loss": 0.4364,
+      "step": 40922
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2566117575940764e-06,
+      "loss": 0.1866,
+      "step": 40923
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.25627592228771e-06,
+      "loss": 0.3027,
+      "step": 40924
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.255940086981345e-06,
+      "loss": 0.3803,
+      "step": 40925
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2556042516749795e-06,
+      "loss": 0.2297,
+      "step": 40926
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2552684163686125e-06,
+      "loss": 0.2894,
+      "step": 40927
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.254932581062247e-06,
+      "loss": 0.2353,
+      "step": 40928
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.254596745755882e-06,
+      "loss": 0.2695,
+      "step": 40929
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.254260910449516e-06,
+      "loss": 0.3136,
+      "step": 40930
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.25392507514315e-06,
+      "loss": 0.3,
+      "step": 40931
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.253589239836785e-06,
+      "loss": 0.3241,
+      "step": 40932
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.253253404530419e-06,
+      "loss": 0.2047,
+      "step": 40933
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.252917569224053e-06,
+      "loss": 0.1977,
+      "step": 40934
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.252581733917687e-06,
+      "loss": 0.2252,
+      "step": 40935
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.252245898611321e-06,
+      "loss": 0.2867,
+      "step": 40936
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2519100633049556e-06,
+      "loss": 0.2197,
+      "step": 40937
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.25157422799859e-06,
+      "loss": 0.2922,
+      "step": 40938
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.251238392692224e-06,
+      "loss": 0.2745,
+      "step": 40939
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.250902557385859e-06,
+      "loss": 0.2333,
+      "step": 40940
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.250566722079493e-06,
+      "loss": 0.3765,
+      "step": 40941
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.250230886773126e-06,
+      "loss": 0.2278,
+      "step": 40942
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.249895051466761e-06,
+      "loss": 0.3138,
+      "step": 40943
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2495592161603956e-06,
+      "loss": 0.3215,
+      "step": 40944
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.249223380854029e-06,
+      "loss": 0.2198,
+      "step": 40945
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.248887545547664e-06,
+      "loss": 0.1613,
+      "step": 40946
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.248551710241299e-06,
+      "loss": 0.3379,
+      "step": 40947
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.248215874934932e-06,
+      "loss": 0.3741,
+      "step": 40948
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.247880039628566e-06,
+      "loss": 0.1376,
+      "step": 40949
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.247544204322201e-06,
+      "loss": 0.3373,
+      "step": 40950
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.247208369015835e-06,
+      "loss": 0.3033,
+      "step": 40951
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.246872533709469e-06,
+      "loss": 0.3501,
+      "step": 40952
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.246536698403104e-06,
+      "loss": 0.1769,
+      "step": 40953
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.246200863096738e-06,
+      "loss": 0.1611,
+      "step": 40954
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2458650277903724e-06,
+      "loss": 0.1753,
+      "step": 40955
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.245529192484007e-06,
+      "loss": 0.1833,
+      "step": 40956
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.24519335717764e-06,
+      "loss": 0.2152,
+      "step": 40957
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.244857521871275e-06,
+      "loss": 0.1637,
+      "step": 40958
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.244521686564909e-06,
+      "loss": 0.3662,
+      "step": 40959
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.244185851258543e-06,
+      "loss": 0.2037,
+      "step": 40960
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.243850015952178e-06,
+      "loss": 0.1682,
+      "step": 40961
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2435141806458124e-06,
+      "loss": 0.3394,
+      "step": 40962
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.243178345339445e-06,
+      "loss": 0.3036,
+      "step": 40963
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.24284251003308e-06,
+      "loss": 0.2164,
+      "step": 40964
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.242506674726715e-06,
+      "loss": 0.3799,
+      "step": 40965
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2421708394203485e-06,
+      "loss": 0.335,
+      "step": 40966
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.241835004113983e-06,
+      "loss": 0.3144,
+      "step": 40967
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.241499168807618e-06,
+      "loss": 0.2205,
+      "step": 40968
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.241163333501251e-06,
+      "loss": 0.2046,
+      "step": 40969
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.240827498194885e-06,
+      "loss": 0.1714,
+      "step": 40970
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.24049166288852e-06,
+      "loss": 0.1552,
+      "step": 40971
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.240155827582154e-06,
+      "loss": 0.2199,
+      "step": 40972
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2398199922757885e-06,
+      "loss": 0.1318,
+      "step": 40973
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.239484156969423e-06,
+      "loss": 0.4153,
+      "step": 40974
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.239148321663057e-06,
+      "loss": 0.2618,
+      "step": 40975
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2388124863566916e-06,
+      "loss": 0.2234,
+      "step": 40976
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.238476651050326e-06,
+      "loss": 0.3476,
+      "step": 40977
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.238140815743959e-06,
+      "loss": 0.3283,
+      "step": 40978
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.237804980437594e-06,
+      "loss": 0.2971,
+      "step": 40979
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2374691451312285e-06,
+      "loss": 0.1232,
+      "step": 40980
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.237133309824862e-06,
+      "loss": 0.4629,
+      "step": 40981
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.236797474518497e-06,
+      "loss": 0.2245,
+      "step": 40982
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2364616392121316e-06,
+      "loss": 0.3135,
+      "step": 40983
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2361258039057645e-06,
+      "loss": 0.2794,
+      "step": 40984
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.235789968599399e-06,
+      "loss": 0.3791,
+      "step": 40985
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.235454133293034e-06,
+      "loss": 0.3445,
+      "step": 40986
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.235118297986668e-06,
+      "loss": 0.218,
+      "step": 40987
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.234782462680302e-06,
+      "loss": 0.2663,
+      "step": 40988
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.234446627373937e-06,
+      "loss": 0.2389,
+      "step": 40989
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.234110792067571e-06,
+      "loss": 0.258,
+      "step": 40990
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.2337749567612045e-06,
+      "loss": 0.2419,
+      "step": 40991
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 6.233439121454839e-06,
+      "loss": 0.1948,
+      "step": 40992
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.233103286148473e-06,
+      "loss": 0.2789,
+      "step": 40993
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.232767450842108e-06,
+      "loss": 0.3219,
+      "step": 40994
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.232431615535742e-06,
+      "loss": 0.2489,
+      "step": 40995
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.232095780229376e-06,
+      "loss": 0.3893,
+      "step": 40996
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.231759944923011e-06,
+      "loss": 0.3731,
+      "step": 40997
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.231424109616645e-06,
+      "loss": 0.1483,
+      "step": 40998
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.231088274310278e-06,
+      "loss": 0.231,
+      "step": 40999
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.230752439003913e-06,
+      "loss": 0.2321,
+      "step": 41000
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.230416603697548e-06,
+      "loss": 0.3251,
+      "step": 41001
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.230080768391181e-06,
+      "loss": 0.3123,
+      "step": 41002
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.229744933084816e-06,
+      "loss": 0.107,
+      "step": 41003
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.229409097778451e-06,
+      "loss": 0.2626,
+      "step": 41004
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.229073262472084e-06,
+      "loss": 0.2906,
+      "step": 41005
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.228737427165718e-06,
+      "loss": 0.3034,
+      "step": 41006
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.228401591859353e-06,
+      "loss": 0.3478,
+      "step": 41007
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.228065756552987e-06,
+      "loss": 0.4223,
+      "step": 41008
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.227729921246621e-06,
+      "loss": 0.2994,
+      "step": 41009
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.227394085940256e-06,
+      "loss": 0.3622,
+      "step": 41010
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.22705825063389e-06,
+      "loss": 0.1757,
+      "step": 41011
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.2267224153275245e-06,
+      "loss": 0.2813,
+      "step": 41012
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.2263865800211574e-06,
+      "loss": 0.3634,
+      "step": 41013
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.226050744714792e-06,
+      "loss": 0.2297,
+      "step": 41014
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.225714909408427e-06,
+      "loss": 0.3993,
+      "step": 41015
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.2253790741020605e-06,
+      "loss": 0.184,
+      "step": 41016
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.225043238795695e-06,
+      "loss": 0.2866,
+      "step": 41017
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.22470740348933e-06,
+      "loss": 0.4448,
+      "step": 41018
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.224371568182963e-06,
+      "loss": 0.1499,
+      "step": 41019
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.2240357328765974e-06,
+      "loss": 0.2611,
+      "step": 41020
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.223699897570232e-06,
+      "loss": 0.2455,
+      "step": 41021
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.223364062263866e-06,
+      "loss": 0.2957,
+      "step": 41022
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.2230282269575005e-06,
+      "loss": 0.3024,
+      "step": 41023
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.222692391651135e-06,
+      "loss": 0.3602,
+      "step": 41024
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.222356556344768e-06,
+      "loss": 0.3735,
+      "step": 41025
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.222020721038403e-06,
+      "loss": 0.2085,
+      "step": 41026
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.221684885732037e-06,
+      "loss": 0.2373,
+      "step": 41027
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.221349050425671e-06,
+      "loss": 0.2821,
+      "step": 41028
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.221013215119306e-06,
+      "loss": 0.1706,
+      "step": 41029
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.2206773798129405e-06,
+      "loss": 0.3212,
+      "step": 41030
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.220341544506574e-06,
+      "loss": 0.2899,
+      "step": 41031
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.220005709200209e-06,
+      "loss": 0.2097,
+      "step": 41032
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.219669873893844e-06,
+      "loss": 0.3673,
+      "step": 41033
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.2193340385874766e-06,
+      "loss": 0.3102,
+      "step": 41034
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.218998203281111e-06,
+      "loss": 0.2129,
+      "step": 41035
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.218662367974746e-06,
+      "loss": 0.2497,
+      "step": 41036
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.21832653266838e-06,
+      "loss": 0.1546,
+      "step": 41037
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.217990697362014e-06,
+      "loss": 0.3251,
+      "step": 41038
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.217654862055649e-06,
+      "loss": 0.1842,
+      "step": 41039
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.217319026749282e-06,
+      "loss": 0.3238,
+      "step": 41040
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.2169831914429165e-06,
+      "loss": 0.1544,
+      "step": 41041
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.216647356136551e-06,
+      "loss": 0.2911,
+      "step": 41042
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.216311520830185e-06,
+      "loss": 0.1427,
+      "step": 41043
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.21597568552382e-06,
+      "loss": 0.2705,
+      "step": 41044
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.215639850217454e-06,
+      "loss": 0.2046,
+      "step": 41045
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.215304014911088e-06,
+      "loss": 0.2738,
+      "step": 41046
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.214968179604723e-06,
+      "loss": 0.2275,
+      "step": 41047
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.2146323442983565e-06,
+      "loss": 0.4235,
+      "step": 41048
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.21429650899199e-06,
+      "loss": 0.2781,
+      "step": 41049
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.213960673685625e-06,
+      "loss": 0.3084,
+      "step": 41050
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.21362483837926e-06,
+      "loss": 0.2307,
+      "step": 41051
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.2132890030728934e-06,
+      "loss": 0.2478,
+      "step": 41052
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.212953167766528e-06,
+      "loss": 0.4637,
+      "step": 41053
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.212617332460163e-06,
+      "loss": 0.2368,
+      "step": 41054
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.212281497153796e-06,
+      "loss": 0.2478,
+      "step": 41055
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.21194566184743e-06,
+      "loss": 0.2525,
+      "step": 41056
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.211609826541065e-06,
+      "loss": 0.2505,
+      "step": 41057
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.211273991234699e-06,
+      "loss": 0.2167,
+      "step": 41058
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.2109381559283334e-06,
+      "loss": 0.3267,
+      "step": 41059
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.210602320621968e-06,
+      "loss": 0.3334,
+      "step": 41060
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.210266485315601e-06,
+      "loss": 0.4477,
+      "step": 41061
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.209930650009236e-06,
+      "loss": 0.336,
+      "step": 41062
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.20959481470287e-06,
+      "loss": 0.1878,
+      "step": 41063
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.209258979396504e-06,
+      "loss": 0.31,
+      "step": 41064
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.208923144090139e-06,
+      "loss": 0.2248,
+      "step": 41065
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.208587308783773e-06,
+      "loss": 0.1579,
+      "step": 41066
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.208251473477407e-06,
+      "loss": 0.2605,
+      "step": 41067
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.207915638171042e-06,
+      "loss": 0.2162,
+      "step": 41068
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.2075798028646765e-06,
+      "loss": 0.4039,
+      "step": 41069
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.2072439675583095e-06,
+      "loss": 0.247,
+      "step": 41070
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.206908132251944e-06,
+      "loss": 0.2561,
+      "step": 41071
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.206572296945579e-06,
+      "loss": 0.2902,
+      "step": 41072
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.2062364616392126e-06,
+      "loss": 0.1393,
+      "step": 41073
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.205900626332847e-06,
+      "loss": 0.2539,
+      "step": 41074
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.205564791026482e-06,
+      "loss": 0.2606,
+      "step": 41075
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.205228955720115e-06,
+      "loss": 0.2478,
+      "step": 41076
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.2048931204137495e-06,
+      "loss": 0.1755,
+      "step": 41077
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.204557285107384e-06,
+      "loss": 0.1992,
+      "step": 41078
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.204221449801018e-06,
+      "loss": 0.5638,
+      "step": 41079
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.2038856144946525e-06,
+      "loss": 0.4839,
+      "step": 41080
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.203549779188287e-06,
+      "loss": 0.1499,
+      "step": 41081
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.20321394388192e-06,
+      "loss": 0.2054,
+      "step": 41082
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.202878108575555e-06,
+      "loss": 0.3241,
+      "step": 41083
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.2025422732691894e-06,
+      "loss": 0.1485,
+      "step": 41084
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.202206437962823e-06,
+      "loss": 0.2922,
+      "step": 41085
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.201870602656458e-06,
+      "loss": 0.2999,
+      "step": 41086
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.2015347673500925e-06,
+      "loss": 0.3822,
+      "step": 41087
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.201198932043726e-06,
+      "loss": 0.2582,
+      "step": 41088
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.200863096737361e-06,
+      "loss": 0.2971,
+      "step": 41089
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.200527261430996e-06,
+      "loss": 0.1625,
+      "step": 41090
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.200191426124629e-06,
+      "loss": 0.2843,
+      "step": 41091
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.199855590818263e-06,
+      "loss": 0.2018,
+      "step": 41092
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.199519755511898e-06,
+      "loss": 0.3212,
+      "step": 41093
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.199183920205532e-06,
+      "loss": 0.1558,
+      "step": 41094
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.198848084899166e-06,
+      "loss": 0.3509,
+      "step": 41095
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.198512249592801e-06,
+      "loss": 0.2216,
+      "step": 41096
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.198176414286434e-06,
+      "loss": 0.3364,
+      "step": 41097
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.197840578980069e-06,
+      "loss": 0.2922,
+      "step": 41098
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.197504743673703e-06,
+      "loss": 0.3073,
+      "step": 41099
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.197168908367337e-06,
+      "loss": 0.338,
+      "step": 41100
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.196833073060972e-06,
+      "loss": 0.0898,
+      "step": 41101
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.1964972377546055e-06,
+      "loss": 0.332,
+      "step": 41102
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.19616140244824e-06,
+      "loss": 0.205,
+      "step": 41103
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.195825567141874e-06,
+      "loss": 0.2755,
+      "step": 41104
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.195489731835508e-06,
+      "loss": 0.2798,
+      "step": 41105
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.195153896529142e-06,
+      "loss": 0.4025,
+      "step": 41106
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.194818061222777e-06,
+      "loss": 0.2763,
+      "step": 41107
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.194482225916411e-06,
+      "loss": 0.275,
+      "step": 41108
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.1941463906100455e-06,
+      "loss": 0.2847,
+      "step": 41109
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.19381055530368e-06,
+      "loss": 0.352,
+      "step": 41110
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.193474719997313e-06,
+      "loss": 0.3081,
+      "step": 41111
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.193138884690948e-06,
+      "loss": 0.2878,
+      "step": 41112
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.192803049384582e-06,
+      "loss": 0.2988,
+      "step": 41113
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.192467214078216e-06,
+      "loss": 0.1324,
+      "step": 41114
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.192131378771851e-06,
+      "loss": 0.1613,
+      "step": 41115
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.1917955434654855e-06,
+      "loss": 0.2137,
+      "step": 41116
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.191459708159118e-06,
+      "loss": 0.2306,
+      "step": 41117
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.191123872852753e-06,
+      "loss": 0.2268,
+      "step": 41118
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.190788037546388e-06,
+      "loss": 0.2474,
+      "step": 41119
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.1904522022400215e-06,
+      "loss": 0.2678,
+      "step": 41120
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.190116366933656e-06,
+      "loss": 0.3179,
+      "step": 41121
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.189780531627291e-06,
+      "loss": 0.1982,
+      "step": 41122
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.189444696320925e-06,
+      "loss": 0.1953,
+      "step": 41123
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.189108861014559e-06,
+      "loss": 0.2884,
+      "step": 41124
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.188773025708194e-06,
+      "loss": 0.341,
+      "step": 41125
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.188437190401827e-06,
+      "loss": 0.1477,
+      "step": 41126
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.1881013550954615e-06,
+      "loss": 0.1972,
+      "step": 41127
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.187765519789096e-06,
+      "loss": 0.3085,
+      "step": 41128
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.18742968448273e-06,
+      "loss": 0.2834,
+      "step": 41129
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.187093849176365e-06,
+      "loss": 0.1989,
+      "step": 41130
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.186758013869999e-06,
+      "loss": 0.1761,
+      "step": 41131
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.186422178563632e-06,
+      "loss": 0.1737,
+      "step": 41132
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.186086343257267e-06,
+      "loss": 0.4448,
+      "step": 41133
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.1857505079509015e-06,
+      "loss": 0.1948,
+      "step": 41134
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.185414672644535e-06,
+      "loss": 0.3153,
+      "step": 41135
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.18507883733817e-06,
+      "loss": 0.2593,
+      "step": 41136
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.184743002031805e-06,
+      "loss": 0.1462,
+      "step": 41137
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.1844071667254375e-06,
+      "loss": 0.1598,
+      "step": 41138
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.184071331419072e-06,
+      "loss": 0.2875,
+      "step": 41139
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.183735496112707e-06,
+      "loss": 0.1288,
+      "step": 41140
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.183399660806341e-06,
+      "loss": 0.1924,
+      "step": 41141
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.183063825499975e-06,
+      "loss": 0.3688,
+      "step": 41142
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.18272799019361e-06,
+      "loss": 0.4535,
+      "step": 41143
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.182392154887244e-06,
+      "loss": 0.4313,
+      "step": 41144
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.182056319580878e-06,
+      "loss": 0.2299,
+      "step": 41145
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.181720484274513e-06,
+      "loss": 0.3851,
+      "step": 41146
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.181384648968146e-06,
+      "loss": 0.2888,
+      "step": 41147
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.181048813661781e-06,
+      "loss": 0.3023,
+      "step": 41148
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.180712978355415e-06,
+      "loss": 0.3795,
+      "step": 41149
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.180377143049049e-06,
+      "loss": 0.2382,
+      "step": 41150
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.180041307742684e-06,
+      "loss": 0.2244,
+      "step": 41151
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.179705472436318e-06,
+      "loss": 0.3545,
+      "step": 41152
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.179369637129951e-06,
+      "loss": 0.3535,
+      "step": 41153
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.179033801823586e-06,
+      "loss": 0.1263,
+      "step": 41154
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.178697966517221e-06,
+      "loss": 0.3829,
+      "step": 41155
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.178362131210854e-06,
+      "loss": 0.2517,
+      "step": 41156
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.178026295904489e-06,
+      "loss": 0.1473,
+      "step": 41157
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.177690460598124e-06,
+      "loss": 0.1726,
+      "step": 41158
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.1773546252917575e-06,
+      "loss": 0.1715,
+      "step": 41159
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.177018789985391e-06,
+      "loss": 0.3352,
+      "step": 41160
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.176682954679026e-06,
+      "loss": 0.2652,
+      "step": 41161
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.17634711937266e-06,
+      "loss": 0.4251,
+      "step": 41162
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.176011284066294e-06,
+      "loss": 0.3295,
+      "step": 41163
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.175675448759929e-06,
+      "loss": 0.1205,
+      "step": 41164
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.175339613453563e-06,
+      "loss": 0.1459,
+      "step": 41165
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.1750037781471975e-06,
+      "loss": 0.2072,
+      "step": 41166
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.174667942840832e-06,
+      "loss": 0.2489,
+      "step": 41167
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.174332107534465e-06,
+      "loss": 0.359,
+      "step": 41168
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.1739962722281e-06,
+      "loss": 0.2622,
+      "step": 41169
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.173660436921734e-06,
+      "loss": 0.2952,
+      "step": 41170
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.173324601615368e-06,
+      "loss": 0.2543,
+      "step": 41171
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.172988766309003e-06,
+      "loss": 0.2377,
+      "step": 41172
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.1726529310026375e-06,
+      "loss": 0.2411,
+      "step": 41173
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.1723170956962704e-06,
+      "loss": 0.2507,
+      "step": 41174
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.171981260389905e-06,
+      "loss": 0.1928,
+      "step": 41175
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.17164542508354e-06,
+      "loss": 0.3144,
+      "step": 41176
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.1713095897771735e-06,
+      "loss": 0.491,
+      "step": 41177
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.170973754470808e-06,
+      "loss": 0.24,
+      "step": 41178
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.170637919164443e-06,
+      "loss": 0.3603,
+      "step": 41179
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.170302083858077e-06,
+      "loss": 0.2242,
+      "step": 41180
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.169966248551711e-06,
+      "loss": 0.1854,
+      "step": 41181
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.169630413245346e-06,
+      "loss": 0.2211,
+      "step": 41182
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.169294577938979e-06,
+      "loss": 0.2403,
+      "step": 41183
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.1689587426326135e-06,
+      "loss": 0.3742,
+      "step": 41184
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.168622907326248e-06,
+      "loss": 0.3612,
+      "step": 41185
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.168287072019882e-06,
+      "loss": 0.3483,
+      "step": 41186
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.167951236713517e-06,
+      "loss": 0.2567,
+      "step": 41187
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.167615401407151e-06,
+      "loss": 0.2899,
+      "step": 41188
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.167279566100784e-06,
+      "loss": 0.3403,
+      "step": 41189
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 6.166943730794419e-06,
+      "loss": 0.3932,
+      "step": 41190
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.166607895488053e-06,
+      "loss": 0.3218,
+      "step": 41191
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.166272060181687e-06,
+      "loss": 0.4013,
+      "step": 41192
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.165936224875322e-06,
+      "loss": 0.1707,
+      "step": 41193
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.165600389568956e-06,
+      "loss": 0.2247,
+      "step": 41194
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1652645542625896e-06,
+      "loss": 0.2041,
+      "step": 41195
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.164928718956224e-06,
+      "loss": 0.3262,
+      "step": 41196
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.164592883649858e-06,
+      "loss": 0.2291,
+      "step": 41197
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.164257048343493e-06,
+      "loss": 0.3967,
+      "step": 41198
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.163921213037127e-06,
+      "loss": 0.2362,
+      "step": 41199
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.163585377730761e-06,
+      "loss": 0.2426,
+      "step": 41200
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.163249542424396e-06,
+      "loss": 0.271,
+      "step": 41201
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.16291370711803e-06,
+      "loss": 0.2652,
+      "step": 41202
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.162577871811663e-06,
+      "loss": 0.2704,
+      "step": 41203
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.162242036505298e-06,
+      "loss": 0.2584,
+      "step": 41204
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.161906201198933e-06,
+      "loss": 0.2185,
+      "step": 41205
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1615703658925665e-06,
+      "loss": 0.3821,
+      "step": 41206
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.161234530586201e-06,
+      "loss": 0.2565,
+      "step": 41207
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.160898695279836e-06,
+      "loss": 0.3261,
+      "step": 41208
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.160562859973469e-06,
+      "loss": 0.2476,
+      "step": 41209
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.160227024667103e-06,
+      "loss": 0.2873,
+      "step": 41210
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.159891189360738e-06,
+      "loss": 0.3337,
+      "step": 41211
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.159555354054372e-06,
+      "loss": 0.2069,
+      "step": 41212
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1592195187480064e-06,
+      "loss": 0.3338,
+      "step": 41213
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.158883683441641e-06,
+      "loss": 0.2006,
+      "step": 41214
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.158547848135275e-06,
+      "loss": 0.1937,
+      "step": 41215
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1582120128289095e-06,
+      "loss": 0.2418,
+      "step": 41216
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.157876177522543e-06,
+      "loss": 0.2378,
+      "step": 41217
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.157540342216177e-06,
+      "loss": 0.2186,
+      "step": 41218
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.157204506909812e-06,
+      "loss": 0.3364,
+      "step": 41219
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1568686716034464e-06,
+      "loss": 0.2542,
+      "step": 41220
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.15653283629708e-06,
+      "loss": 0.2221,
+      "step": 41221
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.156197000990715e-06,
+      "loss": 0.3041,
+      "step": 41222
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1558611656843495e-06,
+      "loss": 0.2314,
+      "step": 41223
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1555253303779825e-06,
+      "loss": 0.3176,
+      "step": 41224
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.155189495071617e-06,
+      "loss": 0.2673,
+      "step": 41225
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.154853659765252e-06,
+      "loss": 0.1889,
+      "step": 41226
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1545178244588856e-06,
+      "loss": 0.4336,
+      "step": 41227
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.15418198915252e-06,
+      "loss": 0.4442,
+      "step": 41228
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.153846153846155e-06,
+      "loss": 0.2327,
+      "step": 41229
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.153510318539788e-06,
+      "loss": 0.3046,
+      "step": 41230
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1531744832334225e-06,
+      "loss": 0.4228,
+      "step": 41231
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.152838647927057e-06,
+      "loss": 0.3261,
+      "step": 41232
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.152502812620691e-06,
+      "loss": 0.2513,
+      "step": 41233
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1521669773143256e-06,
+      "loss": 0.2816,
+      "step": 41234
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.15183114200796e-06,
+      "loss": 0.2726,
+      "step": 41235
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.151495306701594e-06,
+      "loss": 0.2825,
+      "step": 41236
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.151159471395229e-06,
+      "loss": 0.3153,
+      "step": 41237
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.150823636088863e-06,
+      "loss": 0.2406,
+      "step": 41238
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.150487800782496e-06,
+      "loss": 0.181,
+      "step": 41239
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.150151965476131e-06,
+      "loss": 0.2859,
+      "step": 41240
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1498161301697656e-06,
+      "loss": 0.2199,
+      "step": 41241
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.149480294863399e-06,
+      "loss": 0.3671,
+      "step": 41242
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.149144459557034e-06,
+      "loss": 0.3557,
+      "step": 41243
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.148808624250669e-06,
+      "loss": 0.2658,
+      "step": 41244
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.148472788944302e-06,
+      "loss": 0.3696,
+      "step": 41245
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.148136953637936e-06,
+      "loss": 0.3818,
+      "step": 41246
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.147801118331571e-06,
+      "loss": 0.3537,
+      "step": 41247
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.147465283025205e-06,
+      "loss": 0.4136,
+      "step": 41248
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.147129447718839e-06,
+      "loss": 0.2782,
+      "step": 41249
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.146793612412474e-06,
+      "loss": 0.3014,
+      "step": 41250
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.146457777106107e-06,
+      "loss": 0.2552,
+      "step": 41251
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.146121941799742e-06,
+      "loss": 0.3855,
+      "step": 41252
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.145786106493376e-06,
+      "loss": 0.4421,
+      "step": 41253
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.14545027118701e-06,
+      "loss": 0.195,
+      "step": 41254
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.145114435880645e-06,
+      "loss": 0.2467,
+      "step": 41255
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.144778600574279e-06,
+      "loss": 0.1665,
+      "step": 41256
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.144442765267913e-06,
+      "loss": 0.257,
+      "step": 41257
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.144106929961548e-06,
+      "loss": 0.2324,
+      "step": 41258
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1437710946551824e-06,
+      "loss": 0.3961,
+      "step": 41259
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.143435259348815e-06,
+      "loss": 0.1144,
+      "step": 41260
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.14309942404245e-06,
+      "loss": 0.4337,
+      "step": 41261
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.142763588736085e-06,
+      "loss": 0.2079,
+      "step": 41262
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1424277534297185e-06,
+      "loss": 0.228,
+      "step": 41263
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.142091918123353e-06,
+      "loss": 0.3076,
+      "step": 41264
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.141756082816988e-06,
+      "loss": 0.2549,
+      "step": 41265
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.141420247510621e-06,
+      "loss": 0.2993,
+      "step": 41266
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.141084412204255e-06,
+      "loss": 0.2124,
+      "step": 41267
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.14074857689789e-06,
+      "loss": 0.183,
+      "step": 41268
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.140412741591524e-06,
+      "loss": 0.2881,
+      "step": 41269
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1400769062851585e-06,
+      "loss": 0.3563,
+      "step": 41270
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.139741070978793e-06,
+      "loss": 0.2942,
+      "step": 41271
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.139405235672427e-06,
+      "loss": 0.3591,
+      "step": 41272
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.139069400366061e-06,
+      "loss": 0.166,
+      "step": 41273
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.138733565059695e-06,
+      "loss": 0.1276,
+      "step": 41274
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.138397729753329e-06,
+      "loss": 0.2204,
+      "step": 41275
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.138061894446964e-06,
+      "loss": 0.2208,
+      "step": 41276
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1377260591405985e-06,
+      "loss": 0.3638,
+      "step": 41277
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.137390223834232e-06,
+      "loss": 0.2103,
+      "step": 41278
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.137054388527867e-06,
+      "loss": 0.3203,
+      "step": 41279
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1367185532215016e-06,
+      "loss": 0.3001,
+      "step": 41280
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1363827179151345e-06,
+      "loss": 0.2001,
+      "step": 41281
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.136046882608769e-06,
+      "loss": 0.1778,
+      "step": 41282
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.135711047302403e-06,
+      "loss": 0.2002,
+      "step": 41283
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.135375211996038e-06,
+      "loss": 0.4138,
+      "step": 41284
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.135039376689672e-06,
+      "loss": 0.1946,
+      "step": 41285
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.134703541383305e-06,
+      "loss": 0.3062,
+      "step": 41286
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.13436770607694e-06,
+      "loss": 0.2649,
+      "step": 41287
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1340318707705745e-06,
+      "loss": 0.3455,
+      "step": 41288
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.133696035464208e-06,
+      "loss": 0.3553,
+      "step": 41289
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.133360200157843e-06,
+      "loss": 0.2978,
+      "step": 41290
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.133024364851478e-06,
+      "loss": 0.2101,
+      "step": 41291
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.132688529545111e-06,
+      "loss": 0.1818,
+      "step": 41292
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.132352694238746e-06,
+      "loss": 0.304,
+      "step": 41293
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.132016858932381e-06,
+      "loss": 0.414,
+      "step": 41294
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.131681023626014e-06,
+      "loss": 0.304,
+      "step": 41295
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.131345188319648e-06,
+      "loss": 0.2148,
+      "step": 41296
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.131009353013283e-06,
+      "loss": 0.2375,
+      "step": 41297
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.130673517706917e-06,
+      "loss": 0.2023,
+      "step": 41298
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.130337682400551e-06,
+      "loss": 0.2357,
+      "step": 41299
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.130001847094186e-06,
+      "loss": 0.1593,
+      "step": 41300
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.129666011787819e-06,
+      "loss": 0.3277,
+      "step": 41301
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.129330176481454e-06,
+      "loss": 0.2985,
+      "step": 41302
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.128994341175088e-06,
+      "loss": 0.3386,
+      "step": 41303
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.128658505868722e-06,
+      "loss": 0.2328,
+      "step": 41304
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.128322670562357e-06,
+      "loss": 0.196,
+      "step": 41305
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.127986835255991e-06,
+      "loss": 0.3044,
+      "step": 41306
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.127650999949625e-06,
+      "loss": 0.2032,
+      "step": 41307
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.127315164643259e-06,
+      "loss": 0.1965,
+      "step": 41308
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.126979329336894e-06,
+      "loss": 0.2777,
+      "step": 41309
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1266434940305274e-06,
+      "loss": 0.2489,
+      "step": 41310
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.126307658724162e-06,
+      "loss": 0.2015,
+      "step": 41311
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.125971823417797e-06,
+      "loss": 0.2124,
+      "step": 41312
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1256359881114305e-06,
+      "loss": 0.2477,
+      "step": 41313
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.125300152805065e-06,
+      "loss": 0.5758,
+      "step": 41314
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1249643174987e-06,
+      "loss": 0.2162,
+      "step": 41315
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.124628482192333e-06,
+      "loss": 0.2194,
+      "step": 41316
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.124292646885967e-06,
+      "loss": 0.2747,
+      "step": 41317
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.123956811579602e-06,
+      "loss": 0.2936,
+      "step": 41318
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.123620976273236e-06,
+      "loss": 0.2275,
+      "step": 41319
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1232851409668705e-06,
+      "loss": 0.21,
+      "step": 41320
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.122949305660505e-06,
+      "loss": 0.2393,
+      "step": 41321
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.122613470354138e-06,
+      "loss": 0.239,
+      "step": 41322
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.122277635047773e-06,
+      "loss": 0.4131,
+      "step": 41323
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.121941799741407e-06,
+      "loss": 0.2423,
+      "step": 41324
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.121605964435041e-06,
+      "loss": 0.3316,
+      "step": 41325
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.121270129128676e-06,
+      "loss": 0.3539,
+      "step": 41326
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1209342938223105e-06,
+      "loss": 0.4068,
+      "step": 41327
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.120598458515944e-06,
+      "loss": 0.1962,
+      "step": 41328
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.120262623209579e-06,
+      "loss": 0.1467,
+      "step": 41329
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.119926787903213e-06,
+      "loss": 0.3128,
+      "step": 41330
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1195909525968466e-06,
+      "loss": 0.2735,
+      "step": 41331
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.119255117290481e-06,
+      "loss": 0.2249,
+      "step": 41332
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.118919281984116e-06,
+      "loss": 0.2568,
+      "step": 41333
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.11858344667775e-06,
+      "loss": 0.337,
+      "step": 41334
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.118247611371384e-06,
+      "loss": 0.3377,
+      "step": 41335
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.117911776065019e-06,
+      "loss": 0.3514,
+      "step": 41336
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.117575940758652e-06,
+      "loss": 0.3589,
+      "step": 41337
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1172401054522865e-06,
+      "loss": 0.3153,
+      "step": 41338
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.116904270145921e-06,
+      "loss": 0.212,
+      "step": 41339
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.116568434839555e-06,
+      "loss": 0.2255,
+      "step": 41340
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.11623259953319e-06,
+      "loss": 0.3512,
+      "step": 41341
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.115896764226824e-06,
+      "loss": 0.1197,
+      "step": 41342
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.115560928920457e-06,
+      "loss": 0.2755,
+      "step": 41343
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.115225093614092e-06,
+      "loss": 0.2089,
+      "step": 41344
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1148892583077265e-06,
+      "loss": 0.1932,
+      "step": 41345
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.11455342300136e-06,
+      "loss": 0.257,
+      "step": 41346
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.114217587694995e-06,
+      "loss": 0.2156,
+      "step": 41347
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.11388175238863e-06,
+      "loss": 0.2929,
+      "step": 41348
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1135459170822634e-06,
+      "loss": 0.2542,
+      "step": 41349
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.113210081775898e-06,
+      "loss": 0.2283,
+      "step": 41350
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.112874246469533e-06,
+      "loss": 0.2212,
+      "step": 41351
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.112538411163166e-06,
+      "loss": 0.239,
+      "step": 41352
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1122025758568e-06,
+      "loss": 0.2367,
+      "step": 41353
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.111866740550435e-06,
+      "loss": 0.1667,
+      "step": 41354
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.111530905244069e-06,
+      "loss": 0.2547,
+      "step": 41355
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.111195069937703e-06,
+      "loss": 0.3223,
+      "step": 41356
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.110859234631338e-06,
+      "loss": 0.192,
+      "step": 41357
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.110523399324971e-06,
+      "loss": 0.233,
+      "step": 41358
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.110187564018606e-06,
+      "loss": 0.2149,
+      "step": 41359
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.10985172871224e-06,
+      "loss": 0.3623,
+      "step": 41360
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.109515893405874e-06,
+      "loss": 0.4048,
+      "step": 41361
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.109180058099509e-06,
+      "loss": 0.193,
+      "step": 41362
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.108844222793143e-06,
+      "loss": 0.2705,
+      "step": 41363
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.108508387486776e-06,
+      "loss": 0.241,
+      "step": 41364
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.108172552180411e-06,
+      "loss": 0.1275,
+      "step": 41365
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.107836716874046e-06,
+      "loss": 0.3253,
+      "step": 41366
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1075008815676795e-06,
+      "loss": 0.157,
+      "step": 41367
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.107165046261314e-06,
+      "loss": 0.2505,
+      "step": 41368
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.106829210954949e-06,
+      "loss": 0.2359,
+      "step": 41369
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1064933756485826e-06,
+      "loss": 0.2228,
+      "step": 41370
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.106157540342217e-06,
+      "loss": 0.3101,
+      "step": 41371
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.10582170503585e-06,
+      "loss": 0.1532,
+      "step": 41372
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.105485869729485e-06,
+      "loss": 0.3103,
+      "step": 41373
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1051500344231194e-06,
+      "loss": 0.2595,
+      "step": 41374
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.104814199116753e-06,
+      "loss": 0.1356,
+      "step": 41375
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.104478363810388e-06,
+      "loss": 0.274,
+      "step": 41376
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1041425285040225e-06,
+      "loss": 0.2405,
+      "step": 41377
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.1038066931976555e-06,
+      "loss": 0.4259,
+      "step": 41378
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.10347085789129e-06,
+      "loss": 0.1977,
+      "step": 41379
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.103135022584925e-06,
+      "loss": 0.2557,
+      "step": 41380
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.102799187278559e-06,
+      "loss": 0.3048,
+      "step": 41381
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.102463351972193e-06,
+      "loss": 0.3119,
+      "step": 41382
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.102127516665828e-06,
+      "loss": 0.3666,
+      "step": 41383
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.101791681359462e-06,
+      "loss": 0.5337,
+      "step": 41384
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.101455846053096e-06,
+      "loss": 0.2305,
+      "step": 41385
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.10112001074673e-06,
+      "loss": 0.1041,
+      "step": 41386
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.100784175440364e-06,
+      "loss": 0.2161,
+      "step": 41387
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.100448340133999e-06,
+      "loss": 0.2778,
+      "step": 41388
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 6.100112504827633e-06,
+      "loss": 0.1964,
+      "step": 41389
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.099776669521267e-06,
+      "loss": 0.1557,
+      "step": 41390
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.099440834214902e-06,
+      "loss": 0.1472,
+      "step": 41391
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.099104998908536e-06,
+      "loss": 0.1671,
+      "step": 41392
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.098769163602169e-06,
+      "loss": 0.2395,
+      "step": 41393
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.098433328295804e-06,
+      "loss": 0.267,
+      "step": 41394
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0980974929894386e-06,
+      "loss": 0.3832,
+      "step": 41395
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.097761657683072e-06,
+      "loss": 0.1716,
+      "step": 41396
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.097425822376707e-06,
+      "loss": 0.2411,
+      "step": 41397
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.097089987070342e-06,
+      "loss": 0.2479,
+      "step": 41398
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.096754151763975e-06,
+      "loss": 0.4228,
+      "step": 41399
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.096418316457609e-06,
+      "loss": 0.2448,
+      "step": 41400
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.096082481151244e-06,
+      "loss": 0.3828,
+      "step": 41401
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.095746645844878e-06,
+      "loss": 0.3271,
+      "step": 41402
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.095410810538512e-06,
+      "loss": 0.1668,
+      "step": 41403
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.095074975232147e-06,
+      "loss": 0.1381,
+      "step": 41404
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.094739139925781e-06,
+      "loss": 0.2986,
+      "step": 41405
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0944033046194155e-06,
+      "loss": 0.2948,
+      "step": 41406
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.09406746931305e-06,
+      "loss": 0.1816,
+      "step": 41407
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.093731634006683e-06,
+      "loss": 0.2212,
+      "step": 41408
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.093395798700318e-06,
+      "loss": 0.3472,
+      "step": 41409
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.093059963393952e-06,
+      "loss": 0.2978,
+      "step": 41410
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.092724128087586e-06,
+      "loss": 0.1256,
+      "step": 41411
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.092388292781221e-06,
+      "loss": 0.1784,
+      "step": 41412
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0920524574748555e-06,
+      "loss": 0.3418,
+      "step": 41413
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.091716622168488e-06,
+      "loss": 0.2531,
+      "step": 41414
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.091380786862123e-06,
+      "loss": 0.2402,
+      "step": 41415
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.091044951555758e-06,
+      "loss": 0.2786,
+      "step": 41416
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0907091162493915e-06,
+      "loss": 0.4187,
+      "step": 41417
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.090373280943026e-06,
+      "loss": 0.1955,
+      "step": 41418
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.090037445636661e-06,
+      "loss": 0.3443,
+      "step": 41419
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.089701610330294e-06,
+      "loss": 0.3331,
+      "step": 41420
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.089365775023928e-06,
+      "loss": 0.4491,
+      "step": 41421
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.089029939717563e-06,
+      "loss": 0.3406,
+      "step": 41422
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.088694104411197e-06,
+      "loss": 0.2283,
+      "step": 41423
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0883582691048315e-06,
+      "loss": 0.2555,
+      "step": 41424
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.088022433798466e-06,
+      "loss": 0.4088,
+      "step": 41425
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0876865984921e-06,
+      "loss": 0.2396,
+      "step": 41426
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.087350763185735e-06,
+      "loss": 0.3224,
+      "step": 41427
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.087014927879369e-06,
+      "loss": 0.1942,
+      "step": 41428
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.086679092573002e-06,
+      "loss": 0.4177,
+      "step": 41429
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.086343257266637e-06,
+      "loss": 0.2476,
+      "step": 41430
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0860074219602715e-06,
+      "loss": 0.431,
+      "step": 41431
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.085671586653905e-06,
+      "loss": 0.1882,
+      "step": 41432
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.08533575134754e-06,
+      "loss": 0.2631,
+      "step": 41433
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0849999160411746e-06,
+      "loss": 0.3152,
+      "step": 41434
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0846640807348075e-06,
+      "loss": 0.2088,
+      "step": 41435
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.084328245428442e-06,
+      "loss": 0.1334,
+      "step": 41436
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.083992410122077e-06,
+      "loss": 0.4318,
+      "step": 41437
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.083656574815711e-06,
+      "loss": 0.2098,
+      "step": 41438
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.083320739509345e-06,
+      "loss": 0.2961,
+      "step": 41439
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.08298490420298e-06,
+      "loss": 0.2676,
+      "step": 41440
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.082649068896614e-06,
+      "loss": 0.3129,
+      "step": 41441
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.082313233590248e-06,
+      "loss": 0.2954,
+      "step": 41442
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.081977398283882e-06,
+      "loss": 0.2711,
+      "step": 41443
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.081641562977516e-06,
+      "loss": 0.2786,
+      "step": 41444
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.081305727671151e-06,
+      "loss": 0.3396,
+      "step": 41445
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.080969892364785e-06,
+      "loss": 0.1833,
+      "step": 41446
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.080634057058419e-06,
+      "loss": 0.2417,
+      "step": 41447
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.080298221752054e-06,
+      "loss": 0.2454,
+      "step": 41448
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.079962386445688e-06,
+      "loss": 0.2833,
+      "step": 41449
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.079626551139321e-06,
+      "loss": 0.2683,
+      "step": 41450
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.079290715832956e-06,
+      "loss": 0.1915,
+      "step": 41451
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.078954880526591e-06,
+      "loss": 0.256,
+      "step": 41452
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.078619045220224e-06,
+      "loss": 0.2889,
+      "step": 41453
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.078283209913859e-06,
+      "loss": 0.2382,
+      "step": 41454
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.077947374607494e-06,
+      "loss": 0.2274,
+      "step": 41455
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.077611539301127e-06,
+      "loss": 0.0651,
+      "step": 41456
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.077275703994761e-06,
+      "loss": 0.2838,
+      "step": 41457
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.076939868688396e-06,
+      "loss": 0.1508,
+      "step": 41458
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.07660403338203e-06,
+      "loss": 0.2892,
+      "step": 41459
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.076268198075664e-06,
+      "loss": 0.1701,
+      "step": 41460
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.075932362769299e-06,
+      "loss": 0.254,
+      "step": 41461
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.075596527462933e-06,
+      "loss": 0.1343,
+      "step": 41462
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0752606921565675e-06,
+      "loss": 0.2087,
+      "step": 41463
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0749248568502004e-06,
+      "loss": 0.2909,
+      "step": 41464
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.074589021543835e-06,
+      "loss": 0.2145,
+      "step": 41465
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.07425318623747e-06,
+      "loss": 0.3207,
+      "step": 41466
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0739173509311035e-06,
+      "loss": 0.365,
+      "step": 41467
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.073581515624738e-06,
+      "loss": 0.1963,
+      "step": 41468
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.073245680318373e-06,
+      "loss": 0.277,
+      "step": 41469
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.072909845012006e-06,
+      "loss": 0.1736,
+      "step": 41470
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0725740097056404e-06,
+      "loss": 0.3045,
+      "step": 41471
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.072238174399275e-06,
+      "loss": 0.2463,
+      "step": 41472
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.071902339092909e-06,
+      "loss": 0.3759,
+      "step": 41473
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0715665037865435e-06,
+      "loss": 0.2065,
+      "step": 41474
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.071230668480178e-06,
+      "loss": 0.3603,
+      "step": 41475
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.070894833173812e-06,
+      "loss": 0.3961,
+      "step": 41476
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.070558997867446e-06,
+      "loss": 0.2857,
+      "step": 41477
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0702231625610804e-06,
+      "loss": 0.1683,
+      "step": 41478
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.069887327254714e-06,
+      "loss": 0.2562,
+      "step": 41479
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.069551491948349e-06,
+      "loss": 0.3913,
+      "step": 41480
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0692156566419835e-06,
+      "loss": 0.1641,
+      "step": 41481
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.068879821335617e-06,
+      "loss": 0.2994,
+      "step": 41482
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.068543986029252e-06,
+      "loss": 0.2361,
+      "step": 41483
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.068208150722887e-06,
+      "loss": 0.2171,
+      "step": 41484
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0678723154165196e-06,
+      "loss": 0.1283,
+      "step": 41485
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.067536480110154e-06,
+      "loss": 0.4108,
+      "step": 41486
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.067200644803789e-06,
+      "loss": 0.2639,
+      "step": 41487
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.066864809497423e-06,
+      "loss": 0.3828,
+      "step": 41488
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.066528974191057e-06,
+      "loss": 0.152,
+      "step": 41489
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.066193138884692e-06,
+      "loss": 0.1951,
+      "step": 41490
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.065857303578325e-06,
+      "loss": 0.2386,
+      "step": 41491
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0655214682719596e-06,
+      "loss": 0.1867,
+      "step": 41492
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.065185632965594e-06,
+      "loss": 0.2868,
+      "step": 41493
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.064849797659228e-06,
+      "loss": 0.207,
+      "step": 41494
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.064513962352863e-06,
+      "loss": 0.3051,
+      "step": 41495
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.064178127046497e-06,
+      "loss": 0.1935,
+      "step": 41496
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.063842291740131e-06,
+      "loss": 0.6477,
+      "step": 41497
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.063506456433766e-06,
+      "loss": 0.2307,
+      "step": 41498
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0631706211273996e-06,
+      "loss": 0.2785,
+      "step": 41499
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.062834785821033e-06,
+      "loss": 0.1295,
+      "step": 41500
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.062498950514668e-06,
+      "loss": 0.3021,
+      "step": 41501
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.062163115208303e-06,
+      "loss": 0.148,
+      "step": 41502
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0618272799019364e-06,
+      "loss": 0.2045,
+      "step": 41503
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.061491444595571e-06,
+      "loss": 0.3859,
+      "step": 41504
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.061155609289206e-06,
+      "loss": 0.2285,
+      "step": 41505
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.060819773982839e-06,
+      "loss": 0.2043,
+      "step": 41506
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.060483938676473e-06,
+      "loss": 0.363,
+      "step": 41507
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.060148103370108e-06,
+      "loss": 0.3887,
+      "step": 41508
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.059812268063742e-06,
+      "loss": 0.2394,
+      "step": 41509
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0594764327573764e-06,
+      "loss": 0.2211,
+      "step": 41510
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.059140597451011e-06,
+      "loss": 0.3051,
+      "step": 41511
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.058804762144644e-06,
+      "loss": 0.1378,
+      "step": 41512
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.058468926838279e-06,
+      "loss": 0.4555,
+      "step": 41513
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.058133091531913e-06,
+      "loss": 0.2251,
+      "step": 41514
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.057797256225547e-06,
+      "loss": 0.3775,
+      "step": 41515
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.057461420919182e-06,
+      "loss": 0.4205,
+      "step": 41516
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0571255856128164e-06,
+      "loss": 0.2247,
+      "step": 41517
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.05678975030645e-06,
+      "loss": 0.2613,
+      "step": 41518
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.056453915000085e-06,
+      "loss": 0.4332,
+      "step": 41519
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0561180796937195e-06,
+      "loss": 0.3047,
+      "step": 41520
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0557822443873525e-06,
+      "loss": 0.276,
+      "step": 41521
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.055446409080987e-06,
+      "loss": 0.272,
+      "step": 41522
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.055110573774622e-06,
+      "loss": 0.1866,
+      "step": 41523
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0547747384682556e-06,
+      "loss": 0.3662,
+      "step": 41524
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.05443890316189e-06,
+      "loss": 0.177,
+      "step": 41525
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.054103067855525e-06,
+      "loss": 0.1644,
+      "step": 41526
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.053767232549158e-06,
+      "loss": 0.4059,
+      "step": 41527
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0534313972427925e-06,
+      "loss": 0.2597,
+      "step": 41528
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.053095561936427e-06,
+      "loss": 0.2557,
+      "step": 41529
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.052759726630061e-06,
+      "loss": 0.3368,
+      "step": 41530
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0524238913236956e-06,
+      "loss": 0.3199,
+      "step": 41531
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.05208805601733e-06,
+      "loss": 0.3964,
+      "step": 41532
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.051752220710963e-06,
+      "loss": 0.165,
+      "step": 41533
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.051416385404598e-06,
+      "loss": 0.2359,
+      "step": 41534
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0510805500982325e-06,
+      "loss": 0.1815,
+      "step": 41535
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.050744714791866e-06,
+      "loss": 0.3637,
+      "step": 41536
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.050408879485501e-06,
+      "loss": 0.2626,
+      "step": 41537
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0500730441791356e-06,
+      "loss": 0.2752,
+      "step": 41538
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.049737208872769e-06,
+      "loss": 0.1547,
+      "step": 41539
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.049401373566404e-06,
+      "loss": 0.1632,
+      "step": 41540
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.049065538260039e-06,
+      "loss": 0.1475,
+      "step": 41541
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.048729702953672e-06,
+      "loss": 0.2167,
+      "step": 41542
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.048393867647306e-06,
+      "loss": 0.3654,
+      "step": 41543
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.048058032340941e-06,
+      "loss": 0.2557,
+      "step": 41544
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.047722197034575e-06,
+      "loss": 0.3195,
+      "step": 41545
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.047386361728209e-06,
+      "loss": 0.2657,
+      "step": 41546
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.047050526421844e-06,
+      "loss": 0.1886,
+      "step": 41547
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.046714691115477e-06,
+      "loss": 0.2985,
+      "step": 41548
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.046378855809112e-06,
+      "loss": 0.3511,
+      "step": 41549
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.046043020502746e-06,
+      "loss": 0.4622,
+      "step": 41550
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.04570718519638e-06,
+      "loss": 0.1857,
+      "step": 41551
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.045371349890015e-06,
+      "loss": 0.2701,
+      "step": 41552
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0450355145836485e-06,
+      "loss": 0.2368,
+      "step": 41553
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.044699679277283e-06,
+      "loss": 0.2136,
+      "step": 41554
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.044363843970917e-06,
+      "loss": 0.2375,
+      "step": 41555
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.044028008664551e-06,
+      "loss": 0.2349,
+      "step": 41556
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.043692173358185e-06,
+      "loss": 0.2124,
+      "step": 41557
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.04335633805182e-06,
+      "loss": 0.3115,
+      "step": 41558
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.043020502745454e-06,
+      "loss": 0.1545,
+      "step": 41559
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0426846674390885e-06,
+      "loss": 0.1285,
+      "step": 41560
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.042348832132723e-06,
+      "loss": 0.2512,
+      "step": 41561
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.042012996826356e-06,
+      "loss": 0.3494,
+      "step": 41562
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.041677161519991e-06,
+      "loss": 0.2184,
+      "step": 41563
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.041341326213625e-06,
+      "loss": 0.1983,
+      "step": 41564
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.041005490907259e-06,
+      "loss": 0.3607,
+      "step": 41565
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.040669655600894e-06,
+      "loss": 0.341,
+      "step": 41566
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0403338202945285e-06,
+      "loss": 0.295,
+      "step": 41567
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.039997984988161e-06,
+      "loss": 0.2754,
+      "step": 41568
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.039662149681796e-06,
+      "loss": 0.3641,
+      "step": 41569
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.039326314375431e-06,
+      "loss": 0.4284,
+      "step": 41570
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0389904790690645e-06,
+      "loss": 0.4994,
+      "step": 41571
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.038654643762699e-06,
+      "loss": 0.1896,
+      "step": 41572
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.038318808456334e-06,
+      "loss": 0.2349,
+      "step": 41573
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.037982973149968e-06,
+      "loss": 0.1957,
+      "step": 41574
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.037647137843602e-06,
+      "loss": 0.365,
+      "step": 41575
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.037311302537237e-06,
+      "loss": 0.269,
+      "step": 41576
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.03697546723087e-06,
+      "loss": 0.1819,
+      "step": 41577
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0366396319245045e-06,
+      "loss": 0.194,
+      "step": 41578
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.036303796618139e-06,
+      "loss": 0.1935,
+      "step": 41579
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.035967961311773e-06,
+      "loss": 0.1562,
+      "step": 41580
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.035632126005408e-06,
+      "loss": 0.3069,
+      "step": 41581
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.035296290699042e-06,
+      "loss": 0.2684,
+      "step": 41582
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.034960455392675e-06,
+      "loss": 0.3418,
+      "step": 41583
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.03462462008631e-06,
+      "loss": 0.3396,
+      "step": 41584
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.0342887847799445e-06,
+      "loss": 0.3312,
+      "step": 41585
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.033952949473578e-06,
+      "loss": 0.2156,
+      "step": 41586
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 6.033617114167213e-06,
+      "loss": 0.2037,
+      "step": 41587
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.033281278860848e-06,
+      "loss": 0.3355,
+      "step": 41588
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.032945443554481e-06,
+      "loss": 0.201,
+      "step": 41589
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.032609608248115e-06,
+      "loss": 0.2484,
+      "step": 41590
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.03227377294175e-06,
+      "loss": 0.2694,
+      "step": 41591
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.031937937635384e-06,
+      "loss": 0.2981,
+      "step": 41592
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.031602102329018e-06,
+      "loss": 0.2905,
+      "step": 41593
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.031266267022653e-06,
+      "loss": 0.3933,
+      "step": 41594
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.030930431716287e-06,
+      "loss": 0.1964,
+      "step": 41595
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.030594596409921e-06,
+      "loss": 0.233,
+      "step": 41596
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.030258761103556e-06,
+      "loss": 0.1402,
+      "step": 41597
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.029922925797189e-06,
+      "loss": 0.1642,
+      "step": 41598
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.029587090490824e-06,
+      "loss": 0.2138,
+      "step": 41599
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.029251255184458e-06,
+      "loss": 0.3043,
+      "step": 41600
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.028915419878092e-06,
+      "loss": 0.2874,
+      "step": 41601
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.028579584571727e-06,
+      "loss": 0.325,
+      "step": 41602
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.028243749265361e-06,
+      "loss": 0.3655,
+      "step": 41603
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.027907913958994e-06,
+      "loss": 0.2196,
+      "step": 41604
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.027572078652629e-06,
+      "loss": 0.3285,
+      "step": 41605
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.027236243346264e-06,
+      "loss": 0.3054,
+      "step": 41606
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.026900408039897e-06,
+      "loss": 0.2778,
+      "step": 41607
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.026564572733532e-06,
+      "loss": 0.146,
+      "step": 41608
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.026228737427167e-06,
+      "loss": 0.2301,
+      "step": 41609
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.0258929021208005e-06,
+      "loss": 0.1835,
+      "step": 41610
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.025557066814435e-06,
+      "loss": 0.5044,
+      "step": 41611
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.025221231508069e-06,
+      "loss": 0.2452,
+      "step": 41612
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.024885396201703e-06,
+      "loss": 0.2783,
+      "step": 41613
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.024549560895337e-06,
+      "loss": 0.1722,
+      "step": 41614
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.024213725588972e-06,
+      "loss": 0.2975,
+      "step": 41615
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.023877890282606e-06,
+      "loss": 0.2591,
+      "step": 41616
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.0235420549762405e-06,
+      "loss": 0.2005,
+      "step": 41617
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.023206219669875e-06,
+      "loss": 0.2174,
+      "step": 41618
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.022870384363508e-06,
+      "loss": 0.1725,
+      "step": 41619
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.022534549057143e-06,
+      "loss": 0.4132,
+      "step": 41620
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.022198713750777e-06,
+      "loss": 0.1676,
+      "step": 41621
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.021862878444411e-06,
+      "loss": 0.2143,
+      "step": 41622
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.021527043138046e-06,
+      "loss": 0.2053,
+      "step": 41623
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.0211912078316805e-06,
+      "loss": 0.3399,
+      "step": 41624
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.0208553725253135e-06,
+      "loss": 0.3417,
+      "step": 41625
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.020519537218948e-06,
+      "loss": 0.2719,
+      "step": 41626
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.020183701912583e-06,
+      "loss": 0.27,
+      "step": 41627
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.0198478666062165e-06,
+      "loss": 0.1575,
+      "step": 41628
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.019512031299851e-06,
+      "loss": 0.2694,
+      "step": 41629
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.019176195993486e-06,
+      "loss": 0.3022,
+      "step": 41630
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.01884036068712e-06,
+      "loss": 0.2235,
+      "step": 41631
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.018504525380754e-06,
+      "loss": 0.3127,
+      "step": 41632
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.018168690074389e-06,
+      "loss": 0.2238,
+      "step": 41633
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.017832854768022e-06,
+      "loss": 0.2532,
+      "step": 41634
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.0174970194616565e-06,
+      "loss": 0.2175,
+      "step": 41635
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.017161184155291e-06,
+      "loss": 0.4318,
+      "step": 41636
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.016825348848925e-06,
+      "loss": 0.1626,
+      "step": 41637
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.01648951354256e-06,
+      "loss": 0.2828,
+      "step": 41638
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.016153678236194e-06,
+      "loss": 0.2412,
+      "step": 41639
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.015817842929827e-06,
+      "loss": 0.2398,
+      "step": 41640
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.015482007623462e-06,
+      "loss": 0.3253,
+      "step": 41641
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.015146172317096e-06,
+      "loss": 0.2375,
+      "step": 41642
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.01481033701073e-06,
+      "loss": 0.2581,
+      "step": 41643
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.014474501704365e-06,
+      "loss": 0.3389,
+      "step": 41644
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.014138666397999e-06,
+      "loss": 0.2118,
+      "step": 41645
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.013802831091633e-06,
+      "loss": 0.3488,
+      "step": 41646
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.013466995785267e-06,
+      "loss": 0.22,
+      "step": 41647
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.013131160478901e-06,
+      "loss": 0.327,
+      "step": 41648
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.012795325172536e-06,
+      "loss": 0.294,
+      "step": 41649
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.01245948986617e-06,
+      "loss": 0.1864,
+      "step": 41650
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.012123654559804e-06,
+      "loss": 0.274,
+      "step": 41651
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.011787819253439e-06,
+      "loss": 0.2377,
+      "step": 41652
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.011451983947073e-06,
+      "loss": 0.201,
+      "step": 41653
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.011116148640706e-06,
+      "loss": 0.2279,
+      "step": 41654
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.010780313334341e-06,
+      "loss": 0.2797,
+      "step": 41655
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.010444478027976e-06,
+      "loss": 0.2536,
+      "step": 41656
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.0101086427216095e-06,
+      "loss": 0.154,
+      "step": 41657
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.009772807415244e-06,
+      "loss": 0.2808,
+      "step": 41658
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.009436972108879e-06,
+      "loss": 0.2752,
+      "step": 41659
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.009101136802512e-06,
+      "loss": 0.2641,
+      "step": 41660
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.008765301496146e-06,
+      "loss": 0.1281,
+      "step": 41661
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.008429466189781e-06,
+      "loss": 0.2224,
+      "step": 41662
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.008093630883415e-06,
+      "loss": 0.3404,
+      "step": 41663
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.0077577955770495e-06,
+      "loss": 0.2515,
+      "step": 41664
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.007421960270684e-06,
+      "loss": 0.3862,
+      "step": 41665
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.007086124964318e-06,
+      "loss": 0.2127,
+      "step": 41666
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.0067502896579525e-06,
+      "loss": 0.2605,
+      "step": 41667
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.006414454351586e-06,
+      "loss": 0.4688,
+      "step": 41668
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.00607861904522e-06,
+      "loss": 0.2122,
+      "step": 41669
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.005742783738855e-06,
+      "loss": 0.3244,
+      "step": 41670
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.0054069484324894e-06,
+      "loss": 0.2815,
+      "step": 41671
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.005071113126123e-06,
+      "loss": 0.4572,
+      "step": 41672
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.004735277819758e-06,
+      "loss": 0.3865,
+      "step": 41673
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.0043994425133925e-06,
+      "loss": 0.2469,
+      "step": 41674
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.0040636072070255e-06,
+      "loss": 0.2582,
+      "step": 41675
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.00372777190066e-06,
+      "loss": 0.2341,
+      "step": 41676
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.003391936594295e-06,
+      "loss": 0.4839,
+      "step": 41677
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.003056101287929e-06,
+      "loss": 0.4317,
+      "step": 41678
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.002720265981563e-06,
+      "loss": 0.4835,
+      "step": 41679
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.002384430675198e-06,
+      "loss": 0.1784,
+      "step": 41680
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.002048595368831e-06,
+      "loss": 0.3419,
+      "step": 41681
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.0017127600624655e-06,
+      "loss": 0.2269,
+      "step": 41682
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.0013769247561e-06,
+      "loss": 0.1625,
+      "step": 41683
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.001041089449734e-06,
+      "loss": 0.3865,
+      "step": 41684
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.000705254143369e-06,
+      "loss": 0.127,
+      "step": 41685
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.000369418837003e-06,
+      "loss": 0.2501,
+      "step": 41686
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 6.000033583530637e-06,
+      "loss": 0.212,
+      "step": 41687
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.999697748224272e-06,
+      "loss": 0.172,
+      "step": 41688
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.999361912917906e-06,
+      "loss": 0.2183,
+      "step": 41689
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.999026077611539e-06,
+      "loss": 0.4188,
+      "step": 41690
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.998690242305174e-06,
+      "loss": 0.4607,
+      "step": 41691
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.9983544069988086e-06,
+      "loss": 0.2382,
+      "step": 41692
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.998018571692442e-06,
+      "loss": 0.1447,
+      "step": 41693
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.997682736386077e-06,
+      "loss": 0.2486,
+      "step": 41694
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.997346901079712e-06,
+      "loss": 0.3166,
+      "step": 41695
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.997011065773345e-06,
+      "loss": 0.2429,
+      "step": 41696
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.996675230466979e-06,
+      "loss": 0.3118,
+      "step": 41697
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.996339395160614e-06,
+      "loss": 0.1652,
+      "step": 41698
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.996003559854248e-06,
+      "loss": 0.2333,
+      "step": 41699
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.995667724547882e-06,
+      "loss": 0.4422,
+      "step": 41700
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.995331889241517e-06,
+      "loss": 0.2062,
+      "step": 41701
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.99499605393515e-06,
+      "loss": 0.2317,
+      "step": 41702
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.994660218628785e-06,
+      "loss": 0.273,
+      "step": 41703
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.994324383322419e-06,
+      "loss": 0.2684,
+      "step": 41704
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.993988548016053e-06,
+      "loss": 0.462,
+      "step": 41705
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.993652712709688e-06,
+      "loss": 0.3343,
+      "step": 41706
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.993316877403322e-06,
+      "loss": 0.3233,
+      "step": 41707
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.992981042096956e-06,
+      "loss": 0.3101,
+      "step": 41708
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.992645206790591e-06,
+      "loss": 0.2882,
+      "step": 41709
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.9923093714842254e-06,
+      "loss": 0.182,
+      "step": 41710
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.991973536177858e-06,
+      "loss": 0.1841,
+      "step": 41711
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.991637700871493e-06,
+      "loss": 0.1624,
+      "step": 41712
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.991301865565128e-06,
+      "loss": 0.38,
+      "step": 41713
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.9909660302587615e-06,
+      "loss": 0.3261,
+      "step": 41714
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.990630194952396e-06,
+      "loss": 0.2119,
+      "step": 41715
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.990294359646031e-06,
+      "loss": 0.3229,
+      "step": 41716
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.989958524339664e-06,
+      "loss": 0.2723,
+      "step": 41717
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.989622689033298e-06,
+      "loss": 0.2372,
+      "step": 41718
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.989286853726933e-06,
+      "loss": 0.2705,
+      "step": 41719
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.988951018420567e-06,
+      "loss": 0.1975,
+      "step": 41720
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.9886151831142015e-06,
+      "loss": 0.2192,
+      "step": 41721
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.988279347807836e-06,
+      "loss": 0.3678,
+      "step": 41722
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.98794351250147e-06,
+      "loss": 0.1049,
+      "step": 41723
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.987607677195105e-06,
+      "loss": 0.158,
+      "step": 41724
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.987271841888738e-06,
+      "loss": 0.1914,
+      "step": 41725
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.986936006582372e-06,
+      "loss": 0.2375,
+      "step": 41726
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.986600171276007e-06,
+      "loss": 0.236,
+      "step": 41727
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.9862643359696415e-06,
+      "loss": 0.2133,
+      "step": 41728
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.985928500663275e-06,
+      "loss": 0.168,
+      "step": 41729
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.98559266535691e-06,
+      "loss": 0.2003,
+      "step": 41730
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.9852568300505446e-06,
+      "loss": 0.3761,
+      "step": 41731
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.9849209947441775e-06,
+      "loss": 0.2351,
+      "step": 41732
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.984585159437812e-06,
+      "loss": 0.3557,
+      "step": 41733
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.984249324131446e-06,
+      "loss": 0.3194,
+      "step": 41734
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.983913488825081e-06,
+      "loss": 0.2015,
+      "step": 41735
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.983577653518715e-06,
+      "loss": 0.2789,
+      "step": 41736
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.983241818212348e-06,
+      "loss": 0.4255,
+      "step": 41737
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.982905982905983e-06,
+      "loss": 0.2517,
+      "step": 41738
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.9825701475996175e-06,
+      "loss": 0.2198,
+      "step": 41739
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.982234312293251e-06,
+      "loss": 0.2352,
+      "step": 41740
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.981898476986886e-06,
+      "loss": 0.2716,
+      "step": 41741
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.981562641680521e-06,
+      "loss": 0.197,
+      "step": 41742
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.981226806374154e-06,
+      "loss": 0.1943,
+      "step": 41743
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.980890971067789e-06,
+      "loss": 0.2424,
+      "step": 41744
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.980555135761424e-06,
+      "loss": 0.4451,
+      "step": 41745
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.980219300455057e-06,
+      "loss": 0.2976,
+      "step": 41746
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.979883465148691e-06,
+      "loss": 0.2665,
+      "step": 41747
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.979547629842326e-06,
+      "loss": 0.2293,
+      "step": 41748
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.97921179453596e-06,
+      "loss": 0.1289,
+      "step": 41749
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.978875959229594e-06,
+      "loss": 0.2482,
+      "step": 41750
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.978540123923229e-06,
+      "loss": 0.3038,
+      "step": 41751
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.978204288616862e-06,
+      "loss": 0.2313,
+      "step": 41752
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.977868453310497e-06,
+      "loss": 0.3853,
+      "step": 41753
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.977532618004131e-06,
+      "loss": 0.5529,
+      "step": 41754
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.977196782697765e-06,
+      "loss": 0.1856,
+      "step": 41755
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.9768609473914e-06,
+      "loss": 0.1738,
+      "step": 41756
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.976525112085034e-06,
+      "loss": 0.2862,
+      "step": 41757
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.976189276778668e-06,
+      "loss": 0.2399,
+      "step": 41758
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.975853441472302e-06,
+      "loss": 0.2825,
+      "step": 41759
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.975517606165937e-06,
+      "loss": 0.3051,
+      "step": 41760
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.9751817708595704e-06,
+      "loss": 0.384,
+      "step": 41761
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.974845935553205e-06,
+      "loss": 0.3763,
+      "step": 41762
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.97451010024684e-06,
+      "loss": 0.1323,
+      "step": 41763
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.9741742649404735e-06,
+      "loss": 0.5394,
+      "step": 41764
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.973838429634108e-06,
+      "loss": 0.3022,
+      "step": 41765
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.973502594327743e-06,
+      "loss": 0.29,
+      "step": 41766
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.973166759021376e-06,
+      "loss": 0.3462,
+      "step": 41767
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.9728309237150104e-06,
+      "loss": 0.3759,
+      "step": 41768
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.972495088408645e-06,
+      "loss": 0.2498,
+      "step": 41769
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.972159253102279e-06,
+      "loss": 0.3249,
+      "step": 41770
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.9718234177959135e-06,
+      "loss": 0.3496,
+      "step": 41771
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.971487582489548e-06,
+      "loss": 0.319,
+      "step": 41772
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.971151747183181e-06,
+      "loss": 0.2825,
+      "step": 41773
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.970815911876816e-06,
+      "loss": 0.3336,
+      "step": 41774
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.97048007657045e-06,
+      "loss": 0.1889,
+      "step": 41775
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.970144241264084e-06,
+      "loss": 0.3239,
+      "step": 41776
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.969808405957719e-06,
+      "loss": 0.2255,
+      "step": 41777
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.9694725706513535e-06,
+      "loss": 0.2152,
+      "step": 41778
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.969136735344987e-06,
+      "loss": 0.2316,
+      "step": 41779
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.968800900038622e-06,
+      "loss": 0.1966,
+      "step": 41780
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.968465064732256e-06,
+      "loss": 0.3506,
+      "step": 41781
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.9681292294258896e-06,
+      "loss": 0.1417,
+      "step": 41782
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.967793394119524e-06,
+      "loss": 0.2103,
+      "step": 41783
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.967457558813159e-06,
+      "loss": 0.2922,
+      "step": 41784
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.967121723506793e-06,
+      "loss": 0.2163,
+      "step": 41785
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 5.966785888200427e-06,
+      "loss": 0.2458,
+      "step": 41786
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.966450052894062e-06,
+      "loss": 0.1465,
+      "step": 41787
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.966114217587695e-06,
+      "loss": 0.2997,
+      "step": 41788
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9657783822813296e-06,
+      "loss": 0.2261,
+      "step": 41789
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.965442546974964e-06,
+      "loss": 0.2421,
+      "step": 41790
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.965106711668598e-06,
+      "loss": 0.2288,
+      "step": 41791
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.964770876362233e-06,
+      "loss": 0.201,
+      "step": 41792
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.964435041055867e-06,
+      "loss": 0.2204,
+      "step": 41793
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9640992057495e-06,
+      "loss": 0.1797,
+      "step": 41794
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.963763370443135e-06,
+      "loss": 0.177,
+      "step": 41795
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9634275351367695e-06,
+      "loss": 0.4144,
+      "step": 41796
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.963091699830403e-06,
+      "loss": 0.3084,
+      "step": 41797
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.962755864524038e-06,
+      "loss": 0.1961,
+      "step": 41798
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.962420029217673e-06,
+      "loss": 0.157,
+      "step": 41799
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9620841939113064e-06,
+      "loss": 0.3587,
+      "step": 41800
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.961748358604941e-06,
+      "loss": 0.3884,
+      "step": 41801
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.961412523298576e-06,
+      "loss": 0.384,
+      "step": 41802
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.961076687992209e-06,
+      "loss": 0.3671,
+      "step": 41803
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.960740852685843e-06,
+      "loss": 0.4544,
+      "step": 41804
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.960405017379478e-06,
+      "loss": 0.2749,
+      "step": 41805
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.960069182073112e-06,
+      "loss": 0.1765,
+      "step": 41806
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9597333467667464e-06,
+      "loss": 0.2851,
+      "step": 41807
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.959397511460381e-06,
+      "loss": 0.2632,
+      "step": 41808
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.959061676154014e-06,
+      "loss": 0.1109,
+      "step": 41809
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.958725840847649e-06,
+      "loss": 0.2426,
+      "step": 41810
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.958390005541283e-06,
+      "loss": 0.3331,
+      "step": 41811
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.958054170234917e-06,
+      "loss": 0.2929,
+      "step": 41812
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.957718334928552e-06,
+      "loss": 0.2846,
+      "step": 41813
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.957382499622186e-06,
+      "loss": 0.3775,
+      "step": 41814
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.957046664315819e-06,
+      "loss": 0.3266,
+      "step": 41815
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.956710829009454e-06,
+      "loss": 0.1589,
+      "step": 41816
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.956374993703089e-06,
+      "loss": 0.1263,
+      "step": 41817
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9560391583967225e-06,
+      "loss": 0.2915,
+      "step": 41818
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.955703323090357e-06,
+      "loss": 0.3261,
+      "step": 41819
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.955367487783992e-06,
+      "loss": 0.3508,
+      "step": 41820
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9550316524776256e-06,
+      "loss": 0.3728,
+      "step": 41821
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.95469581717126e-06,
+      "loss": 0.3158,
+      "step": 41822
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.954359981864893e-06,
+      "loss": 0.22,
+      "step": 41823
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.954024146558528e-06,
+      "loss": 0.2803,
+      "step": 41824
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9536883112521625e-06,
+      "loss": 0.1497,
+      "step": 41825
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.953352475945796e-06,
+      "loss": 0.2846,
+      "step": 41826
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.953016640639431e-06,
+      "loss": 0.3016,
+      "step": 41827
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9526808053330656e-06,
+      "loss": 0.2549,
+      "step": 41828
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9523449700266985e-06,
+      "loss": 0.2732,
+      "step": 41829
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.952009134720333e-06,
+      "loss": 0.1507,
+      "step": 41830
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.951673299413968e-06,
+      "loss": 0.155,
+      "step": 41831
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.951337464107602e-06,
+      "loss": 0.2732,
+      "step": 41832
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.951001628801236e-06,
+      "loss": 0.2236,
+      "step": 41833
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.950665793494871e-06,
+      "loss": 0.2786,
+      "step": 41834
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.950329958188505e-06,
+      "loss": 0.2411,
+      "step": 41835
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.949994122882139e-06,
+      "loss": 0.2065,
+      "step": 41836
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.949658287575774e-06,
+      "loss": 0.2224,
+      "step": 41837
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.949322452269407e-06,
+      "loss": 0.3106,
+      "step": 41838
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.948986616963042e-06,
+      "loss": 0.2302,
+      "step": 41839
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.948650781656676e-06,
+      "loss": 0.3221,
+      "step": 41840
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.94831494635031e-06,
+      "loss": 0.1786,
+      "step": 41841
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.947979111043945e-06,
+      "loss": 0.2834,
+      "step": 41842
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.947643275737579e-06,
+      "loss": 0.247,
+      "step": 41843
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.947307440431212e-06,
+      "loss": 0.2066,
+      "step": 41844
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.946971605124847e-06,
+      "loss": 0.2921,
+      "step": 41845
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.946635769818482e-06,
+      "loss": 0.2517,
+      "step": 41846
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.946299934512115e-06,
+      "loss": 0.2314,
+      "step": 41847
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.94596409920575e-06,
+      "loss": 0.2872,
+      "step": 41848
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.945628263899385e-06,
+      "loss": 0.2909,
+      "step": 41849
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.945292428593018e-06,
+      "loss": 0.2001,
+      "step": 41850
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.944956593286652e-06,
+      "loss": 0.3507,
+      "step": 41851
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.944620757980287e-06,
+      "loss": 0.4865,
+      "step": 41852
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.944284922673921e-06,
+      "loss": 0.2344,
+      "step": 41853
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.943949087367555e-06,
+      "loss": 0.3399,
+      "step": 41854
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.94361325206119e-06,
+      "loss": 0.288,
+      "step": 41855
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.943277416754824e-06,
+      "loss": 0.2029,
+      "step": 41856
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9429415814484585e-06,
+      "loss": 0.3144,
+      "step": 41857
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.942605746142093e-06,
+      "loss": 0.2313,
+      "step": 41858
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.942269910835726e-06,
+      "loss": 0.3318,
+      "step": 41859
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.941934075529361e-06,
+      "loss": 0.3702,
+      "step": 41860
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.941598240222995e-06,
+      "loss": 0.1955,
+      "step": 41861
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.941262404916629e-06,
+      "loss": 0.1752,
+      "step": 41862
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.940926569610264e-06,
+      "loss": 0.2978,
+      "step": 41863
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9405907343038985e-06,
+      "loss": 0.251,
+      "step": 41864
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.940254898997531e-06,
+      "loss": 0.1945,
+      "step": 41865
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.939919063691166e-06,
+      "loss": 0.2397,
+      "step": 41866
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.939583228384801e-06,
+      "loss": 0.1738,
+      "step": 41867
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9392473930784345e-06,
+      "loss": 0.2531,
+      "step": 41868
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.938911557772069e-06,
+      "loss": 0.2928,
+      "step": 41869
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.938575722465704e-06,
+      "loss": 0.4039,
+      "step": 41870
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.938239887159338e-06,
+      "loss": 0.2882,
+      "step": 41871
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.937904051852971e-06,
+      "loss": 0.2157,
+      "step": 41872
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.937568216546606e-06,
+      "loss": 0.2806,
+      "step": 41873
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.93723238124024e-06,
+      "loss": 0.2334,
+      "step": 41874
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9368965459338745e-06,
+      "loss": 0.2682,
+      "step": 41875
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.936560710627509e-06,
+      "loss": 0.3113,
+      "step": 41876
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.936224875321143e-06,
+      "loss": 0.2295,
+      "step": 41877
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.935889040014778e-06,
+      "loss": 0.1235,
+      "step": 41878
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.935553204708412e-06,
+      "loss": 0.2504,
+      "step": 41879
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.935217369402045e-06,
+      "loss": 0.2091,
+      "step": 41880
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.93488153409568e-06,
+      "loss": 0.4305,
+      "step": 41881
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9345456987893145e-06,
+      "loss": 0.2889,
+      "step": 41882
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.934209863482948e-06,
+      "loss": 0.2652,
+      "step": 41883
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.933874028176583e-06,
+      "loss": 0.3915,
+      "step": 41884
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.933538192870218e-06,
+      "loss": 0.2025,
+      "step": 41885
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9332023575638505e-06,
+      "loss": 0.2393,
+      "step": 41886
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.932866522257485e-06,
+      "loss": 0.324,
+      "step": 41887
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.93253068695112e-06,
+      "loss": 0.2263,
+      "step": 41888
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.932194851644754e-06,
+      "loss": 0.4062,
+      "step": 41889
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.931859016338388e-06,
+      "loss": 0.3801,
+      "step": 41890
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.931523181032023e-06,
+      "loss": 0.2676,
+      "step": 41891
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.931187345725657e-06,
+      "loss": 0.1794,
+      "step": 41892
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.930851510419291e-06,
+      "loss": 0.2768,
+      "step": 41893
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.930515675112925e-06,
+      "loss": 0.1502,
+      "step": 41894
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.930179839806559e-06,
+      "loss": 0.2556,
+      "step": 41895
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.929844004500194e-06,
+      "loss": 0.2387,
+      "step": 41896
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.929508169193828e-06,
+      "loss": 0.365,
+      "step": 41897
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.929172333887462e-06,
+      "loss": 0.3452,
+      "step": 41898
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.928836498581097e-06,
+      "loss": 0.3566,
+      "step": 41899
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.928500663274731e-06,
+      "loss": 0.2049,
+      "step": 41900
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.928164827968364e-06,
+      "loss": 0.3939,
+      "step": 41901
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.927828992661999e-06,
+      "loss": 0.2525,
+      "step": 41902
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.927493157355634e-06,
+      "loss": 0.2416,
+      "step": 41903
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.927157322049267e-06,
+      "loss": 0.2908,
+      "step": 41904
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.926821486742902e-06,
+      "loss": 0.2337,
+      "step": 41905
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.926485651436537e-06,
+      "loss": 0.41,
+      "step": 41906
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.92614981613017e-06,
+      "loss": 0.3269,
+      "step": 41907
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.925813980823804e-06,
+      "loss": 0.2625,
+      "step": 41908
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.925478145517439e-06,
+      "loss": 0.2445,
+      "step": 41909
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.925142310211073e-06,
+      "loss": 0.2755,
+      "step": 41910
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.924806474904707e-06,
+      "loss": 0.2535,
+      "step": 41911
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.924470639598342e-06,
+      "loss": 0.1974,
+      "step": 41912
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.924134804291976e-06,
+      "loss": 0.207,
+      "step": 41913
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9237989689856105e-06,
+      "loss": 0.3496,
+      "step": 41914
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9234631336792435e-06,
+      "loss": 0.1866,
+      "step": 41915
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.923127298372878e-06,
+      "loss": 0.2953,
+      "step": 41916
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.922791463066513e-06,
+      "loss": 0.211,
+      "step": 41917
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9224556277601466e-06,
+      "loss": 0.1248,
+      "step": 41918
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.922119792453781e-06,
+      "loss": 0.2719,
+      "step": 41919
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.921783957147416e-06,
+      "loss": 0.3103,
+      "step": 41920
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.921448121841049e-06,
+      "loss": 0.3439,
+      "step": 41921
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9211122865346834e-06,
+      "loss": 0.1906,
+      "step": 41922
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.920776451228318e-06,
+      "loss": 0.363,
+      "step": 41923
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.920440615921952e-06,
+      "loss": 0.2245,
+      "step": 41924
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9201047806155865e-06,
+      "loss": 0.3471,
+      "step": 41925
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.919768945309221e-06,
+      "loss": 0.2693,
+      "step": 41926
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.919433110002855e-06,
+      "loss": 0.212,
+      "step": 41927
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.919097274696489e-06,
+      "loss": 0.5308,
+      "step": 41928
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9187614393901234e-06,
+      "loss": 0.3083,
+      "step": 41929
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.918425604083757e-06,
+      "loss": 0.2155,
+      "step": 41930
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.918089768777392e-06,
+      "loss": 0.2444,
+      "step": 41931
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9177539334710265e-06,
+      "loss": 0.2801,
+      "step": 41932
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.91741809816466e-06,
+      "loss": 0.3955,
+      "step": 41933
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.917082262858295e-06,
+      "loss": 0.3205,
+      "step": 41934
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.91674642755193e-06,
+      "loss": 0.3582,
+      "step": 41935
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.916410592245563e-06,
+      "loss": 0.2165,
+      "step": 41936
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.916074756939197e-06,
+      "loss": 0.3162,
+      "step": 41937
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.915738921632832e-06,
+      "loss": 0.2426,
+      "step": 41938
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.915403086326466e-06,
+      "loss": 0.2365,
+      "step": 41939
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9150672510201e-06,
+      "loss": 0.2898,
+      "step": 41940
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.914731415713735e-06,
+      "loss": 0.2742,
+      "step": 41941
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.914395580407368e-06,
+      "loss": 0.2958,
+      "step": 41942
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9140597451010026e-06,
+      "loss": 0.3032,
+      "step": 41943
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.913723909794637e-06,
+      "loss": 0.1864,
+      "step": 41944
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.913388074488271e-06,
+      "loss": 0.3016,
+      "step": 41945
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.913052239181906e-06,
+      "loss": 0.1472,
+      "step": 41946
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.91271640387554e-06,
+      "loss": 0.2447,
+      "step": 41947
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.912380568569174e-06,
+      "loss": 0.3813,
+      "step": 41948
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.912044733262809e-06,
+      "loss": 0.2084,
+      "step": 41949
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9117088979564426e-06,
+      "loss": 0.2705,
+      "step": 41950
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.911373062650076e-06,
+      "loss": 0.1871,
+      "step": 41951
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.911037227343711e-06,
+      "loss": 0.2189,
+      "step": 41952
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.910701392037346e-06,
+      "loss": 0.2357,
+      "step": 41953
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9103655567309795e-06,
+      "loss": 0.1964,
+      "step": 41954
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.910029721424614e-06,
+      "loss": 0.2416,
+      "step": 41955
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.909693886118249e-06,
+      "loss": 0.2976,
+      "step": 41956
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.909358050811882e-06,
+      "loss": 0.2789,
+      "step": 41957
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.909022215505516e-06,
+      "loss": 0.4095,
+      "step": 41958
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.908686380199151e-06,
+      "loss": 0.1803,
+      "step": 41959
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.908350544892785e-06,
+      "loss": 0.2093,
+      "step": 41960
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9080147095864194e-06,
+      "loss": 0.3785,
+      "step": 41961
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.907678874280054e-06,
+      "loss": 0.192,
+      "step": 41962
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.907343038973687e-06,
+      "loss": 0.2179,
+      "step": 41963
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.907007203667322e-06,
+      "loss": 0.2888,
+      "step": 41964
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.906671368360956e-06,
+      "loss": 0.235,
+      "step": 41965
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.90633553305459e-06,
+      "loss": 0.265,
+      "step": 41966
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.905999697748225e-06,
+      "loss": 0.283,
+      "step": 41967
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9056638624418594e-06,
+      "loss": 0.3084,
+      "step": 41968
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.905328027135493e-06,
+      "loss": 0.3096,
+      "step": 41969
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.904992191829128e-06,
+      "loss": 0.3204,
+      "step": 41970
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9046563565227625e-06,
+      "loss": 0.2951,
+      "step": 41971
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9043205212163955e-06,
+      "loss": 0.3414,
+      "step": 41972
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.90398468591003e-06,
+      "loss": 0.3518,
+      "step": 41973
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.903648850603665e-06,
+      "loss": 0.2185,
+      "step": 41974
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.903313015297299e-06,
+      "loss": 0.2249,
+      "step": 41975
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.902977179990933e-06,
+      "loss": 0.2036,
+      "step": 41976
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.902641344684568e-06,
+      "loss": 0.4104,
+      "step": 41977
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.902305509378201e-06,
+      "loss": 0.2349,
+      "step": 41978
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9019696740718355e-06,
+      "loss": 0.0962,
+      "step": 41979
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.90163383876547e-06,
+      "loss": 0.3158,
+      "step": 41980
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.901298003459104e-06,
+      "loss": 0.2038,
+      "step": 41981
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.9009621681527386e-06,
+      "loss": 0.2399,
+      "step": 41982
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.900626332846373e-06,
+      "loss": 0.4208,
+      "step": 41983
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 5.900290497540007e-06,
+      "loss": 0.4081,
+      "step": 41984
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.899954662233641e-06,
+      "loss": 0.2385,
+      "step": 41985
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8996188269272755e-06,
+      "loss": 0.1976,
+      "step": 41986
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.899282991620909e-06,
+      "loss": 0.4534,
+      "step": 41987
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.898947156314544e-06,
+      "loss": 0.1255,
+      "step": 41988
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8986113210081786e-06,
+      "loss": 0.2106,
+      "step": 41989
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.898275485701812e-06,
+      "loss": 0.2834,
+      "step": 41990
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.897939650395447e-06,
+      "loss": 0.1677,
+      "step": 41991
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.897603815089082e-06,
+      "loss": 0.3131,
+      "step": 41992
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.897267979782715e-06,
+      "loss": 0.1956,
+      "step": 41993
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.896932144476349e-06,
+      "loss": 0.3359,
+      "step": 41994
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.896596309169984e-06,
+      "loss": 0.3317,
+      "step": 41995
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.896260473863618e-06,
+      "loss": 0.2783,
+      "step": 41996
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.895924638557252e-06,
+      "loss": 0.379,
+      "step": 41997
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.895588803250887e-06,
+      "loss": 0.2244,
+      "step": 41998
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.89525296794452e-06,
+      "loss": 0.2365,
+      "step": 41999
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.894917132638155e-06,
+      "loss": 0.3249,
+      "step": 42000
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.894581297331789e-06,
+      "loss": 0.351,
+      "step": 42001
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.894245462025423e-06,
+      "loss": 0.2139,
+      "step": 42002
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.893909626719058e-06,
+      "loss": 0.1893,
+      "step": 42003
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8935737914126915e-06,
+      "loss": 0.1858,
+      "step": 42004
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.893237956106326e-06,
+      "loss": 0.2881,
+      "step": 42005
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.892902120799961e-06,
+      "loss": 0.2218,
+      "step": 42006
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.892566285493594e-06,
+      "loss": 0.2202,
+      "step": 42007
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.892230450187228e-06,
+      "loss": 0.2327,
+      "step": 42008
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.891894614880863e-06,
+      "loss": 0.3854,
+      "step": 42009
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.891558779574497e-06,
+      "loss": 0.2715,
+      "step": 42010
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8912229442681315e-06,
+      "loss": 0.2068,
+      "step": 42011
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.890887108961766e-06,
+      "loss": 0.1237,
+      "step": 42012
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.890551273655399e-06,
+      "loss": 0.2332,
+      "step": 42013
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.890215438349034e-06,
+      "loss": 0.1349,
+      "step": 42014
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.889879603042668e-06,
+      "loss": 0.2406,
+      "step": 42015
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.889543767736302e-06,
+      "loss": 0.2937,
+      "step": 42016
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.889207932429937e-06,
+      "loss": 0.2289,
+      "step": 42017
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8888720971235715e-06,
+      "loss": 0.2594,
+      "step": 42018
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8885362618172044e-06,
+      "loss": 0.3279,
+      "step": 42019
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.888200426510839e-06,
+      "loss": 0.2798,
+      "step": 42020
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.887864591204474e-06,
+      "loss": 0.3202,
+      "step": 42021
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8875287558981075e-06,
+      "loss": 0.4702,
+      "step": 42022
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.887192920591742e-06,
+      "loss": 0.2814,
+      "step": 42023
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.886857085285377e-06,
+      "loss": 0.1604,
+      "step": 42024
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.886521249979011e-06,
+      "loss": 0.2186,
+      "step": 42025
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.886185414672645e-06,
+      "loss": 0.2672,
+      "step": 42026
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.88584957936628e-06,
+      "loss": 0.1565,
+      "step": 42027
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.885513744059913e-06,
+      "loss": 0.2627,
+      "step": 42028
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8851779087535475e-06,
+      "loss": 0.3454,
+      "step": 42029
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.884842073447182e-06,
+      "loss": 0.2364,
+      "step": 42030
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.884506238140816e-06,
+      "loss": 0.2629,
+      "step": 42031
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.884170402834451e-06,
+      "loss": 0.3723,
+      "step": 42032
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.883834567528085e-06,
+      "loss": 0.3481,
+      "step": 42033
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.883498732221718e-06,
+      "loss": 0.3526,
+      "step": 42034
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.883162896915353e-06,
+      "loss": 0.3053,
+      "step": 42035
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8828270616089875e-06,
+      "loss": 0.4555,
+      "step": 42036
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.882491226302621e-06,
+      "loss": 0.2943,
+      "step": 42037
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.882155390996256e-06,
+      "loss": 0.2515,
+      "step": 42038
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.881819555689891e-06,
+      "loss": 0.2651,
+      "step": 42039
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.881483720383524e-06,
+      "loss": 0.3856,
+      "step": 42040
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.881147885077158e-06,
+      "loss": 0.287,
+      "step": 42041
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.880812049770793e-06,
+      "loss": 0.1576,
+      "step": 42042
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.880476214464427e-06,
+      "loss": 0.2226,
+      "step": 42043
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.880140379158061e-06,
+      "loss": 0.2404,
+      "step": 42044
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.879804543851696e-06,
+      "loss": 0.2032,
+      "step": 42045
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.87946870854533e-06,
+      "loss": 0.3649,
+      "step": 42046
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.879132873238964e-06,
+      "loss": 0.1846,
+      "step": 42047
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.878797037932599e-06,
+      "loss": 0.1441,
+      "step": 42048
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.878461202626232e-06,
+      "loss": 0.2625,
+      "step": 42049
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.878125367319867e-06,
+      "loss": 0.3086,
+      "step": 42050
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.877789532013501e-06,
+      "loss": 0.2613,
+      "step": 42051
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.877453696707135e-06,
+      "loss": 0.1928,
+      "step": 42052
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.87711786140077e-06,
+      "loss": 0.2563,
+      "step": 42053
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.876782026094404e-06,
+      "loss": 0.2847,
+      "step": 42054
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.876446190788037e-06,
+      "loss": 0.3492,
+      "step": 42055
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.876110355481672e-06,
+      "loss": 0.2401,
+      "step": 42056
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.875774520175307e-06,
+      "loss": 0.1848,
+      "step": 42057
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8754386848689404e-06,
+      "loss": 0.1901,
+      "step": 42058
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.875102849562575e-06,
+      "loss": 0.3634,
+      "step": 42059
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.87476701425621e-06,
+      "loss": 0.0731,
+      "step": 42060
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8744311789498435e-06,
+      "loss": 0.2176,
+      "step": 42061
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.874095343643478e-06,
+      "loss": 0.2831,
+      "step": 42062
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.873759508337112e-06,
+      "loss": 0.3736,
+      "step": 42063
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.873423673030746e-06,
+      "loss": 0.328,
+      "step": 42064
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8730878377243804e-06,
+      "loss": 0.3069,
+      "step": 42065
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.872752002418015e-06,
+      "loss": 0.3347,
+      "step": 42066
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.872416167111649e-06,
+      "loss": 0.2879,
+      "step": 42067
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8720803318052835e-06,
+      "loss": 0.1323,
+      "step": 42068
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.871744496498918e-06,
+      "loss": 0.2774,
+      "step": 42069
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.871408661192551e-06,
+      "loss": 0.3828,
+      "step": 42070
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.871072825886186e-06,
+      "loss": 0.3776,
+      "step": 42071
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.87073699057982e-06,
+      "loss": 0.1321,
+      "step": 42072
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.870401155273454e-06,
+      "loss": 0.2882,
+      "step": 42073
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.870065319967089e-06,
+      "loss": 0.2469,
+      "step": 42074
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8697294846607235e-06,
+      "loss": 0.2738,
+      "step": 42075
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8693936493543565e-06,
+      "loss": 0.3002,
+      "step": 42076
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.869057814047991e-06,
+      "loss": 0.1896,
+      "step": 42077
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.868721978741626e-06,
+      "loss": 0.2468,
+      "step": 42078
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8683861434352596e-06,
+      "loss": 0.3113,
+      "step": 42079
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.868050308128894e-06,
+      "loss": 0.2563,
+      "step": 42080
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.867714472822529e-06,
+      "loss": 0.3953,
+      "step": 42081
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.867378637516163e-06,
+      "loss": 0.2619,
+      "step": 42082
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.867042802209797e-06,
+      "loss": 0.3378,
+      "step": 42083
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.866706966903432e-06,
+      "loss": 0.2532,
+      "step": 42084
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.866371131597065e-06,
+      "loss": 0.2278,
+      "step": 42085
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8660352962906995e-06,
+      "loss": 0.2314,
+      "step": 42086
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.865699460984334e-06,
+      "loss": 0.1675,
+      "step": 42087
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.865363625677968e-06,
+      "loss": 0.2535,
+      "step": 42088
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.865027790371603e-06,
+      "loss": 0.1755,
+      "step": 42089
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.864691955065237e-06,
+      "loss": 0.3621,
+      "step": 42090
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.86435611975887e-06,
+      "loss": 0.166,
+      "step": 42091
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.864020284452505e-06,
+      "loss": 0.2555,
+      "step": 42092
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8636844491461395e-06,
+      "loss": 0.384,
+      "step": 42093
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.863348613839773e-06,
+      "loss": 0.1925,
+      "step": 42094
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.863012778533408e-06,
+      "loss": 0.2328,
+      "step": 42095
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.862676943227042e-06,
+      "loss": 0.1761,
+      "step": 42096
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.862341107920676e-06,
+      "loss": 0.3263,
+      "step": 42097
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.86200527261431e-06,
+      "loss": 0.3038,
+      "step": 42098
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.861669437307944e-06,
+      "loss": 0.1419,
+      "step": 42099
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.861333602001579e-06,
+      "loss": 0.3059,
+      "step": 42100
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.860997766695213e-06,
+      "loss": 0.2546,
+      "step": 42101
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.860661931388847e-06,
+      "loss": 0.2419,
+      "step": 42102
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.860326096082482e-06,
+      "loss": 0.2051,
+      "step": 42103
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8599902607761164e-06,
+      "loss": 0.1549,
+      "step": 42104
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.859654425469749e-06,
+      "loss": 0.099,
+      "step": 42105
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.859318590163384e-06,
+      "loss": 0.1909,
+      "step": 42106
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.858982754857019e-06,
+      "loss": 0.2301,
+      "step": 42107
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8586469195506525e-06,
+      "loss": 0.2371,
+      "step": 42108
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.858311084244287e-06,
+      "loss": 0.3365,
+      "step": 42109
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.857975248937922e-06,
+      "loss": 0.3486,
+      "step": 42110
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.857639413631555e-06,
+      "loss": 0.3124,
+      "step": 42111
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.857303578325189e-06,
+      "loss": 0.1259,
+      "step": 42112
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.856967743018824e-06,
+      "loss": 0.1622,
+      "step": 42113
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.856631907712458e-06,
+      "loss": 0.3317,
+      "step": 42114
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8562960724060925e-06,
+      "loss": 0.3653,
+      "step": 42115
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.855960237099727e-06,
+      "loss": 0.4194,
+      "step": 42116
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.855624401793361e-06,
+      "loss": 0.1242,
+      "step": 42117
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8552885664869956e-06,
+      "loss": 0.5566,
+      "step": 42118
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.85495273118063e-06,
+      "loss": 0.1961,
+      "step": 42119
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.854616895874263e-06,
+      "loss": 0.1809,
+      "step": 42120
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.854281060567898e-06,
+      "loss": 0.3379,
+      "step": 42121
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8539452252615325e-06,
+      "loss": 0.1984,
+      "step": 42122
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.853609389955166e-06,
+      "loss": 0.235,
+      "step": 42123
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.853273554648801e-06,
+      "loss": 0.2486,
+      "step": 42124
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8529377193424356e-06,
+      "loss": 0.2921,
+      "step": 42125
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8526018840360685e-06,
+      "loss": 0.2777,
+      "step": 42126
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.852266048729703e-06,
+      "loss": 0.2481,
+      "step": 42127
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.851930213423338e-06,
+      "loss": 0.2032,
+      "step": 42128
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.851594378116972e-06,
+      "loss": 0.1217,
+      "step": 42129
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.851258542810606e-06,
+      "loss": 0.3699,
+      "step": 42130
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.850922707504241e-06,
+      "loss": 0.2763,
+      "step": 42131
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.850586872197874e-06,
+      "loss": 0.3491,
+      "step": 42132
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8502510368915085e-06,
+      "loss": 0.2228,
+      "step": 42133
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.849915201585143e-06,
+      "loss": 0.3079,
+      "step": 42134
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.849579366278777e-06,
+      "loss": 0.2882,
+      "step": 42135
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.849243530972412e-06,
+      "loss": 0.1868,
+      "step": 42136
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.848907695666046e-06,
+      "loss": 0.2597,
+      "step": 42137
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.84857186035968e-06,
+      "loss": 0.1717,
+      "step": 42138
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.848236025053315e-06,
+      "loss": 0.2725,
+      "step": 42139
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.847900189746949e-06,
+      "loss": 0.2477,
+      "step": 42140
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.847564354440582e-06,
+      "loss": 0.303,
+      "step": 42141
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.847228519134217e-06,
+      "loss": 0.2847,
+      "step": 42142
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.846892683827852e-06,
+      "loss": 0.2429,
+      "step": 42143
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.846556848521485e-06,
+      "loss": 0.3152,
+      "step": 42144
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.84622101321512e-06,
+      "loss": 0.2651,
+      "step": 42145
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.845885177908755e-06,
+      "loss": 0.2047,
+      "step": 42146
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.845549342602388e-06,
+      "loss": 0.2769,
+      "step": 42147
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.845213507296022e-06,
+      "loss": 0.2608,
+      "step": 42148
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.844877671989657e-06,
+      "loss": 0.4991,
+      "step": 42149
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.844541836683291e-06,
+      "loss": 0.3648,
+      "step": 42150
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.844206001376925e-06,
+      "loss": 0.2632,
+      "step": 42151
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.84387016607056e-06,
+      "loss": 0.1887,
+      "step": 42152
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.843534330764194e-06,
+      "loss": 0.2711,
+      "step": 42153
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.843198495457828e-06,
+      "loss": 0.329,
+      "step": 42154
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.842862660151462e-06,
+      "loss": 0.3057,
+      "step": 42155
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.842526824845096e-06,
+      "loss": 0.3258,
+      "step": 42156
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.842190989538731e-06,
+      "loss": 0.2433,
+      "step": 42157
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.841855154232365e-06,
+      "loss": 0.1412,
+      "step": 42158
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.841519318925999e-06,
+      "loss": 0.2563,
+      "step": 42159
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.841183483619634e-06,
+      "loss": 0.2603,
+      "step": 42160
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8408476483132685e-06,
+      "loss": 0.3833,
+      "step": 42161
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.840511813006901e-06,
+      "loss": 0.4075,
+      "step": 42162
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.840175977700536e-06,
+      "loss": 0.2154,
+      "step": 42163
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.839840142394171e-06,
+      "loss": 0.3251,
+      "step": 42164
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8395043070878045e-06,
+      "loss": 0.2112,
+      "step": 42165
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.839168471781439e-06,
+      "loss": 0.2252,
+      "step": 42166
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.838832636475074e-06,
+      "loss": 0.3248,
+      "step": 42167
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.838496801168707e-06,
+      "loss": 0.3162,
+      "step": 42168
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.838160965862341e-06,
+      "loss": 0.3524,
+      "step": 42169
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.837825130555976e-06,
+      "loss": 0.2332,
+      "step": 42170
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.83748929524961e-06,
+      "loss": 0.361,
+      "step": 42171
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8371534599432445e-06,
+      "loss": 0.2087,
+      "step": 42172
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.836817624636879e-06,
+      "loss": 0.2901,
+      "step": 42173
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.836481789330513e-06,
+      "loss": 0.1625,
+      "step": 42174
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.836145954024148e-06,
+      "loss": 0.2729,
+      "step": 42175
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.835810118717781e-06,
+      "loss": 0.2439,
+      "step": 42176
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.835474283411415e-06,
+      "loss": 0.2167,
+      "step": 42177
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.83513844810505e-06,
+      "loss": 0.1618,
+      "step": 42178
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8348026127986845e-06,
+      "loss": 0.2484,
+      "step": 42179
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.834466777492318e-06,
+      "loss": 0.3089,
+      "step": 42180
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.834130942185953e-06,
+      "loss": 0.2722,
+      "step": 42181
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.833795106879588e-06,
+      "loss": 0.2583,
+      "step": 42182
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 5.8334592715732205e-06,
+      "loss": 0.406,
+      "step": 42183
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.833123436266855e-06,
+      "loss": 0.2872,
+      "step": 42184
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.832787600960489e-06,
+      "loss": 0.4034,
+      "step": 42185
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.832451765654124e-06,
+      "loss": 0.2982,
+      "step": 42186
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.832115930347758e-06,
+      "loss": 0.2988,
+      "step": 42187
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.831780095041391e-06,
+      "loss": 0.3461,
+      "step": 42188
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.831444259735026e-06,
+      "loss": 0.2146,
+      "step": 42189
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8311084244286605e-06,
+      "loss": 0.2029,
+      "step": 42190
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.830772589122294e-06,
+      "loss": 0.1817,
+      "step": 42191
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.830436753815929e-06,
+      "loss": 0.2181,
+      "step": 42192
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.830100918509564e-06,
+      "loss": 0.3141,
+      "step": 42193
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.829765083203197e-06,
+      "loss": 0.347,
+      "step": 42194
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.829429247896832e-06,
+      "loss": 0.2824,
+      "step": 42195
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.829093412590467e-06,
+      "loss": 0.327,
+      "step": 42196
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8287575772841e-06,
+      "loss": 0.2838,
+      "step": 42197
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.828421741977734e-06,
+      "loss": 0.2296,
+      "step": 42198
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.828085906671369e-06,
+      "loss": 0.2897,
+      "step": 42199
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.827750071365003e-06,
+      "loss": 0.3393,
+      "step": 42200
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.827414236058637e-06,
+      "loss": 0.2173,
+      "step": 42201
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.827078400752272e-06,
+      "loss": 0.206,
+      "step": 42202
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.826742565445905e-06,
+      "loss": 0.2008,
+      "step": 42203
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.82640673013954e-06,
+      "loss": 0.2962,
+      "step": 42204
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.826070894833174e-06,
+      "loss": 0.1629,
+      "step": 42205
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.825735059526808e-06,
+      "loss": 0.3169,
+      "step": 42206
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.825399224220443e-06,
+      "loss": 0.2148,
+      "step": 42207
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.825063388914077e-06,
+      "loss": 0.2745,
+      "step": 42208
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.824727553607711e-06,
+      "loss": 0.1442,
+      "step": 42209
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.824391718301345e-06,
+      "loss": 0.4485,
+      "step": 42210
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.82405588299498e-06,
+      "loss": 0.3379,
+      "step": 42211
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8237200476886135e-06,
+      "loss": 0.2717,
+      "step": 42212
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.823384212382248e-06,
+      "loss": 0.4353,
+      "step": 42213
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.823048377075883e-06,
+      "loss": 0.1797,
+      "step": 42214
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8227125417695165e-06,
+      "loss": 0.3106,
+      "step": 42215
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.822376706463151e-06,
+      "loss": 0.2976,
+      "step": 42216
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.822040871156786e-06,
+      "loss": 0.3783,
+      "step": 42217
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.821705035850419e-06,
+      "loss": 0.2679,
+      "step": 42218
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8213692005440534e-06,
+      "loss": 0.1848,
+      "step": 42219
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.821033365237688e-06,
+      "loss": 0.2051,
+      "step": 42220
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.820697529931322e-06,
+      "loss": 0.4408,
+      "step": 42221
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8203616946249565e-06,
+      "loss": 0.3052,
+      "step": 42222
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.820025859318591e-06,
+      "loss": 0.2842,
+      "step": 42223
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.819690024012224e-06,
+      "loss": 0.2098,
+      "step": 42224
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.819354188705859e-06,
+      "loss": 0.2313,
+      "step": 42225
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8190183533994934e-06,
+      "loss": 0.1653,
+      "step": 42226
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.818682518093127e-06,
+      "loss": 0.2842,
+      "step": 42227
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.818346682786762e-06,
+      "loss": 0.1844,
+      "step": 42228
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8180108474803965e-06,
+      "loss": 0.4203,
+      "step": 42229
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.81767501217403e-06,
+      "loss": 0.312,
+      "step": 42230
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.817339176867665e-06,
+      "loss": 0.4194,
+      "step": 42231
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8170033415613e-06,
+      "loss": 0.3906,
+      "step": 42232
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8166675062549326e-06,
+      "loss": 0.2525,
+      "step": 42233
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.816331670948567e-06,
+      "loss": 0.3518,
+      "step": 42234
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.815995835642202e-06,
+      "loss": 0.1714,
+      "step": 42235
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.815660000335836e-06,
+      "loss": 0.3996,
+      "step": 42236
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.81532416502947e-06,
+      "loss": 0.3358,
+      "step": 42237
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.814988329723105e-06,
+      "loss": 0.3939,
+      "step": 42238
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.814652494416738e-06,
+      "loss": 0.2024,
+      "step": 42239
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8143166591103726e-06,
+      "loss": 0.2266,
+      "step": 42240
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.813980823804007e-06,
+      "loss": 0.3124,
+      "step": 42241
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.813644988497641e-06,
+      "loss": 0.2211,
+      "step": 42242
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.813309153191276e-06,
+      "loss": 0.1723,
+      "step": 42243
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.81297331788491e-06,
+      "loss": 0.2307,
+      "step": 42244
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.812637482578543e-06,
+      "loss": 0.3113,
+      "step": 42245
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.812301647272178e-06,
+      "loss": 0.4947,
+      "step": 42246
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8119658119658126e-06,
+      "loss": 0.1561,
+      "step": 42247
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.811629976659446e-06,
+      "loss": 0.3205,
+      "step": 42248
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.811294141353081e-06,
+      "loss": 0.3016,
+      "step": 42249
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.810958306046716e-06,
+      "loss": 0.2645,
+      "step": 42250
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8106224707403495e-06,
+      "loss": 0.425,
+      "step": 42251
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.810286635433984e-06,
+      "loss": 0.1739,
+      "step": 42252
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.809950800127619e-06,
+      "loss": 0.2565,
+      "step": 42253
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.809614964821252e-06,
+      "loss": 0.1686,
+      "step": 42254
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.809279129514886e-06,
+      "loss": 0.2108,
+      "step": 42255
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.808943294208521e-06,
+      "loss": 0.3294,
+      "step": 42256
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.808607458902155e-06,
+      "loss": 0.2493,
+      "step": 42257
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8082716235957894e-06,
+      "loss": 0.3048,
+      "step": 42258
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.807935788289424e-06,
+      "loss": 0.3232,
+      "step": 42259
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.807599952983057e-06,
+      "loss": 0.3777,
+      "step": 42260
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.807264117676692e-06,
+      "loss": 0.1718,
+      "step": 42261
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.806928282370326e-06,
+      "loss": 0.4354,
+      "step": 42262
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.80659244706396e-06,
+      "loss": 0.4174,
+      "step": 42263
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.806256611757595e-06,
+      "loss": 0.2496,
+      "step": 42264
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8059207764512294e-06,
+      "loss": 0.2012,
+      "step": 42265
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.805584941144863e-06,
+      "loss": 0.3146,
+      "step": 42266
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.805249105838497e-06,
+      "loss": 0.1929,
+      "step": 42267
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.804913270532132e-06,
+      "loss": 0.244,
+      "step": 42268
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8045774352257655e-06,
+      "loss": 0.2542,
+      "step": 42269
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8042415999194e-06,
+      "loss": 0.2581,
+      "step": 42270
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.803905764613035e-06,
+      "loss": 0.1941,
+      "step": 42271
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.803569929306669e-06,
+      "loss": 0.3638,
+      "step": 42272
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.803234094000303e-06,
+      "loss": 0.2918,
+      "step": 42273
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.802898258693936e-06,
+      "loss": 0.3152,
+      "step": 42274
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.802562423387571e-06,
+      "loss": 0.2343,
+      "step": 42275
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8022265880812055e-06,
+      "loss": 0.1833,
+      "step": 42276
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.801890752774839e-06,
+      "loss": 0.3115,
+      "step": 42277
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.801554917468474e-06,
+      "loss": 0.7061,
+      "step": 42278
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8012190821621086e-06,
+      "loss": 0.2634,
+      "step": 42279
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.8008832468557415e-06,
+      "loss": 0.3057,
+      "step": 42280
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.800547411549376e-06,
+      "loss": 0.2922,
+      "step": 42281
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.800211576243011e-06,
+      "loss": 0.2019,
+      "step": 42282
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.799875740936645e-06,
+      "loss": 0.449,
+      "step": 42283
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.799539905630279e-06,
+      "loss": 0.1847,
+      "step": 42284
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.799204070323914e-06,
+      "loss": 0.2172,
+      "step": 42285
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.798868235017548e-06,
+      "loss": 0.2075,
+      "step": 42286
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.798532399711182e-06,
+      "loss": 0.2957,
+      "step": 42287
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.798196564404817e-06,
+      "loss": 0.177,
+      "step": 42288
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.79786072909845e-06,
+      "loss": 0.1426,
+      "step": 42289
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.797524893792085e-06,
+      "loss": 0.2775,
+      "step": 42290
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.797189058485719e-06,
+      "loss": 0.2476,
+      "step": 42291
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.796853223179353e-06,
+      "loss": 0.3182,
+      "step": 42292
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.796517387872988e-06,
+      "loss": 0.1521,
+      "step": 42293
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.796181552566622e-06,
+      "loss": 0.2633,
+      "step": 42294
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.795845717260255e-06,
+      "loss": 0.2174,
+      "step": 42295
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.79550988195389e-06,
+      "loss": 0.4807,
+      "step": 42296
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.795174046647525e-06,
+      "loss": 0.2622,
+      "step": 42297
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.794838211341158e-06,
+      "loss": 0.3147,
+      "step": 42298
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.794502376034793e-06,
+      "loss": 0.1654,
+      "step": 42299
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.794166540728428e-06,
+      "loss": 0.2275,
+      "step": 42300
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.793830705422061e-06,
+      "loss": 0.3529,
+      "step": 42301
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.793494870115695e-06,
+      "loss": 0.1471,
+      "step": 42302
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.79315903480933e-06,
+      "loss": 0.1886,
+      "step": 42303
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.792823199502964e-06,
+      "loss": 0.2297,
+      "step": 42304
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.792487364196598e-06,
+      "loss": 0.3429,
+      "step": 42305
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.792151528890233e-06,
+      "loss": 0.2787,
+      "step": 42306
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.791815693583867e-06,
+      "loss": 0.2132,
+      "step": 42307
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.7914798582775015e-06,
+      "loss": 0.0972,
+      "step": 42308
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.791144022971136e-06,
+      "loss": 0.2194,
+      "step": 42309
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.790808187664769e-06,
+      "loss": 0.3377,
+      "step": 42310
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.790472352358404e-06,
+      "loss": 0.2302,
+      "step": 42311
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.790136517052038e-06,
+      "loss": 0.1493,
+      "step": 42312
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.789800681745672e-06,
+      "loss": 0.2721,
+      "step": 42313
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.789464846439307e-06,
+      "loss": 0.3211,
+      "step": 42314
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.7891290111329415e-06,
+      "loss": 0.4362,
+      "step": 42315
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.7887931758265744e-06,
+      "loss": 0.5462,
+      "step": 42316
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.788457340520209e-06,
+      "loss": 0.2417,
+      "step": 42317
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.788121505213844e-06,
+      "loss": 0.1793,
+      "step": 42318
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.7877856699074775e-06,
+      "loss": 0.5058,
+      "step": 42319
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.787449834601112e-06,
+      "loss": 0.3363,
+      "step": 42320
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.787113999294747e-06,
+      "loss": 0.2884,
+      "step": 42321
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.786778163988381e-06,
+      "loss": 0.2592,
+      "step": 42322
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.786442328682014e-06,
+      "loss": 0.3333,
+      "step": 42323
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.786106493375649e-06,
+      "loss": 0.2391,
+      "step": 42324
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.785770658069283e-06,
+      "loss": 0.2482,
+      "step": 42325
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.7854348227629175e-06,
+      "loss": 0.1871,
+      "step": 42326
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.785098987456552e-06,
+      "loss": 0.1525,
+      "step": 42327
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.784763152150186e-06,
+      "loss": 0.2749,
+      "step": 42328
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.784427316843821e-06,
+      "loss": 0.2527,
+      "step": 42329
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.784091481537455e-06,
+      "loss": 0.1206,
+      "step": 42330
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.783755646231088e-06,
+      "loss": 0.213,
+      "step": 42331
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.783419810924723e-06,
+      "loss": 0.2405,
+      "step": 42332
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.7830839756183575e-06,
+      "loss": 0.182,
+      "step": 42333
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.782748140311991e-06,
+      "loss": 0.2739,
+      "step": 42334
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.782412305005626e-06,
+      "loss": 0.2482,
+      "step": 42335
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.782076469699261e-06,
+      "loss": 0.364,
+      "step": 42336
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.7817406343928936e-06,
+      "loss": 0.1851,
+      "step": 42337
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.781404799086528e-06,
+      "loss": 0.3296,
+      "step": 42338
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.781068963780163e-06,
+      "loss": 0.1761,
+      "step": 42339
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.780733128473797e-06,
+      "loss": 0.1905,
+      "step": 42340
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.780397293167431e-06,
+      "loss": 0.1456,
+      "step": 42341
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.780061457861066e-06,
+      "loss": 0.2488,
+      "step": 42342
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.7797256225547e-06,
+      "loss": 0.203,
+      "step": 42343
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.779389787248334e-06,
+      "loss": 0.319,
+      "step": 42344
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.779053951941968e-06,
+      "loss": 0.4628,
+      "step": 42345
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.778718116635602e-06,
+      "loss": 0.2178,
+      "step": 42346
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.778382281329237e-06,
+      "loss": 0.4311,
+      "step": 42347
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.778046446022871e-06,
+      "loss": 0.2486,
+      "step": 42348
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.777710610716505e-06,
+      "loss": 0.2651,
+      "step": 42349
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.77737477541014e-06,
+      "loss": 0.2231,
+      "step": 42350
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.777038940103774e-06,
+      "loss": 0.2177,
+      "step": 42351
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.776703104797407e-06,
+      "loss": 0.2799,
+      "step": 42352
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.776367269491042e-06,
+      "loss": 0.264,
+      "step": 42353
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.776031434184677e-06,
+      "loss": 0.3139,
+      "step": 42354
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.7756955988783104e-06,
+      "loss": 0.2175,
+      "step": 42355
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.775359763571945e-06,
+      "loss": 0.2032,
+      "step": 42356
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.77502392826558e-06,
+      "loss": 0.2552,
+      "step": 42357
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.774688092959213e-06,
+      "loss": 0.2219,
+      "step": 42358
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.774352257652847e-06,
+      "loss": 0.3328,
+      "step": 42359
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.774016422346482e-06,
+      "loss": 0.2955,
+      "step": 42360
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.773680587040116e-06,
+      "loss": 0.3946,
+      "step": 42361
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.77334475173375e-06,
+      "loss": 0.2469,
+      "step": 42362
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.773008916427385e-06,
+      "loss": 0.2159,
+      "step": 42363
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.772673081121019e-06,
+      "loss": 0.2144,
+      "step": 42364
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.7723372458146535e-06,
+      "loss": 0.2265,
+      "step": 42365
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.7720014105082865e-06,
+      "loss": 0.2488,
+      "step": 42366
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.771665575201921e-06,
+      "loss": 0.2099,
+      "step": 42367
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.771329739895556e-06,
+      "loss": 0.2665,
+      "step": 42368
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.7709939045891896e-06,
+      "loss": 0.2005,
+      "step": 42369
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.770658069282824e-06,
+      "loss": 0.3953,
+      "step": 42370
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.770322233976459e-06,
+      "loss": 0.3204,
+      "step": 42371
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.769986398670092e-06,
+      "loss": 0.3331,
+      "step": 42372
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.7696505633637265e-06,
+      "loss": 0.2387,
+      "step": 42373
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.769314728057361e-06,
+      "loss": 0.3549,
+      "step": 42374
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.768978892750995e-06,
+      "loss": 0.2644,
+      "step": 42375
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.7686430574446296e-06,
+      "loss": 0.2621,
+      "step": 42376
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.768307222138264e-06,
+      "loss": 0.3373,
+      "step": 42377
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.767971386831898e-06,
+      "loss": 0.3412,
+      "step": 42378
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.767635551525533e-06,
+      "loss": 0.247,
+      "step": 42379
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.7672997162191664e-06,
+      "loss": 0.2464,
+      "step": 42380
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 5.7669638809128e-06,
+      "loss": 0.1049,
+      "step": 42381
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.766628045606435e-06,
+      "loss": 0.1909,
+      "step": 42382
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7662922103000695e-06,
+      "loss": 0.3153,
+      "step": 42383
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.765956374993703e-06,
+      "loss": 0.466,
+      "step": 42384
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.765620539687338e-06,
+      "loss": 0.2062,
+      "step": 42385
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.765284704380973e-06,
+      "loss": 0.3674,
+      "step": 42386
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.764948869074606e-06,
+      "loss": 0.1585,
+      "step": 42387
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.76461303376824e-06,
+      "loss": 0.2325,
+      "step": 42388
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.764277198461875e-06,
+      "loss": 0.2184,
+      "step": 42389
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.763941363155509e-06,
+      "loss": 0.2435,
+      "step": 42390
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.763605527849143e-06,
+      "loss": 0.2734,
+      "step": 42391
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.763269692542778e-06,
+      "loss": 0.2346,
+      "step": 42392
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.762933857236411e-06,
+      "loss": 0.1157,
+      "step": 42393
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.762598021930046e-06,
+      "loss": 0.2893,
+      "step": 42394
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.76226218662368e-06,
+      "loss": 0.3611,
+      "step": 42395
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.761926351317314e-06,
+      "loss": 0.2521,
+      "step": 42396
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.761590516010949e-06,
+      "loss": 0.1951,
+      "step": 42397
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.761254680704583e-06,
+      "loss": 0.3695,
+      "step": 42398
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.760918845398217e-06,
+      "loss": 0.1844,
+      "step": 42399
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.760583010091852e-06,
+      "loss": 0.2817,
+      "step": 42400
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.760247174785486e-06,
+      "loss": 0.1658,
+      "step": 42401
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.759911339479119e-06,
+      "loss": 0.2328,
+      "step": 42402
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.759575504172754e-06,
+      "loss": 0.1639,
+      "step": 42403
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.759239668866389e-06,
+      "loss": 0.3213,
+      "step": 42404
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7589038335600225e-06,
+      "loss": 0.2152,
+      "step": 42405
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.758567998253657e-06,
+      "loss": 0.2352,
+      "step": 42406
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.758232162947292e-06,
+      "loss": 0.1775,
+      "step": 42407
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.757896327640925e-06,
+      "loss": 0.2423,
+      "step": 42408
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.757560492334559e-06,
+      "loss": 0.2888,
+      "step": 42409
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.757224657028194e-06,
+      "loss": 0.169,
+      "step": 42410
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.756888821721828e-06,
+      "loss": 0.257,
+      "step": 42411
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7565529864154625e-06,
+      "loss": 0.3397,
+      "step": 42412
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.756217151109097e-06,
+      "loss": 0.3145,
+      "step": 42413
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.75588131580273e-06,
+      "loss": 0.2932,
+      "step": 42414
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.755545480496365e-06,
+      "loss": 0.1566,
+      "step": 42415
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.755209645189999e-06,
+      "loss": 0.3032,
+      "step": 42416
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.754873809883633e-06,
+      "loss": 0.3517,
+      "step": 42417
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.754537974577268e-06,
+      "loss": 0.4068,
+      "step": 42418
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7542021392709025e-06,
+      "loss": 0.1905,
+      "step": 42419
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.753866303964536e-06,
+      "loss": 0.2303,
+      "step": 42420
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.753530468658171e-06,
+      "loss": 0.3262,
+      "step": 42421
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7531946333518055e-06,
+      "loss": 0.4478,
+      "step": 42422
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7528587980454385e-06,
+      "loss": 0.3971,
+      "step": 42423
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.752522962739073e-06,
+      "loss": 0.3327,
+      "step": 42424
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.752187127432708e-06,
+      "loss": 0.1792,
+      "step": 42425
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.751851292126342e-06,
+      "loss": 0.3457,
+      "step": 42426
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.751515456819976e-06,
+      "loss": 0.1369,
+      "step": 42427
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.751179621513611e-06,
+      "loss": 0.2504,
+      "step": 42428
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.750843786207244e-06,
+      "loss": 0.3699,
+      "step": 42429
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7505079509008785e-06,
+      "loss": 0.2313,
+      "step": 42430
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.750172115594513e-06,
+      "loss": 0.2734,
+      "step": 42431
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.749836280288147e-06,
+      "loss": 0.3446,
+      "step": 42432
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.749500444981782e-06,
+      "loss": 0.3379,
+      "step": 42433
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.749164609675416e-06,
+      "loss": 0.2465,
+      "step": 42434
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.74882877436905e-06,
+      "loss": 0.2542,
+      "step": 42435
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.748492939062684e-06,
+      "loss": 0.4167,
+      "step": 42436
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7481571037563185e-06,
+      "loss": 0.2227,
+      "step": 42437
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.747821268449952e-06,
+      "loss": 0.3223,
+      "step": 42438
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.747485433143587e-06,
+      "loss": 0.3141,
+      "step": 42439
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7471495978372216e-06,
+      "loss": 0.2628,
+      "step": 42440
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.746813762530855e-06,
+      "loss": 0.2188,
+      "step": 42441
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.74647792722449e-06,
+      "loss": 0.2985,
+      "step": 42442
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.746142091918125e-06,
+      "loss": 0.2408,
+      "step": 42443
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.745806256611758e-06,
+      "loss": 0.2765,
+      "step": 42444
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.745470421305392e-06,
+      "loss": 0.1855,
+      "step": 42445
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.745134585999027e-06,
+      "loss": 0.2373,
+      "step": 42446
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.744798750692661e-06,
+      "loss": 0.5041,
+      "step": 42447
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.744462915386295e-06,
+      "loss": 0.3688,
+      "step": 42448
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.74412708007993e-06,
+      "loss": 0.1855,
+      "step": 42449
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.743791244773563e-06,
+      "loss": 0.4376,
+      "step": 42450
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.743455409467198e-06,
+      "loss": 0.2157,
+      "step": 42451
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.743119574160832e-06,
+      "loss": 0.4573,
+      "step": 42452
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.742783738854466e-06,
+      "loss": 0.2182,
+      "step": 42453
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.742447903548101e-06,
+      "loss": 0.1558,
+      "step": 42454
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7421120682417345e-06,
+      "loss": 0.279,
+      "step": 42455
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.741776232935369e-06,
+      "loss": 0.2363,
+      "step": 42456
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.741440397629004e-06,
+      "loss": 0.1744,
+      "step": 42457
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.741104562322637e-06,
+      "loss": 0.2714,
+      "step": 42458
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.740768727016271e-06,
+      "loss": 0.2682,
+      "step": 42459
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.740432891709906e-06,
+      "loss": 0.1878,
+      "step": 42460
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.74009705640354e-06,
+      "loss": 0.3023,
+      "step": 42461
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7397612210971745e-06,
+      "loss": 0.2526,
+      "step": 42462
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.739425385790809e-06,
+      "loss": 0.3773,
+      "step": 42463
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.739089550484442e-06,
+      "loss": 0.1577,
+      "step": 42464
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.738753715178077e-06,
+      "loss": 0.3999,
+      "step": 42465
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.738417879871711e-06,
+      "loss": 0.3618,
+      "step": 42466
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.738082044565345e-06,
+      "loss": 0.2597,
+      "step": 42467
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.73774620925898e-06,
+      "loss": 0.4059,
+      "step": 42468
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7374103739526145e-06,
+      "loss": 0.2187,
+      "step": 42469
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7370745386462474e-06,
+      "loss": 0.215,
+      "step": 42470
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.736738703339882e-06,
+      "loss": 0.1257,
+      "step": 42471
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.736402868033517e-06,
+      "loss": 0.271,
+      "step": 42472
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7360670327271505e-06,
+      "loss": 0.2515,
+      "step": 42473
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.735731197420785e-06,
+      "loss": 0.3949,
+      "step": 42474
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.73539536211442e-06,
+      "loss": 0.2188,
+      "step": 42475
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.735059526808054e-06,
+      "loss": 0.352,
+      "step": 42476
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.734723691501688e-06,
+      "loss": 0.1627,
+      "step": 42477
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.734387856195323e-06,
+      "loss": 0.4892,
+      "step": 42478
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.734052020888956e-06,
+      "loss": 0.2047,
+      "step": 42479
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7337161855825905e-06,
+      "loss": 0.24,
+      "step": 42480
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.733380350276225e-06,
+      "loss": 0.2629,
+      "step": 42481
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.733044514969859e-06,
+      "loss": 0.1907,
+      "step": 42482
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.732708679663494e-06,
+      "loss": 0.2452,
+      "step": 42483
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.732372844357128e-06,
+      "loss": 0.2739,
+      "step": 42484
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.732037009050761e-06,
+      "loss": 0.3433,
+      "step": 42485
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.731701173744396e-06,
+      "loss": 0.2371,
+      "step": 42486
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7313653384380305e-06,
+      "loss": 0.2963,
+      "step": 42487
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.731029503131664e-06,
+      "loss": 0.1832,
+      "step": 42488
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.730693667825299e-06,
+      "loss": 0.2293,
+      "step": 42489
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.730357832518934e-06,
+      "loss": 0.2658,
+      "step": 42490
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.730021997212567e-06,
+      "loss": 0.1484,
+      "step": 42491
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.729686161906201e-06,
+      "loss": 0.3728,
+      "step": 42492
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.729350326599836e-06,
+      "loss": 0.2698,
+      "step": 42493
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.72901449129347e-06,
+      "loss": 0.1876,
+      "step": 42494
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.728678655987104e-06,
+      "loss": 0.2644,
+      "step": 42495
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.728342820680739e-06,
+      "loss": 0.3751,
+      "step": 42496
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.728006985374373e-06,
+      "loss": 0.3053,
+      "step": 42497
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.727671150068007e-06,
+      "loss": 0.211,
+      "step": 42498
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.727335314761642e-06,
+      "loss": 0.3057,
+      "step": 42499
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.726999479455275e-06,
+      "loss": 0.3093,
+      "step": 42500
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.72666364414891e-06,
+      "loss": 0.3005,
+      "step": 42501
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.726327808842544e-06,
+      "loss": 0.1747,
+      "step": 42502
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.725991973536178e-06,
+      "loss": 0.2213,
+      "step": 42503
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.725656138229813e-06,
+      "loss": 0.2585,
+      "step": 42504
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.725320302923447e-06,
+      "loss": 0.2023,
+      "step": 42505
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.72498446761708e-06,
+      "loss": 0.2378,
+      "step": 42506
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.724648632310715e-06,
+      "loss": 0.2539,
+      "step": 42507
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.72431279700435e-06,
+      "loss": 0.1389,
+      "step": 42508
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7239769616979834e-06,
+      "loss": 0.1999,
+      "step": 42509
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.723641126391618e-06,
+      "loss": 0.1903,
+      "step": 42510
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.723305291085253e-06,
+      "loss": 0.3629,
+      "step": 42511
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7229694557788865e-06,
+      "loss": 0.2615,
+      "step": 42512
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.722633620472521e-06,
+      "loss": 0.2564,
+      "step": 42513
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.722297785166156e-06,
+      "loss": 0.271,
+      "step": 42514
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.721961949859789e-06,
+      "loss": 0.3222,
+      "step": 42515
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7216261145534234e-06,
+      "loss": 0.287,
+      "step": 42516
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.721290279247058e-06,
+      "loss": 0.3126,
+      "step": 42517
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.720954443940692e-06,
+      "loss": 0.145,
+      "step": 42518
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7206186086343265e-06,
+      "loss": 0.1806,
+      "step": 42519
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.720282773327961e-06,
+      "loss": 0.2195,
+      "step": 42520
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.719946938021594e-06,
+      "loss": 0.2257,
+      "step": 42521
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.719611102715229e-06,
+      "loss": 0.2077,
+      "step": 42522
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7192752674088634e-06,
+      "loss": 0.4108,
+      "step": 42523
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.718939432102497e-06,
+      "loss": 0.2647,
+      "step": 42524
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.718603596796132e-06,
+      "loss": 0.3149,
+      "step": 42525
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7182677614897665e-06,
+      "loss": 0.2488,
+      "step": 42526
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7179319261833995e-06,
+      "loss": 0.4524,
+      "step": 42527
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.717596090877034e-06,
+      "loss": 0.2756,
+      "step": 42528
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.717260255570669e-06,
+      "loss": 0.2888,
+      "step": 42529
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7169244202643026e-06,
+      "loss": 0.2368,
+      "step": 42530
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.716588584957937e-06,
+      "loss": 0.2125,
+      "step": 42531
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.716252749651572e-06,
+      "loss": 0.3143,
+      "step": 42532
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.715916914345206e-06,
+      "loss": 0.244,
+      "step": 42533
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.71558107903884e-06,
+      "loss": 0.209,
+      "step": 42534
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.715245243732475e-06,
+      "loss": 0.2004,
+      "step": 42535
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.714909408426108e-06,
+      "loss": 0.3568,
+      "step": 42536
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7145735731197426e-06,
+      "loss": 0.2409,
+      "step": 42537
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.714237737813377e-06,
+      "loss": 0.222,
+      "step": 42538
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.713901902507011e-06,
+      "loss": 0.2274,
+      "step": 42539
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.713566067200646e-06,
+      "loss": 0.2873,
+      "step": 42540
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.71323023189428e-06,
+      "loss": 0.4273,
+      "step": 42541
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.712894396587913e-06,
+      "loss": 0.2908,
+      "step": 42542
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.712558561281548e-06,
+      "loss": 0.2685,
+      "step": 42543
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7122227259751826e-06,
+      "loss": 0.1338,
+      "step": 42544
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.711886890668816e-06,
+      "loss": 0.3341,
+      "step": 42545
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.711551055362451e-06,
+      "loss": 0.1876,
+      "step": 42546
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.711215220056085e-06,
+      "loss": 0.319,
+      "step": 42547
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7108793847497194e-06,
+      "loss": 0.145,
+      "step": 42548
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.710543549443353e-06,
+      "loss": 0.1761,
+      "step": 42549
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.710207714136987e-06,
+      "loss": 0.4691,
+      "step": 42550
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.709871878830622e-06,
+      "loss": 0.248,
+      "step": 42551
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.709536043524256e-06,
+      "loss": 0.1923,
+      "step": 42552
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.70920020821789e-06,
+      "loss": 0.1506,
+      "step": 42553
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.708864372911525e-06,
+      "loss": 0.264,
+      "step": 42554
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7085285376051594e-06,
+      "loss": 0.3424,
+      "step": 42555
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.708192702298792e-06,
+      "loss": 0.1958,
+      "step": 42556
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.707856866992427e-06,
+      "loss": 0.3544,
+      "step": 42557
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.707521031686062e-06,
+      "loss": 0.1639,
+      "step": 42558
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7071851963796955e-06,
+      "loss": 0.121,
+      "step": 42559
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.70684936107333e-06,
+      "loss": 0.3173,
+      "step": 42560
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.706513525766965e-06,
+      "loss": 0.2542,
+      "step": 42561
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.706177690460598e-06,
+      "loss": 0.2475,
+      "step": 42562
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.705841855154232e-06,
+      "loss": 0.1672,
+      "step": 42563
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.705506019847867e-06,
+      "loss": 0.1438,
+      "step": 42564
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.705170184541501e-06,
+      "loss": 0.2891,
+      "step": 42565
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7048343492351355e-06,
+      "loss": 0.3691,
+      "step": 42566
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.70449851392877e-06,
+      "loss": 0.2339,
+      "step": 42567
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.704162678622404e-06,
+      "loss": 0.4426,
+      "step": 42568
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7038268433160386e-06,
+      "loss": 0.1978,
+      "step": 42569
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.703491008009673e-06,
+      "loss": 0.3534,
+      "step": 42570
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.703155172703306e-06,
+      "loss": 0.2712,
+      "step": 42571
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.702819337396941e-06,
+      "loss": 0.1859,
+      "step": 42572
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7024835020905755e-06,
+      "loss": 0.2534,
+      "step": 42573
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.702147666784209e-06,
+      "loss": 0.1727,
+      "step": 42574
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.701811831477844e-06,
+      "loss": 0.2863,
+      "step": 42575
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7014759961714786e-06,
+      "loss": 0.3483,
+      "step": 42576
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.7011401608651115e-06,
+      "loss": 0.2001,
+      "step": 42577
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.700804325558746e-06,
+      "loss": 0.2863,
+      "step": 42578
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.700468490252381e-06,
+      "loss": 0.2704,
+      "step": 42579
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 5.700132654946015e-06,
+      "loss": 0.3296,
+      "step": 42580
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.699796819639649e-06,
+      "loss": 0.2855,
+      "step": 42581
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.699460984333284e-06,
+      "loss": 0.3887,
+      "step": 42582
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.699125149026917e-06,
+      "loss": 0.1783,
+      "step": 42583
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6987893137205515e-06,
+      "loss": 0.2774,
+      "step": 42584
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.698453478414186e-06,
+      "loss": 0.2019,
+      "step": 42585
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.69811764310782e-06,
+      "loss": 0.4244,
+      "step": 42586
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.697781807801455e-06,
+      "loss": 0.2122,
+      "step": 42587
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.697445972495089e-06,
+      "loss": 0.2771,
+      "step": 42588
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.697110137188723e-06,
+      "loss": 0.208,
+      "step": 42589
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.696774301882358e-06,
+      "loss": 0.3057,
+      "step": 42590
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.696438466575992e-06,
+      "loss": 0.2267,
+      "step": 42591
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.696102631269625e-06,
+      "loss": 0.4286,
+      "step": 42592
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.69576679596326e-06,
+      "loss": 0.3168,
+      "step": 42593
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.695430960656895e-06,
+      "loss": 0.1952,
+      "step": 42594
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.695095125350528e-06,
+      "loss": 0.359,
+      "step": 42595
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.694759290044163e-06,
+      "loss": 0.2248,
+      "step": 42596
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.694423454737798e-06,
+      "loss": 0.2559,
+      "step": 42597
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.694087619431431e-06,
+      "loss": 0.285,
+      "step": 42598
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.693751784125065e-06,
+      "loss": 0.2403,
+      "step": 42599
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6934159488187e-06,
+      "loss": 0.2716,
+      "step": 42600
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.693080113512334e-06,
+      "loss": 0.2876,
+      "step": 42601
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.692744278205968e-06,
+      "loss": 0.2338,
+      "step": 42602
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.692408442899603e-06,
+      "loss": 0.2843,
+      "step": 42603
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.692072607593237e-06,
+      "loss": 0.1891,
+      "step": 42604
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.691736772286871e-06,
+      "loss": 0.2636,
+      "step": 42605
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.691400936980505e-06,
+      "loss": 0.2687,
+      "step": 42606
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.691065101674139e-06,
+      "loss": 0.154,
+      "step": 42607
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.690729266367774e-06,
+      "loss": 0.3267,
+      "step": 42608
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.690393431061408e-06,
+      "loss": 0.369,
+      "step": 42609
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.690057595755042e-06,
+      "loss": 0.2542,
+      "step": 42610
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.689721760448677e-06,
+      "loss": 0.2059,
+      "step": 42611
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6893859251423115e-06,
+      "loss": 0.274,
+      "step": 42612
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.689050089835944e-06,
+      "loss": 0.2484,
+      "step": 42613
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.688714254529579e-06,
+      "loss": 0.1855,
+      "step": 42614
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.688378419223214e-06,
+      "loss": 0.3029,
+      "step": 42615
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6880425839168475e-06,
+      "loss": 0.2999,
+      "step": 42616
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.687706748610482e-06,
+      "loss": 0.3076,
+      "step": 42617
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.687370913304117e-06,
+      "loss": 0.2629,
+      "step": 42618
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.68703507799775e-06,
+      "loss": 0.3306,
+      "step": 42619
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.686699242691384e-06,
+      "loss": 0.1878,
+      "step": 42620
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.686363407385019e-06,
+      "loss": 0.1851,
+      "step": 42621
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.686027572078653e-06,
+      "loss": 0.3149,
+      "step": 42622
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6856917367722875e-06,
+      "loss": 0.3899,
+      "step": 42623
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.685355901465922e-06,
+      "loss": 0.2567,
+      "step": 42624
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.685020066159556e-06,
+      "loss": 0.1775,
+      "step": 42625
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.684684230853191e-06,
+      "loss": 0.4273,
+      "step": 42626
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.684348395546824e-06,
+      "loss": 0.242,
+      "step": 42627
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.684012560240458e-06,
+      "loss": 0.4183,
+      "step": 42628
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.683676724934093e-06,
+      "loss": 0.3589,
+      "step": 42629
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6833408896277275e-06,
+      "loss": 0.2275,
+      "step": 42630
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.683005054321361e-06,
+      "loss": 0.2991,
+      "step": 42631
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.682669219014996e-06,
+      "loss": 0.1999,
+      "step": 42632
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.682333383708631e-06,
+      "loss": 0.1558,
+      "step": 42633
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6819975484022635e-06,
+      "loss": 0.3277,
+      "step": 42634
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.681661713095898e-06,
+      "loss": 0.1956,
+      "step": 42635
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.681325877789532e-06,
+      "loss": 0.2552,
+      "step": 42636
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.680990042483167e-06,
+      "loss": 0.2409,
+      "step": 42637
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.680654207176801e-06,
+      "loss": 0.3085,
+      "step": 42638
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.680318371870435e-06,
+      "loss": 0.4842,
+      "step": 42639
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.679982536564069e-06,
+      "loss": 0.2692,
+      "step": 42640
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6796467012577035e-06,
+      "loss": 0.2048,
+      "step": 42641
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.679310865951337e-06,
+      "loss": 0.3053,
+      "step": 42642
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.678975030644972e-06,
+      "loss": 0.2062,
+      "step": 42643
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.678639195338607e-06,
+      "loss": 0.1152,
+      "step": 42644
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6783033600322404e-06,
+      "loss": 0.3351,
+      "step": 42645
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.677967524725875e-06,
+      "loss": 0.2506,
+      "step": 42646
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.67763168941951e-06,
+      "loss": 0.2984,
+      "step": 42647
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.677295854113143e-06,
+      "loss": 0.3474,
+      "step": 42648
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.676960018806777e-06,
+      "loss": 0.2639,
+      "step": 42649
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.676624183500412e-06,
+      "loss": 0.1934,
+      "step": 42650
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.676288348194046e-06,
+      "loss": 0.1064,
+      "step": 42651
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6759525128876804e-06,
+      "loss": 0.2788,
+      "step": 42652
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.675616677581315e-06,
+      "loss": 0.3608,
+      "step": 42653
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.675280842274948e-06,
+      "loss": 0.2243,
+      "step": 42654
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.674945006968583e-06,
+      "loss": 0.2353,
+      "step": 42655
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.674609171662217e-06,
+      "loss": 0.1954,
+      "step": 42656
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.674273336355851e-06,
+      "loss": 0.3349,
+      "step": 42657
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.673937501049486e-06,
+      "loss": 0.2052,
+      "step": 42658
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.67360166574312e-06,
+      "loss": 0.269,
+      "step": 42659
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.673265830436754e-06,
+      "loss": 0.1631,
+      "step": 42660
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.672929995130389e-06,
+      "loss": 0.2772,
+      "step": 42661
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.672594159824023e-06,
+      "loss": 0.224,
+      "step": 42662
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6722583245176565e-06,
+      "loss": 0.1978,
+      "step": 42663
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.671922489211291e-06,
+      "loss": 0.2683,
+      "step": 42664
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.671586653904926e-06,
+      "loss": 0.2693,
+      "step": 42665
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6712508185985596e-06,
+      "loss": 0.2714,
+      "step": 42666
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.670914983292194e-06,
+      "loss": 0.2157,
+      "step": 42667
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.670579147985829e-06,
+      "loss": 0.3322,
+      "step": 42668
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.670243312679462e-06,
+      "loss": 0.2167,
+      "step": 42669
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6699074773730965e-06,
+      "loss": 0.1665,
+      "step": 42670
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.669571642066731e-06,
+      "loss": 0.155,
+      "step": 42671
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.669235806760365e-06,
+      "loss": 0.1699,
+      "step": 42672
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6688999714539995e-06,
+      "loss": 0.3393,
+      "step": 42673
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.668564136147634e-06,
+      "loss": 0.191,
+      "step": 42674
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.668228300841267e-06,
+      "loss": 0.1766,
+      "step": 42675
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.667892465534902e-06,
+      "loss": 0.1952,
+      "step": 42676
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6675566302285364e-06,
+      "loss": 0.1522,
+      "step": 42677
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.66722079492217e-06,
+      "loss": 0.2376,
+      "step": 42678
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.666884959615805e-06,
+      "loss": 0.2189,
+      "step": 42679
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6665491243094395e-06,
+      "loss": 0.3093,
+      "step": 42680
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.666213289003073e-06,
+      "loss": 0.337,
+      "step": 42681
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.665877453696708e-06,
+      "loss": 0.2082,
+      "step": 42682
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.665541618390343e-06,
+      "loss": 0.3151,
+      "step": 42683
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.665205783083976e-06,
+      "loss": 0.3,
+      "step": 42684
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.66486994777761e-06,
+      "loss": 0.178,
+      "step": 42685
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.664534112471245e-06,
+      "loss": 0.2238,
+      "step": 42686
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.664198277164879e-06,
+      "loss": 0.1779,
+      "step": 42687
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.663862441858513e-06,
+      "loss": 0.1977,
+      "step": 42688
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.663526606552148e-06,
+      "loss": 0.2579,
+      "step": 42689
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.663190771245781e-06,
+      "loss": 0.1778,
+      "step": 42690
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.662854935939416e-06,
+      "loss": 0.3171,
+      "step": 42691
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.66251910063305e-06,
+      "loss": 0.2312,
+      "step": 42692
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.662183265326684e-06,
+      "loss": 0.5889,
+      "step": 42693
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.661847430020319e-06,
+      "loss": 0.2595,
+      "step": 42694
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.661511594713953e-06,
+      "loss": 0.3832,
+      "step": 42695
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.661175759407586e-06,
+      "loss": 0.3569,
+      "step": 42696
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.660839924101221e-06,
+      "loss": 0.3559,
+      "step": 42697
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6605040887948556e-06,
+      "loss": 0.2694,
+      "step": 42698
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.660168253488489e-06,
+      "loss": 0.1702,
+      "step": 42699
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.659832418182124e-06,
+      "loss": 0.356,
+      "step": 42700
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.659496582875759e-06,
+      "loss": 0.311,
+      "step": 42701
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6591607475693925e-06,
+      "loss": 0.4809,
+      "step": 42702
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.658824912263027e-06,
+      "loss": 0.3407,
+      "step": 42703
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.658489076956662e-06,
+      "loss": 0.2512,
+      "step": 42704
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.658153241650295e-06,
+      "loss": 0.3137,
+      "step": 42705
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.657817406343929e-06,
+      "loss": 0.1886,
+      "step": 42706
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.657481571037564e-06,
+      "loss": 0.2531,
+      "step": 42707
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.657145735731198e-06,
+      "loss": 0.2565,
+      "step": 42708
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6568099004248325e-06,
+      "loss": 0.2567,
+      "step": 42709
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.656474065118467e-06,
+      "loss": 0.1952,
+      "step": 42710
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6561382298121e-06,
+      "loss": 0.2131,
+      "step": 42711
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.655802394505735e-06,
+      "loss": 0.2937,
+      "step": 42712
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.655466559199369e-06,
+      "loss": 0.2662,
+      "step": 42713
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.655130723893003e-06,
+      "loss": 0.1716,
+      "step": 42714
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.654794888586638e-06,
+      "loss": 0.2666,
+      "step": 42715
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6544590532802724e-06,
+      "loss": 0.3578,
+      "step": 42716
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.654123217973906e-06,
+      "loss": 0.1709,
+      "step": 42717
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.65378738266754e-06,
+      "loss": 0.3541,
+      "step": 42718
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.653451547361175e-06,
+      "loss": 0.3545,
+      "step": 42719
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6531157120548085e-06,
+      "loss": 0.2678,
+      "step": 42720
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.652779876748443e-06,
+      "loss": 0.4067,
+      "step": 42721
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.652444041442078e-06,
+      "loss": 0.1704,
+      "step": 42722
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.652108206135712e-06,
+      "loss": 0.3345,
+      "step": 42723
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.651772370829346e-06,
+      "loss": 0.3037,
+      "step": 42724
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.651436535522979e-06,
+      "loss": 0.176,
+      "step": 42725
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.651100700216614e-06,
+      "loss": 0.4189,
+      "step": 42726
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6507648649102485e-06,
+      "loss": 0.3559,
+      "step": 42727
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.650429029603882e-06,
+      "loss": 0.1555,
+      "step": 42728
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.650093194297517e-06,
+      "loss": 0.1682,
+      "step": 42729
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.649757358991152e-06,
+      "loss": 0.3509,
+      "step": 42730
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6494215236847845e-06,
+      "loss": 0.219,
+      "step": 42731
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.649085688378419e-06,
+      "loss": 0.2532,
+      "step": 42732
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.648749853072054e-06,
+      "loss": 0.2657,
+      "step": 42733
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.648414017765688e-06,
+      "loss": 0.296,
+      "step": 42734
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.648078182459322e-06,
+      "loss": 0.3639,
+      "step": 42735
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.647742347152957e-06,
+      "loss": 0.4518,
+      "step": 42736
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.647406511846591e-06,
+      "loss": 0.2916,
+      "step": 42737
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.647070676540225e-06,
+      "loss": 0.3006,
+      "step": 42738
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.64673484123386e-06,
+      "loss": 0.3118,
+      "step": 42739
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.646399005927493e-06,
+      "loss": 0.2164,
+      "step": 42740
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.646063170621128e-06,
+      "loss": 0.1627,
+      "step": 42741
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.645727335314762e-06,
+      "loss": 0.371,
+      "step": 42742
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.645391500008396e-06,
+      "loss": 0.4553,
+      "step": 42743
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.645055664702031e-06,
+      "loss": 0.261,
+      "step": 42744
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.644719829395665e-06,
+      "loss": 0.2472,
+      "step": 42745
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.644383994089298e-06,
+      "loss": 0.1772,
+      "step": 42746
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.644048158782933e-06,
+      "loss": 0.2936,
+      "step": 42747
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.643712323476568e-06,
+      "loss": 0.2566,
+      "step": 42748
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.643376488170201e-06,
+      "loss": 0.286,
+      "step": 42749
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.643040652863836e-06,
+      "loss": 0.2428,
+      "step": 42750
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.642704817557471e-06,
+      "loss": 0.3545,
+      "step": 42751
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.642368982251104e-06,
+      "loss": 0.175,
+      "step": 42752
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.642033146944738e-06,
+      "loss": 0.2662,
+      "step": 42753
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.641697311638373e-06,
+      "loss": 0.2722,
+      "step": 42754
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.641361476332007e-06,
+      "loss": 0.1236,
+      "step": 42755
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.641025641025641e-06,
+      "loss": 0.3149,
+      "step": 42756
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.640689805719276e-06,
+      "loss": 0.2693,
+      "step": 42757
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.64035397041291e-06,
+      "loss": 0.3023,
+      "step": 42758
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6400181351065445e-06,
+      "loss": 0.2934,
+      "step": 42759
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.639682299800179e-06,
+      "loss": 0.2563,
+      "step": 42760
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.639346464493812e-06,
+      "loss": 0.2817,
+      "step": 42761
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.639010629187447e-06,
+      "loss": 0.1911,
+      "step": 42762
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.638674793881081e-06,
+      "loss": 0.2638,
+      "step": 42763
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.638338958574715e-06,
+      "loss": 0.2518,
+      "step": 42764
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.63800312326835e-06,
+      "loss": 0.2942,
+      "step": 42765
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6376672879619845e-06,
+      "loss": 0.2623,
+      "step": 42766
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6373314526556174e-06,
+      "loss": 0.2262,
+      "step": 42767
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.636995617349252e-06,
+      "loss": 0.4297,
+      "step": 42768
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.636659782042887e-06,
+      "loss": 0.245,
+      "step": 42769
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6363239467365205e-06,
+      "loss": 0.1316,
+      "step": 42770
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.635988111430155e-06,
+      "loss": 0.3112,
+      "step": 42771
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.63565227612379e-06,
+      "loss": 0.2872,
+      "step": 42772
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.635316440817424e-06,
+      "loss": 0.1397,
+      "step": 42773
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.634980605511058e-06,
+      "loss": 0.2416,
+      "step": 42774
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.634644770204692e-06,
+      "loss": 0.2157,
+      "step": 42775
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.634308934898326e-06,
+      "loss": 0.2794,
+      "step": 42776
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.6339730995919605e-06,
+      "loss": 0.2635,
+      "step": 42777
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 5.633637264285595e-06,
+      "loss": 0.2953,
+      "step": 42778
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.633301428979229e-06,
+      "loss": 0.172,
+      "step": 42779
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.632965593672864e-06,
+      "loss": 0.2808,
+      "step": 42780
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.632629758366498e-06,
+      "loss": 0.4041,
+      "step": 42781
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.632293923060131e-06,
+      "loss": 0.3671,
+      "step": 42782
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.631958087753766e-06,
+      "loss": 0.3299,
+      "step": 42783
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.6316222524474005e-06,
+      "loss": 0.2312,
+      "step": 42784
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.631286417141034e-06,
+      "loss": 0.2385,
+      "step": 42785
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.630950581834669e-06,
+      "loss": 0.2986,
+      "step": 42786
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.630614746528304e-06,
+      "loss": 0.446,
+      "step": 42787
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.6302789112219366e-06,
+      "loss": 0.378,
+      "step": 42788
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.629943075915571e-06,
+      "loss": 0.2862,
+      "step": 42789
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.629607240609206e-06,
+      "loss": 0.2379,
+      "step": 42790
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.62927140530284e-06,
+      "loss": 0.1973,
+      "step": 42791
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.628935569996474e-06,
+      "loss": 0.3653,
+      "step": 42792
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.628599734690109e-06,
+      "loss": 0.4775,
+      "step": 42793
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.628263899383743e-06,
+      "loss": 0.3571,
+      "step": 42794
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.627928064077377e-06,
+      "loss": 0.3061,
+      "step": 42795
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.627592228771012e-06,
+      "loss": 0.1214,
+      "step": 42796
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.627256393464645e-06,
+      "loss": 0.2541,
+      "step": 42797
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.62692055815828e-06,
+      "loss": 0.2353,
+      "step": 42798
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.626584722851914e-06,
+      "loss": 0.2377,
+      "step": 42799
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.626248887545548e-06,
+      "loss": 0.2346,
+      "step": 42800
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.625913052239183e-06,
+      "loss": 0.3165,
+      "step": 42801
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.625577216932817e-06,
+      "loss": 0.3527,
+      "step": 42802
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.62524138162645e-06,
+      "loss": 0.2586,
+      "step": 42803
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.624905546320085e-06,
+      "loss": 0.2897,
+      "step": 42804
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.62456971101372e-06,
+      "loss": 0.3528,
+      "step": 42805
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.6242338757073534e-06,
+      "loss": 0.3103,
+      "step": 42806
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.623898040400988e-06,
+      "loss": 0.2559,
+      "step": 42807
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.623562205094623e-06,
+      "loss": 0.2486,
+      "step": 42808
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.623226369788256e-06,
+      "loss": 0.2075,
+      "step": 42809
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.62289053448189e-06,
+      "loss": 0.3342,
+      "step": 42810
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.622554699175525e-06,
+      "loss": 0.14,
+      "step": 42811
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.622218863869159e-06,
+      "loss": 0.1842,
+      "step": 42812
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.6218830285627934e-06,
+      "loss": 0.2334,
+      "step": 42813
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.621547193256428e-06,
+      "loss": 0.2232,
+      "step": 42814
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.621211357950062e-06,
+      "loss": 0.1693,
+      "step": 42815
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.6208755226436965e-06,
+      "loss": 0.1573,
+      "step": 42816
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.6205396873373295e-06,
+      "loss": 0.3576,
+      "step": 42817
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.620203852030964e-06,
+      "loss": 0.2443,
+      "step": 42818
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.619868016724599e-06,
+      "loss": 0.2437,
+      "step": 42819
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.6195321814182326e-06,
+      "loss": 0.1338,
+      "step": 42820
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.619196346111867e-06,
+      "loss": 0.2977,
+      "step": 42821
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.618860510805502e-06,
+      "loss": 0.3745,
+      "step": 42822
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.618524675499135e-06,
+      "loss": 0.2129,
+      "step": 42823
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.6181888401927695e-06,
+      "loss": 0.2219,
+      "step": 42824
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.617853004886404e-06,
+      "loss": 0.3096,
+      "step": 42825
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.617517169580038e-06,
+      "loss": 0.3248,
+      "step": 42826
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.6171813342736726e-06,
+      "loss": 0.3079,
+      "step": 42827
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.616845498967307e-06,
+      "loss": 0.1857,
+      "step": 42828
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.616509663660941e-06,
+      "loss": 0.2269,
+      "step": 42829
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.616173828354576e-06,
+      "loss": 0.074,
+      "step": 42830
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.6158379930482095e-06,
+      "loss": 0.2405,
+      "step": 42831
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.615502157741843e-06,
+      "loss": 0.1123,
+      "step": 42832
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.615166322435478e-06,
+      "loss": 0.1813,
+      "step": 42833
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.6148304871291126e-06,
+      "loss": 0.2814,
+      "step": 42834
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.614494651822746e-06,
+      "loss": 0.4637,
+      "step": 42835
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.614158816516381e-06,
+      "loss": 0.2419,
+      "step": 42836
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.613822981210016e-06,
+      "loss": 0.3053,
+      "step": 42837
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.613487145903649e-06,
+      "loss": 0.3835,
+      "step": 42838
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.613151310597283e-06,
+      "loss": 0.3859,
+      "step": 42839
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.612815475290918e-06,
+      "loss": 0.2884,
+      "step": 42840
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.612479639984552e-06,
+      "loss": 0.1859,
+      "step": 42841
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.612143804678186e-06,
+      "loss": 0.2066,
+      "step": 42842
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.611807969371821e-06,
+      "loss": 0.3695,
+      "step": 42843
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.611472134065454e-06,
+      "loss": 0.3086,
+      "step": 42844
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.611136298759089e-06,
+      "loss": 0.3294,
+      "step": 42845
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.610800463452723e-06,
+      "loss": 0.2839,
+      "step": 42846
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.610464628146357e-06,
+      "loss": 0.2688,
+      "step": 42847
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.610128792839992e-06,
+      "loss": 0.1058,
+      "step": 42848
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.609792957533626e-06,
+      "loss": 0.3146,
+      "step": 42849
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.60945712222726e-06,
+      "loss": 0.1729,
+      "step": 42850
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.609121286920895e-06,
+      "loss": 0.2557,
+      "step": 42851
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.6087854516145294e-06,
+      "loss": 0.2979,
+      "step": 42852
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.608449616308162e-06,
+      "loss": 0.3449,
+      "step": 42853
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.608113781001797e-06,
+      "loss": 0.2443,
+      "step": 42854
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.607777945695432e-06,
+      "loss": 0.3188,
+      "step": 42855
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.6074421103890655e-06,
+      "loss": 0.2155,
+      "step": 42856
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.6071062750827e-06,
+      "loss": 0.3199,
+      "step": 42857
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.606770439776335e-06,
+      "loss": 0.2418,
+      "step": 42858
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.606434604469968e-06,
+      "loss": 0.3516,
+      "step": 42859
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.606098769163602e-06,
+      "loss": 0.3333,
+      "step": 42860
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.605762933857237e-06,
+      "loss": 0.2392,
+      "step": 42861
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.605427098550871e-06,
+      "loss": 0.2664,
+      "step": 42862
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.6050912632445055e-06,
+      "loss": 0.1966,
+      "step": 42863
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.60475542793814e-06,
+      "loss": 0.3025,
+      "step": 42864
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.604419592631773e-06,
+      "loss": 0.1977,
+      "step": 42865
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.604083757325408e-06,
+      "loss": 0.3199,
+      "step": 42866
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.603747922019042e-06,
+      "loss": 0.3436,
+      "step": 42867
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.603412086712676e-06,
+      "loss": 0.582,
+      "step": 42868
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.603076251406311e-06,
+      "loss": 0.283,
+      "step": 42869
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.6027404160999455e-06,
+      "loss": 0.1676,
+      "step": 42870
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.602404580793579e-06,
+      "loss": 0.2233,
+      "step": 42871
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.602068745487214e-06,
+      "loss": 0.2675,
+      "step": 42872
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.6017329101808486e-06,
+      "loss": 0.1544,
+      "step": 42873
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.6013970748744815e-06,
+      "loss": 0.2124,
+      "step": 42874
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.601061239568116e-06,
+      "loss": 0.1455,
+      "step": 42875
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.600725404261751e-06,
+      "loss": 0.3057,
+      "step": 42876
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.600389568955385e-06,
+      "loss": 0.25,
+      "step": 42877
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.600053733649019e-06,
+      "loss": 0.3385,
+      "step": 42878
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.599717898342654e-06,
+      "loss": 0.4724,
+      "step": 42879
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.599382063036287e-06,
+      "loss": 0.1971,
+      "step": 42880
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.5990462277299215e-06,
+      "loss": 0.3612,
+      "step": 42881
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.598710392423556e-06,
+      "loss": 0.2284,
+      "step": 42882
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.59837455711719e-06,
+      "loss": 0.214,
+      "step": 42883
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.598038721810825e-06,
+      "loss": 0.3956,
+      "step": 42884
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.597702886504459e-06,
+      "loss": 0.3365,
+      "step": 42885
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.597367051198093e-06,
+      "loss": 0.4504,
+      "step": 42886
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.597031215891727e-06,
+      "loss": 0.2662,
+      "step": 42887
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.5966953805853615e-06,
+      "loss": 0.2387,
+      "step": 42888
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.596359545278995e-06,
+      "loss": 0.3919,
+      "step": 42889
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.59602370997263e-06,
+      "loss": 0.2431,
+      "step": 42890
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.595687874666265e-06,
+      "loss": 0.2241,
+      "step": 42891
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.595352039359898e-06,
+      "loss": 0.4722,
+      "step": 42892
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.595016204053533e-06,
+      "loss": 0.3583,
+      "step": 42893
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.594680368747168e-06,
+      "loss": 0.2892,
+      "step": 42894
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.594344533440801e-06,
+      "loss": 0.2704,
+      "step": 42895
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.594008698134435e-06,
+      "loss": 0.2865,
+      "step": 42896
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.59367286282807e-06,
+      "loss": 0.3055,
+      "step": 42897
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.593337027521704e-06,
+      "loss": 0.3932,
+      "step": 42898
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.593001192215338e-06,
+      "loss": 0.3452,
+      "step": 42899
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.592665356908973e-06,
+      "loss": 0.1488,
+      "step": 42900
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.592329521602606e-06,
+      "loss": 0.2913,
+      "step": 42901
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.591993686296241e-06,
+      "loss": 0.2335,
+      "step": 42902
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.591657850989875e-06,
+      "loss": 0.1935,
+      "step": 42903
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.591322015683509e-06,
+      "loss": 0.3208,
+      "step": 42904
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.590986180377144e-06,
+      "loss": 0.3493,
+      "step": 42905
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.5906503450707775e-06,
+      "loss": 0.1783,
+      "step": 42906
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.590314509764412e-06,
+      "loss": 0.2684,
+      "step": 42907
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.589978674458047e-06,
+      "loss": 0.2844,
+      "step": 42908
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.58964283915168e-06,
+      "loss": 0.3968,
+      "step": 42909
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.589307003845314e-06,
+      "loss": 0.141,
+      "step": 42910
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.588971168538949e-06,
+      "loss": 0.2573,
+      "step": 42911
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.588635333232583e-06,
+      "loss": 0.1643,
+      "step": 42912
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.5882994979262175e-06,
+      "loss": 0.4246,
+      "step": 42913
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.587963662619852e-06,
+      "loss": 0.3559,
+      "step": 42914
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.587627827313485e-06,
+      "loss": 0.1434,
+      "step": 42915
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.58729199200712e-06,
+      "loss": 0.4296,
+      "step": 42916
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.586956156700754e-06,
+      "loss": 0.2628,
+      "step": 42917
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.586620321394388e-06,
+      "loss": 0.298,
+      "step": 42918
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.586284486088023e-06,
+      "loss": 0.2973,
+      "step": 42919
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.5859486507816575e-06,
+      "loss": 0.2262,
+      "step": 42920
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.585612815475291e-06,
+      "loss": 0.1791,
+      "step": 42921
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.585276980168925e-06,
+      "loss": 0.3206,
+      "step": 42922
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.58494114486256e-06,
+      "loss": 0.2956,
+      "step": 42923
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.5846053095561936e-06,
+      "loss": 0.3473,
+      "step": 42924
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.584269474249828e-06,
+      "loss": 0.2501,
+      "step": 42925
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.583933638943463e-06,
+      "loss": 0.2736,
+      "step": 42926
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.583597803637097e-06,
+      "loss": 0.3806,
+      "step": 42927
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.583261968330731e-06,
+      "loss": 0.3166,
+      "step": 42928
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.582926133024366e-06,
+      "loss": 0.3307,
+      "step": 42929
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.582590297717999e-06,
+      "loss": 0.4465,
+      "step": 42930
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.5822544624116335e-06,
+      "loss": 0.3342,
+      "step": 42931
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.581918627105268e-06,
+      "loss": 0.2673,
+      "step": 42932
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.581582791798902e-06,
+      "loss": 0.1855,
+      "step": 42933
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.581246956492537e-06,
+      "loss": 0.1908,
+      "step": 42934
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.580911121186171e-06,
+      "loss": 0.3733,
+      "step": 42935
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.580575285879804e-06,
+      "loss": 0.1744,
+      "step": 42936
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.580239450573439e-06,
+      "loss": 0.183,
+      "step": 42937
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.5799036152670735e-06,
+      "loss": 0.257,
+      "step": 42938
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.579567779960707e-06,
+      "loss": 0.2627,
+      "step": 42939
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.579231944654342e-06,
+      "loss": 0.1736,
+      "step": 42940
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.578896109347977e-06,
+      "loss": 0.3688,
+      "step": 42941
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.5785602740416104e-06,
+      "loss": 0.2666,
+      "step": 42942
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.578224438735245e-06,
+      "loss": 0.1232,
+      "step": 42943
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.577888603428879e-06,
+      "loss": 0.3502,
+      "step": 42944
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.577552768122513e-06,
+      "loss": 0.2502,
+      "step": 42945
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.577216932816147e-06,
+      "loss": 0.2746,
+      "step": 42946
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.576881097509782e-06,
+      "loss": 0.2703,
+      "step": 42947
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.576545262203416e-06,
+      "loss": 0.192,
+      "step": 42948
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.57620942689705e-06,
+      "loss": 0.2309,
+      "step": 42949
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.575873591590685e-06,
+      "loss": 0.2184,
+      "step": 42950
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.575537756284318e-06,
+      "loss": 0.3116,
+      "step": 42951
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.575201920977953e-06,
+      "loss": 0.1613,
+      "step": 42952
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.574866085671587e-06,
+      "loss": 0.2744,
+      "step": 42953
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.574530250365221e-06,
+      "loss": 0.49,
+      "step": 42954
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.574194415058856e-06,
+      "loss": 0.3182,
+      "step": 42955
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.57385857975249e-06,
+      "loss": 0.2962,
+      "step": 42956
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.573522744446123e-06,
+      "loss": 0.2551,
+      "step": 42957
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.573186909139758e-06,
+      "loss": 0.3052,
+      "step": 42958
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.572851073833393e-06,
+      "loss": 0.193,
+      "step": 42959
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.5725152385270265e-06,
+      "loss": 0.3656,
+      "step": 42960
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.572179403220661e-06,
+      "loss": 0.3497,
+      "step": 42961
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.571843567914296e-06,
+      "loss": 0.3135,
+      "step": 42962
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.5715077326079296e-06,
+      "loss": 0.2156,
+      "step": 42963
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.571171897301564e-06,
+      "loss": 0.224,
+      "step": 42964
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.570836061995199e-06,
+      "loss": 0.2947,
+      "step": 42965
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.570500226688832e-06,
+      "loss": 0.2771,
+      "step": 42966
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.5701643913824664e-06,
+      "loss": 0.347,
+      "step": 42967
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.569828556076101e-06,
+      "loss": 0.2553,
+      "step": 42968
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.569492720769735e-06,
+      "loss": 0.1466,
+      "step": 42969
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.5691568854633695e-06,
+      "loss": 0.3486,
+      "step": 42970
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.568821050157004e-06,
+      "loss": 0.2886,
+      "step": 42971
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.568485214850637e-06,
+      "loss": 0.2738,
+      "step": 42972
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.568149379544272e-06,
+      "loss": 0.4201,
+      "step": 42973
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.5678135442379064e-06,
+      "loss": 0.1837,
+      "step": 42974
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.56747770893154e-06,
+      "loss": 0.2273,
+      "step": 42975
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.567141873625175e-06,
+      "loss": 0.2518,
+      "step": 42976
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 5.5668060383188095e-06,
+      "loss": 0.1962,
+      "step": 42977
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5664702030124425e-06,
+      "loss": 0.3299,
+      "step": 42978
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.566134367706077e-06,
+      "loss": 0.2847,
+      "step": 42979
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.565798532399712e-06,
+      "loss": 0.1663,
+      "step": 42980
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.565462697093346e-06,
+      "loss": 0.3049,
+      "step": 42981
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.56512686178698e-06,
+      "loss": 0.3144,
+      "step": 42982
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.564791026480615e-06,
+      "loss": 0.244,
+      "step": 42983
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.564455191174249e-06,
+      "loss": 0.2023,
+      "step": 42984
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.564119355867883e-06,
+      "loss": 0.3906,
+      "step": 42985
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.563783520561518e-06,
+      "loss": 0.2956,
+      "step": 42986
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.563447685255151e-06,
+      "loss": 0.3034,
+      "step": 42987
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5631118499487856e-06,
+      "loss": 0.2687,
+      "step": 42988
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.56277601464242e-06,
+      "loss": 0.1942,
+      "step": 42989
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.562440179336054e-06,
+      "loss": 0.3553,
+      "step": 42990
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.562104344029689e-06,
+      "loss": 0.379,
+      "step": 42991
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.561768508723323e-06,
+      "loss": 0.3058,
+      "step": 42992
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.561432673416956e-06,
+      "loss": 0.1937,
+      "step": 42993
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.561096838110591e-06,
+      "loss": 0.2398,
+      "step": 42994
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5607610028042256e-06,
+      "loss": 0.2069,
+      "step": 42995
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.560425167497859e-06,
+      "loss": 0.2464,
+      "step": 42996
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.560089332191494e-06,
+      "loss": 0.3245,
+      "step": 42997
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.559753496885128e-06,
+      "loss": 0.2231,
+      "step": 42998
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5594176615787625e-06,
+      "loss": 0.2616,
+      "step": 42999
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.559081826272396e-06,
+      "loss": 0.2646,
+      "step": 43000
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.55874599096603e-06,
+      "loss": 0.2145,
+      "step": 43001
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.558410155659665e-06,
+      "loss": 0.2922,
+      "step": 43002
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.558074320353299e-06,
+      "loss": 0.2303,
+      "step": 43003
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.557738485046933e-06,
+      "loss": 0.2519,
+      "step": 43004
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.557402649740568e-06,
+      "loss": 0.3461,
+      "step": 43005
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5570668144342024e-06,
+      "loss": 0.2736,
+      "step": 43006
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.556730979127835e-06,
+      "loss": 0.2383,
+      "step": 43007
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.55639514382147e-06,
+      "loss": 0.2588,
+      "step": 43008
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.556059308515105e-06,
+      "loss": 0.2922,
+      "step": 43009
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5557234732087385e-06,
+      "loss": 0.173,
+      "step": 43010
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.555387637902373e-06,
+      "loss": 0.2399,
+      "step": 43011
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.555051802596008e-06,
+      "loss": 0.2609,
+      "step": 43012
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.554715967289641e-06,
+      "loss": 0.266,
+      "step": 43013
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.554380131983275e-06,
+      "loss": 0.1797,
+      "step": 43014
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.55404429667691e-06,
+      "loss": 0.3444,
+      "step": 43015
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.553708461370544e-06,
+      "loss": 0.12,
+      "step": 43016
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5533726260641785e-06,
+      "loss": 0.266,
+      "step": 43017
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.553036790757813e-06,
+      "loss": 0.1747,
+      "step": 43018
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.552700955451447e-06,
+      "loss": 0.221,
+      "step": 43019
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.552365120145082e-06,
+      "loss": 0.3779,
+      "step": 43020
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.552029284838716e-06,
+      "loss": 0.2285,
+      "step": 43021
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.551693449532349e-06,
+      "loss": 0.3312,
+      "step": 43022
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.551357614225984e-06,
+      "loss": 0.2001,
+      "step": 43023
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5510217789196185e-06,
+      "loss": 0.2265,
+      "step": 43024
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.550685943613252e-06,
+      "loss": 0.3241,
+      "step": 43025
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.550350108306887e-06,
+      "loss": 0.2749,
+      "step": 43026
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5500142730005216e-06,
+      "loss": 0.3796,
+      "step": 43027
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5496784376941545e-06,
+      "loss": 0.2294,
+      "step": 43028
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.549342602387789e-06,
+      "loss": 0.4155,
+      "step": 43029
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.549006767081424e-06,
+      "loss": 0.2106,
+      "step": 43030
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.548670931775058e-06,
+      "loss": 0.2697,
+      "step": 43031
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.548335096468692e-06,
+      "loss": 0.2066,
+      "step": 43032
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.547999261162327e-06,
+      "loss": 0.244,
+      "step": 43033
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.54766342585596e-06,
+      "loss": 0.1091,
+      "step": 43034
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5473275905495945e-06,
+      "loss": 0.2553,
+      "step": 43035
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.546991755243229e-06,
+      "loss": 0.2982,
+      "step": 43036
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.546655919936863e-06,
+      "loss": 0.1998,
+      "step": 43037
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.546320084630498e-06,
+      "loss": 0.3224,
+      "step": 43038
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.545984249324132e-06,
+      "loss": 0.202,
+      "step": 43039
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.545648414017766e-06,
+      "loss": 0.2884,
+      "step": 43040
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.545312578711401e-06,
+      "loss": 0.1599,
+      "step": 43041
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.544976743405035e-06,
+      "loss": 0.1097,
+      "step": 43042
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.544640908098668e-06,
+      "loss": 0.2007,
+      "step": 43043
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.544305072792303e-06,
+      "loss": 0.2553,
+      "step": 43044
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.543969237485938e-06,
+      "loss": 0.4178,
+      "step": 43045
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.543633402179571e-06,
+      "loss": 0.2757,
+      "step": 43046
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.543297566873206e-06,
+      "loss": 0.4354,
+      "step": 43047
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.542961731566841e-06,
+      "loss": 0.1785,
+      "step": 43048
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.542625896260474e-06,
+      "loss": 0.4054,
+      "step": 43049
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.542290060954108e-06,
+      "loss": 0.3438,
+      "step": 43050
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.541954225647743e-06,
+      "loss": 0.2707,
+      "step": 43051
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.541618390341377e-06,
+      "loss": 0.1614,
+      "step": 43052
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.541282555035011e-06,
+      "loss": 0.3107,
+      "step": 43053
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.540946719728646e-06,
+      "loss": 0.2698,
+      "step": 43054
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.54061088442228e-06,
+      "loss": 0.2419,
+      "step": 43055
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5402750491159145e-06,
+      "loss": 0.3068,
+      "step": 43056
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.539939213809548e-06,
+      "loss": 0.2884,
+      "step": 43057
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.539603378503182e-06,
+      "loss": 0.3852,
+      "step": 43058
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.539267543196817e-06,
+      "loss": 0.2605,
+      "step": 43059
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.538931707890451e-06,
+      "loss": 0.3904,
+      "step": 43060
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.538595872584085e-06,
+      "loss": 0.3547,
+      "step": 43061
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.53826003727772e-06,
+      "loss": 0.2109,
+      "step": 43062
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5379242019713545e-06,
+      "loss": 0.3113,
+      "step": 43063
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5375883666649874e-06,
+      "loss": 0.2342,
+      "step": 43064
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.537252531358622e-06,
+      "loss": 0.4339,
+      "step": 43065
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.536916696052257e-06,
+      "loss": 0.3614,
+      "step": 43066
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5365808607458905e-06,
+      "loss": 0.1836,
+      "step": 43067
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.536245025439525e-06,
+      "loss": 0.2053,
+      "step": 43068
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.53590919013316e-06,
+      "loss": 0.3091,
+      "step": 43069
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.535573354826793e-06,
+      "loss": 0.3373,
+      "step": 43070
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5352375195204274e-06,
+      "loss": 0.4018,
+      "step": 43071
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.534901684214062e-06,
+      "loss": 0.194,
+      "step": 43072
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.534565848907696e-06,
+      "loss": 0.2067,
+      "step": 43073
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5342300136013305e-06,
+      "loss": 0.1727,
+      "step": 43074
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.533894178294965e-06,
+      "loss": 0.2236,
+      "step": 43075
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.533558342988599e-06,
+      "loss": 0.1888,
+      "step": 43076
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.533222507682234e-06,
+      "loss": 0.278,
+      "step": 43077
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.532886672375868e-06,
+      "loss": 0.1688,
+      "step": 43078
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.532550837069501e-06,
+      "loss": 0.1658,
+      "step": 43079
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.532215001763136e-06,
+      "loss": 0.3507,
+      "step": 43080
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5318791664567705e-06,
+      "loss": 0.2672,
+      "step": 43081
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.531543331150404e-06,
+      "loss": 0.2885,
+      "step": 43082
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.531207495844039e-06,
+      "loss": 0.3734,
+      "step": 43083
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.530871660537674e-06,
+      "loss": 0.2817,
+      "step": 43084
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5305358252313066e-06,
+      "loss": 0.2906,
+      "step": 43085
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.530199989924941e-06,
+      "loss": 0.2759,
+      "step": 43086
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.529864154618575e-06,
+      "loss": 0.1794,
+      "step": 43087
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.52952831931221e-06,
+      "loss": 0.4489,
+      "step": 43088
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.529192484005844e-06,
+      "loss": 0.3188,
+      "step": 43089
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.528856648699478e-06,
+      "loss": 0.28,
+      "step": 43090
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.528520813393112e-06,
+      "loss": 0.395,
+      "step": 43091
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5281849780867465e-06,
+      "loss": 0.3111,
+      "step": 43092
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.52784914278038e-06,
+      "loss": 0.2626,
+      "step": 43093
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.527513307474015e-06,
+      "loss": 0.4153,
+      "step": 43094
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.52717747216765e-06,
+      "loss": 0.2926,
+      "step": 43095
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5268416368612834e-06,
+      "loss": 0.3661,
+      "step": 43096
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.526505801554918e-06,
+      "loss": 0.4,
+      "step": 43097
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.526169966248553e-06,
+      "loss": 0.324,
+      "step": 43098
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.525834130942186e-06,
+      "loss": 0.2005,
+      "step": 43099
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.52549829563582e-06,
+      "loss": 0.1675,
+      "step": 43100
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.525162460329455e-06,
+      "loss": 0.2166,
+      "step": 43101
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.524826625023089e-06,
+      "loss": 0.352,
+      "step": 43102
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5244907897167234e-06,
+      "loss": 0.1685,
+      "step": 43103
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.524154954410358e-06,
+      "loss": 0.238,
+      "step": 43104
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.523819119103991e-06,
+      "loss": 0.3769,
+      "step": 43105
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.523483283797626e-06,
+      "loss": 0.1882,
+      "step": 43106
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.52314744849126e-06,
+      "loss": 0.2044,
+      "step": 43107
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.522811613184894e-06,
+      "loss": 0.4122,
+      "step": 43108
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.522475777878529e-06,
+      "loss": 0.2009,
+      "step": 43109
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5221399425721634e-06,
+      "loss": 0.2125,
+      "step": 43110
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.521804107265797e-06,
+      "loss": 0.2938,
+      "step": 43111
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.521468271959432e-06,
+      "loss": 0.2968,
+      "step": 43112
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.521132436653066e-06,
+      "loss": 0.1726,
+      "step": 43113
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5207966013466995e-06,
+      "loss": 0.4506,
+      "step": 43114
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.520460766040334e-06,
+      "loss": 0.4574,
+      "step": 43115
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.520124930733969e-06,
+      "loss": 0.2339,
+      "step": 43116
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5197890954276026e-06,
+      "loss": 0.3946,
+      "step": 43117
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.519453260121237e-06,
+      "loss": 0.2062,
+      "step": 43118
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.519117424814872e-06,
+      "loss": 0.2398,
+      "step": 43119
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.518781589508505e-06,
+      "loss": 0.3156,
+      "step": 43120
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5184457542021395e-06,
+      "loss": 0.2082,
+      "step": 43121
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.518109918895774e-06,
+      "loss": 0.1752,
+      "step": 43122
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.517774083589408e-06,
+      "loss": 0.2269,
+      "step": 43123
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5174382482830426e-06,
+      "loss": 0.2486,
+      "step": 43124
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.517102412976677e-06,
+      "loss": 0.2602,
+      "step": 43125
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.51676657767031e-06,
+      "loss": 0.1788,
+      "step": 43126
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.516430742363945e-06,
+      "loss": 0.2034,
+      "step": 43127
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5160949070575795e-06,
+      "loss": 0.2142,
+      "step": 43128
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.515759071751213e-06,
+      "loss": 0.2968,
+      "step": 43129
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.515423236444848e-06,
+      "loss": 0.2623,
+      "step": 43130
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5150874011384826e-06,
+      "loss": 0.2718,
+      "step": 43131
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.514751565832116e-06,
+      "loss": 0.2917,
+      "step": 43132
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.514415730525751e-06,
+      "loss": 0.2188,
+      "step": 43133
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.514079895219386e-06,
+      "loss": 0.2863,
+      "step": 43134
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.513744059913019e-06,
+      "loss": 0.142,
+      "step": 43135
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.513408224606653e-06,
+      "loss": 0.1372,
+      "step": 43136
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.513072389300288e-06,
+      "loss": 0.2806,
+      "step": 43137
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.512736553993922e-06,
+      "loss": 0.1782,
+      "step": 43138
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.512400718687556e-06,
+      "loss": 0.3501,
+      "step": 43139
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.512064883381191e-06,
+      "loss": 0.3442,
+      "step": 43140
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.511729048074824e-06,
+      "loss": 0.236,
+      "step": 43141
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.511393212768459e-06,
+      "loss": 0.1857,
+      "step": 43142
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.511057377462093e-06,
+      "loss": 0.2053,
+      "step": 43143
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.510721542155727e-06,
+      "loss": 0.2716,
+      "step": 43144
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.510385706849362e-06,
+      "loss": 0.2289,
+      "step": 43145
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.510049871542996e-06,
+      "loss": 0.3703,
+      "step": 43146
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.509714036236629e-06,
+      "loss": 0.3187,
+      "step": 43147
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.509378200930264e-06,
+      "loss": 0.2462,
+      "step": 43148
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.509042365623899e-06,
+      "loss": 0.2664,
+      "step": 43149
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.508706530317532e-06,
+      "loss": 0.3312,
+      "step": 43150
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.508370695011167e-06,
+      "loss": 0.1007,
+      "step": 43151
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.508034859704802e-06,
+      "loss": 0.2425,
+      "step": 43152
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5076990243984355e-06,
+      "loss": 0.3672,
+      "step": 43153
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.50736318909207e-06,
+      "loss": 0.2163,
+      "step": 43154
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.507027353785705e-06,
+      "loss": 0.3231,
+      "step": 43155
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.506691518479338e-06,
+      "loss": 0.4821,
+      "step": 43156
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.506355683172972e-06,
+      "loss": 0.4336,
+      "step": 43157
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.506019847866607e-06,
+      "loss": 0.2655,
+      "step": 43158
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.505684012560241e-06,
+      "loss": 0.2647,
+      "step": 43159
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5053481772538755e-06,
+      "loss": 0.1946,
+      "step": 43160
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.50501234194751e-06,
+      "loss": 0.2608,
+      "step": 43161
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.504676506641143e-06,
+      "loss": 0.3771,
+      "step": 43162
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.504340671334778e-06,
+      "loss": 0.2705,
+      "step": 43163
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.504004836028412e-06,
+      "loss": 0.2231,
+      "step": 43164
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.503669000722046e-06,
+      "loss": 0.2683,
+      "step": 43165
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.503333165415681e-06,
+      "loss": 0.2806,
+      "step": 43166
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5029973301093155e-06,
+      "loss": 0.1678,
+      "step": 43167
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.502661494802949e-06,
+      "loss": 0.2554,
+      "step": 43168
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.502325659496584e-06,
+      "loss": 0.262,
+      "step": 43169
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.501989824190218e-06,
+      "loss": 0.247,
+      "step": 43170
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.5016539888838515e-06,
+      "loss": 0.1827,
+      "step": 43171
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.501318153577486e-06,
+      "loss": 0.2551,
+      "step": 43172
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.500982318271121e-06,
+      "loss": 0.3127,
+      "step": 43173
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.500646482964755e-06,
+      "loss": 0.2998,
+      "step": 43174
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 5.500310647658389e-06,
+      "loss": 0.3145,
+      "step": 43175
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.499974812352022e-06,
+      "loss": 0.2145,
+      "step": 43176
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.499638977045657e-06,
+      "loss": 0.2595,
+      "step": 43177
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4993031417392915e-06,
+      "loss": 0.1724,
+      "step": 43178
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.498967306432925e-06,
+      "loss": 0.2238,
+      "step": 43179
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.49863147112656e-06,
+      "loss": 0.3122,
+      "step": 43180
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.498295635820195e-06,
+      "loss": 0.3381,
+      "step": 43181
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4979598005138275e-06,
+      "loss": 0.2147,
+      "step": 43182
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.497623965207462e-06,
+      "loss": 0.2052,
+      "step": 43183
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.497288129901097e-06,
+      "loss": 0.3536,
+      "step": 43184
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.496952294594731e-06,
+      "loss": 0.3543,
+      "step": 43185
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.496616459288365e-06,
+      "loss": 0.2764,
+      "step": 43186
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.496280623982e-06,
+      "loss": 0.4335,
+      "step": 43187
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.495944788675634e-06,
+      "loss": 0.1169,
+      "step": 43188
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.495608953369268e-06,
+      "loss": 0.2855,
+      "step": 43189
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.495273118062903e-06,
+      "loss": 0.1442,
+      "step": 43190
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.494937282756536e-06,
+      "loss": 0.2196,
+      "step": 43191
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.494601447450171e-06,
+      "loss": 0.3687,
+      "step": 43192
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.494265612143805e-06,
+      "loss": 0.4565,
+      "step": 43193
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.493929776837439e-06,
+      "loss": 0.3082,
+      "step": 43194
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.493593941531074e-06,
+      "loss": 0.2548,
+      "step": 43195
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.493258106224708e-06,
+      "loss": 0.5842,
+      "step": 43196
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.492922270918341e-06,
+      "loss": 0.2726,
+      "step": 43197
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.492586435611976e-06,
+      "loss": 0.2438,
+      "step": 43198
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.492250600305611e-06,
+      "loss": 0.23,
+      "step": 43199
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.491914764999244e-06,
+      "loss": 0.1639,
+      "step": 43200
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.491578929692879e-06,
+      "loss": 0.1638,
+      "step": 43201
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.491243094386514e-06,
+      "loss": 0.1368,
+      "step": 43202
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4909072590801475e-06,
+      "loss": 0.2494,
+      "step": 43203
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.490571423773781e-06,
+      "loss": 0.2069,
+      "step": 43204
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.490235588467416e-06,
+      "loss": 0.1856,
+      "step": 43205
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.48989975316105e-06,
+      "loss": 0.2048,
+      "step": 43206
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.489563917854684e-06,
+      "loss": 0.1624,
+      "step": 43207
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.489228082548319e-06,
+      "loss": 0.3095,
+      "step": 43208
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.488892247241953e-06,
+      "loss": 0.395,
+      "step": 43209
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4885564119355875e-06,
+      "loss": 0.2794,
+      "step": 43210
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.488220576629222e-06,
+      "loss": 0.2732,
+      "step": 43211
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.487884741322855e-06,
+      "loss": 0.4344,
+      "step": 43212
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.48754890601649e-06,
+      "loss": 0.2497,
+      "step": 43213
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.487213070710124e-06,
+      "loss": 0.1951,
+      "step": 43214
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.486877235403758e-06,
+      "loss": 0.3863,
+      "step": 43215
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.486541400097393e-06,
+      "loss": 0.2555,
+      "step": 43216
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4862055647910275e-06,
+      "loss": 0.1247,
+      "step": 43217
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4858697294846605e-06,
+      "loss": 0.1649,
+      "step": 43218
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.485533894178295e-06,
+      "loss": 0.1296,
+      "step": 43219
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.48519805887193e-06,
+      "loss": 0.3464,
+      "step": 43220
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4848622235655635e-06,
+      "loss": 0.1405,
+      "step": 43221
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.484526388259198e-06,
+      "loss": 0.2216,
+      "step": 43222
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.484190552952833e-06,
+      "loss": 0.2717,
+      "step": 43223
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.483854717646467e-06,
+      "loss": 0.1184,
+      "step": 43224
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.483518882340101e-06,
+      "loss": 0.2385,
+      "step": 43225
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.483183047033735e-06,
+      "loss": 0.1822,
+      "step": 43226
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.482847211727369e-06,
+      "loss": 0.3295,
+      "step": 43227
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4825113764210035e-06,
+      "loss": 0.197,
+      "step": 43228
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.482175541114638e-06,
+      "loss": 0.1138,
+      "step": 43229
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.481839705808272e-06,
+      "loss": 0.2894,
+      "step": 43230
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.481503870501907e-06,
+      "loss": 0.2159,
+      "step": 43231
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.481168035195541e-06,
+      "loss": 0.4944,
+      "step": 43232
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.480832199889174e-06,
+      "loss": 0.2563,
+      "step": 43233
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.480496364582809e-06,
+      "loss": 0.1726,
+      "step": 43234
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4801605292764435e-06,
+      "loss": 0.2355,
+      "step": 43235
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.479824693970077e-06,
+      "loss": 0.1739,
+      "step": 43236
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.479488858663712e-06,
+      "loss": 0.2154,
+      "step": 43237
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.479153023357347e-06,
+      "loss": 0.2784,
+      "step": 43238
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4788171880509796e-06,
+      "loss": 0.2571,
+      "step": 43239
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.478481352744614e-06,
+      "loss": 0.2808,
+      "step": 43240
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.478145517438249e-06,
+      "loss": 0.4234,
+      "step": 43241
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.477809682131883e-06,
+      "loss": 0.19,
+      "step": 43242
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.477473846825517e-06,
+      "loss": 0.3172,
+      "step": 43243
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.477138011519152e-06,
+      "loss": 0.2142,
+      "step": 43244
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.476802176212786e-06,
+      "loss": 0.2281,
+      "step": 43245
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.47646634090642e-06,
+      "loss": 0.3648,
+      "step": 43246
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.476130505600055e-06,
+      "loss": 0.2995,
+      "step": 43247
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.475794670293688e-06,
+      "loss": 0.3884,
+      "step": 43248
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.475458834987323e-06,
+      "loss": 0.2517,
+      "step": 43249
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.475122999680957e-06,
+      "loss": 0.1961,
+      "step": 43250
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.474787164374591e-06,
+      "loss": 0.1866,
+      "step": 43251
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.474451329068226e-06,
+      "loss": 0.155,
+      "step": 43252
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.47411549376186e-06,
+      "loss": 0.2091,
+      "step": 43253
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.473779658455493e-06,
+      "loss": 0.2517,
+      "step": 43254
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.473443823149128e-06,
+      "loss": 0.4209,
+      "step": 43255
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.473107987842763e-06,
+      "loss": 0.2799,
+      "step": 43256
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4727721525363965e-06,
+      "loss": 0.3134,
+      "step": 43257
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.472436317230031e-06,
+      "loss": 0.2152,
+      "step": 43258
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.472100481923666e-06,
+      "loss": 0.1805,
+      "step": 43259
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.471764646617299e-06,
+      "loss": 0.3992,
+      "step": 43260
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.471428811310933e-06,
+      "loss": 0.1651,
+      "step": 43261
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.471092976004568e-06,
+      "loss": 0.4055,
+      "step": 43262
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.470757140698202e-06,
+      "loss": 0.376,
+      "step": 43263
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4704213053918364e-06,
+      "loss": 0.2094,
+      "step": 43264
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.470085470085471e-06,
+      "loss": 0.1897,
+      "step": 43265
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.469749634779105e-06,
+      "loss": 0.2391,
+      "step": 43266
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4694137994727395e-06,
+      "loss": 0.4396,
+      "step": 43267
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4690779641663725e-06,
+      "loss": 0.2446,
+      "step": 43268
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.468742128860007e-06,
+      "loss": 0.1507,
+      "step": 43269
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.468406293553642e-06,
+      "loss": 0.3023,
+      "step": 43270
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.468070458247276e-06,
+      "loss": 0.2972,
+      "step": 43271
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.46773462294091e-06,
+      "loss": 0.1288,
+      "step": 43272
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.467398787634545e-06,
+      "loss": 0.2704,
+      "step": 43273
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.467062952328178e-06,
+      "loss": 0.3595,
+      "step": 43274
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4667271170218125e-06,
+      "loss": 0.2378,
+      "step": 43275
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.466391281715447e-06,
+      "loss": 0.1203,
+      "step": 43276
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.466055446409081e-06,
+      "loss": 0.2937,
+      "step": 43277
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.465719611102716e-06,
+      "loss": 0.2048,
+      "step": 43278
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.46538377579635e-06,
+      "loss": 0.2951,
+      "step": 43279
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.465047940489984e-06,
+      "loss": 0.1939,
+      "step": 43280
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.464712105183619e-06,
+      "loss": 0.2168,
+      "step": 43281
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4643762698772525e-06,
+      "loss": 0.334,
+      "step": 43282
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.464040434570886e-06,
+      "loss": 0.335,
+      "step": 43283
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.463704599264521e-06,
+      "loss": 0.3421,
+      "step": 43284
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4633687639581556e-06,
+      "loss": 0.2498,
+      "step": 43285
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.463032928651789e-06,
+      "loss": 0.3162,
+      "step": 43286
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.462697093345424e-06,
+      "loss": 0.2589,
+      "step": 43287
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.462361258039059e-06,
+      "loss": 0.3274,
+      "step": 43288
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.462025422732692e-06,
+      "loss": 0.3187,
+      "step": 43289
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.461689587426326e-06,
+      "loss": 0.1491,
+      "step": 43290
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.461353752119961e-06,
+      "loss": 0.3016,
+      "step": 43291
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.461017916813595e-06,
+      "loss": 0.2126,
+      "step": 43292
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.460682081507229e-06,
+      "loss": 0.2301,
+      "step": 43293
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.460346246200864e-06,
+      "loss": 0.1971,
+      "step": 43294
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.460010410894497e-06,
+      "loss": 0.2804,
+      "step": 43295
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.459674575588132e-06,
+      "loss": 0.3621,
+      "step": 43296
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.459338740281766e-06,
+      "loss": 0.3394,
+      "step": 43297
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4590029049754e-06,
+      "loss": 0.2914,
+      "step": 43298
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.458667069669035e-06,
+      "loss": 0.1701,
+      "step": 43299
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.458331234362669e-06,
+      "loss": 0.3767,
+      "step": 43300
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.457995399056303e-06,
+      "loss": 0.1787,
+      "step": 43301
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.457659563749938e-06,
+      "loss": 0.3652,
+      "step": 43302
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4573237284435724e-06,
+      "loss": 0.3111,
+      "step": 43303
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.456987893137205e-06,
+      "loss": 0.1899,
+      "step": 43304
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.45665205783084e-06,
+      "loss": 0.2441,
+      "step": 43305
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.456316222524475e-06,
+      "loss": 0.3732,
+      "step": 43306
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4559803872181085e-06,
+      "loss": 0.2564,
+      "step": 43307
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.455644551911743e-06,
+      "loss": 0.2092,
+      "step": 43308
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.455308716605378e-06,
+      "loss": 0.215,
+      "step": 43309
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.454972881299011e-06,
+      "loss": 0.2796,
+      "step": 43310
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.454637045992645e-06,
+      "loss": 0.3944,
+      "step": 43311
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.45430121068628e-06,
+      "loss": 0.2247,
+      "step": 43312
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.453965375379914e-06,
+      "loss": 0.2796,
+      "step": 43313
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4536295400735485e-06,
+      "loss": 0.1529,
+      "step": 43314
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.453293704767183e-06,
+      "loss": 0.266,
+      "step": 43315
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.452957869460817e-06,
+      "loss": 0.2816,
+      "step": 43316
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.452622034154451e-06,
+      "loss": 0.4117,
+      "step": 43317
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.452286198848085e-06,
+      "loss": 0.1506,
+      "step": 43318
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.451950363541719e-06,
+      "loss": 0.2998,
+      "step": 43319
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.451614528235354e-06,
+      "loss": 0.1627,
+      "step": 43320
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4512786929289885e-06,
+      "loss": 0.1552,
+      "step": 43321
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.450942857622622e-06,
+      "loss": 0.268,
+      "step": 43322
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.450607022316257e-06,
+      "loss": 0.2234,
+      "step": 43323
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4502711870098916e-06,
+      "loss": 0.2335,
+      "step": 43324
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4499353517035245e-06,
+      "loss": 0.2571,
+      "step": 43325
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.449599516397159e-06,
+      "loss": 0.3988,
+      "step": 43326
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.449263681090794e-06,
+      "loss": 0.1994,
+      "step": 43327
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.448927845784428e-06,
+      "loss": 0.1867,
+      "step": 43328
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.448592010478062e-06,
+      "loss": 0.1383,
+      "step": 43329
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.448256175171697e-06,
+      "loss": 0.292,
+      "step": 43330
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.44792033986533e-06,
+      "loss": 0.2019,
+      "step": 43331
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4475845045589645e-06,
+      "loss": 0.2616,
+      "step": 43332
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.447248669252599e-06,
+      "loss": 0.3805,
+      "step": 43333
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.446912833946233e-06,
+      "loss": 0.3,
+      "step": 43334
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.446576998639868e-06,
+      "loss": 0.2803,
+      "step": 43335
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.446241163333502e-06,
+      "loss": 0.2784,
+      "step": 43336
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.445905328027136e-06,
+      "loss": 0.1804,
+      "step": 43337
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.445569492720771e-06,
+      "loss": 0.2353,
+      "step": 43338
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4452336574144045e-06,
+      "loss": 0.2448,
+      "step": 43339
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.444897822108038e-06,
+      "loss": 0.3107,
+      "step": 43340
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.444561986801673e-06,
+      "loss": 0.2526,
+      "step": 43341
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.444226151495308e-06,
+      "loss": 0.2194,
+      "step": 43342
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.443890316188941e-06,
+      "loss": 0.3276,
+      "step": 43343
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.443554480882576e-06,
+      "loss": 0.2395,
+      "step": 43344
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.443218645576211e-06,
+      "loss": 0.1979,
+      "step": 43345
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.442882810269844e-06,
+      "loss": 0.2141,
+      "step": 43346
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.442546974963478e-06,
+      "loss": 0.315,
+      "step": 43347
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.442211139657113e-06,
+      "loss": 0.2695,
+      "step": 43348
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.441875304350747e-06,
+      "loss": 0.2341,
+      "step": 43349
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.441539469044381e-06,
+      "loss": 0.4003,
+      "step": 43350
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.441203633738016e-06,
+      "loss": 0.3167,
+      "step": 43351
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.440867798431649e-06,
+      "loss": 0.3131,
+      "step": 43352
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.440531963125284e-06,
+      "loss": 0.2853,
+      "step": 43353
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.440196127818918e-06,
+      "loss": 0.3847,
+      "step": 43354
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.439860292512552e-06,
+      "loss": 0.2412,
+      "step": 43355
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.439524457206187e-06,
+      "loss": 0.1887,
+      "step": 43356
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4391886218998205e-06,
+      "loss": 0.2127,
+      "step": 43357
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.438852786593455e-06,
+      "loss": 0.2004,
+      "step": 43358
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.43851695128709e-06,
+      "loss": 0.4764,
+      "step": 43359
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.438181115980723e-06,
+      "loss": 0.1526,
+      "step": 43360
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4378452806743574e-06,
+      "loss": 0.3817,
+      "step": 43361
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.437509445367992e-06,
+      "loss": 0.2073,
+      "step": 43362
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.437173610061626e-06,
+      "loss": 0.2612,
+      "step": 43363
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4368377747552605e-06,
+      "loss": 0.2089,
+      "step": 43364
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.436501939448895e-06,
+      "loss": 0.1722,
+      "step": 43365
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.436166104142528e-06,
+      "loss": 0.2937,
+      "step": 43366
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.435830268836163e-06,
+      "loss": 0.1689,
+      "step": 43367
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.435494433529797e-06,
+      "loss": 0.2352,
+      "step": 43368
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.435158598223431e-06,
+      "loss": 0.3978,
+      "step": 43369
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.434822762917066e-06,
+      "loss": 0.2567,
+      "step": 43370
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.4344869276107005e-06,
+      "loss": 0.26,
+      "step": 43371
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.434151092304334e-06,
+      "loss": 0.2088,
+      "step": 43372
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.433815256997968e-06,
+      "loss": 0.2717,
+      "step": 43373
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 5.433479421691603e-06,
+      "loss": 0.4902,
+      "step": 43374
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.4331435863852366e-06,
+      "loss": 0.1855,
+      "step": 43375
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.432807751078871e-06,
+      "loss": 0.1746,
+      "step": 43376
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.432471915772506e-06,
+      "loss": 0.5044,
+      "step": 43377
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.43213608046614e-06,
+      "loss": 0.3002,
+      "step": 43378
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.431800245159774e-06,
+      "loss": 0.1766,
+      "step": 43379
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.431464409853409e-06,
+      "loss": 0.164,
+      "step": 43380
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.431128574547042e-06,
+      "loss": 0.2439,
+      "step": 43381
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.4307927392406766e-06,
+      "loss": 0.2292,
+      "step": 43382
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.430456903934311e-06,
+      "loss": 0.1405,
+      "step": 43383
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.430121068627945e-06,
+      "loss": 0.1909,
+      "step": 43384
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.42978523332158e-06,
+      "loss": 0.3223,
+      "step": 43385
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.429449398015214e-06,
+      "loss": 0.2882,
+      "step": 43386
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.429113562708847e-06,
+      "loss": 0.2724,
+      "step": 43387
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.428777727402482e-06,
+      "loss": 0.366,
+      "step": 43388
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.4284418920961165e-06,
+      "loss": 0.2187,
+      "step": 43389
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.42810605678975e-06,
+      "loss": 0.2144,
+      "step": 43390
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.427770221483385e-06,
+      "loss": 0.2728,
+      "step": 43391
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.42743438617702e-06,
+      "loss": 0.3052,
+      "step": 43392
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.4270985508706534e-06,
+      "loss": 0.1676,
+      "step": 43393
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.426762715564288e-06,
+      "loss": 0.1856,
+      "step": 43394
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.426426880257922e-06,
+      "loss": 0.1603,
+      "step": 43395
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.426091044951556e-06,
+      "loss": 0.4262,
+      "step": 43396
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.42575520964519e-06,
+      "loss": 0.2502,
+      "step": 43397
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.425419374338825e-06,
+      "loss": 0.1884,
+      "step": 43398
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.425083539032459e-06,
+      "loss": 0.2551,
+      "step": 43399
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.4247477037260934e-06,
+      "loss": 0.1623,
+      "step": 43400
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.424411868419728e-06,
+      "loss": 0.3772,
+      "step": 43401
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.424076033113361e-06,
+      "loss": 0.27,
+      "step": 43402
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.423740197806996e-06,
+      "loss": 0.2731,
+      "step": 43403
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.42340436250063e-06,
+      "loss": 0.1755,
+      "step": 43404
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.423068527194264e-06,
+      "loss": 0.2104,
+      "step": 43405
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.422732691887899e-06,
+      "loss": 0.3181,
+      "step": 43406
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.422396856581533e-06,
+      "loss": 0.2829,
+      "step": 43407
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.422061021275166e-06,
+      "loss": 0.2488,
+      "step": 43408
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.421725185968801e-06,
+      "loss": 0.1637,
+      "step": 43409
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.421389350662436e-06,
+      "loss": 0.2691,
+      "step": 43410
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.4210535153560695e-06,
+      "loss": 0.2613,
+      "step": 43411
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.420717680049704e-06,
+      "loss": 0.1316,
+      "step": 43412
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.420381844743339e-06,
+      "loss": 0.2792,
+      "step": 43413
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.4200460094369726e-06,
+      "loss": 0.1918,
+      "step": 43414
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.419710174130607e-06,
+      "loss": 0.5075,
+      "step": 43415
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.419374338824242e-06,
+      "loss": 0.1251,
+      "step": 43416
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.419038503517875e-06,
+      "loss": 0.2558,
+      "step": 43417
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.4187026682115095e-06,
+      "loss": 0.209,
+      "step": 43418
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.418366832905144e-06,
+      "loss": 0.3059,
+      "step": 43419
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.418030997598778e-06,
+      "loss": 0.2276,
+      "step": 43420
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.4176951622924126e-06,
+      "loss": 0.3767,
+      "step": 43421
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.417359326986047e-06,
+      "loss": 0.3838,
+      "step": 43422
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.41702349167968e-06,
+      "loss": 0.1809,
+      "step": 43423
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.416687656373315e-06,
+      "loss": 0.2596,
+      "step": 43424
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.4163518210669495e-06,
+      "loss": 0.2305,
+      "step": 43425
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.416015985760583e-06,
+      "loss": 0.2388,
+      "step": 43426
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.415680150454218e-06,
+      "loss": 0.2594,
+      "step": 43427
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.4153443151478525e-06,
+      "loss": 0.256,
+      "step": 43428
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.4150084798414855e-06,
+      "loss": 0.1999,
+      "step": 43429
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.41467264453512e-06,
+      "loss": 0.2571,
+      "step": 43430
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.414336809228755e-06,
+      "loss": 0.1553,
+      "step": 43431
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.414000973922389e-06,
+      "loss": 0.2415,
+      "step": 43432
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.413665138616023e-06,
+      "loss": 0.2719,
+      "step": 43433
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.413329303309658e-06,
+      "loss": 0.3111,
+      "step": 43434
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.412993468003292e-06,
+      "loss": 0.4108,
+      "step": 43435
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.412657632696926e-06,
+      "loss": 0.2932,
+      "step": 43436
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.412321797390561e-06,
+      "loss": 0.2363,
+      "step": 43437
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.411985962084194e-06,
+      "loss": 0.1985,
+      "step": 43438
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.411650126777829e-06,
+      "loss": 0.2667,
+      "step": 43439
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.411314291471463e-06,
+      "loss": 0.2031,
+      "step": 43440
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.410978456165097e-06,
+      "loss": 0.3283,
+      "step": 43441
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.410642620858732e-06,
+      "loss": 0.3476,
+      "step": 43442
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.410306785552366e-06,
+      "loss": 0.3644,
+      "step": 43443
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.409970950245999e-06,
+      "loss": 0.3169,
+      "step": 43444
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.409635114939634e-06,
+      "loss": 0.2061,
+      "step": 43445
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.4092992796332686e-06,
+      "loss": 0.1841,
+      "step": 43446
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.408963444326902e-06,
+      "loss": 0.2187,
+      "step": 43447
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.408627609020537e-06,
+      "loss": 0.1708,
+      "step": 43448
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.408291773714171e-06,
+      "loss": 0.224,
+      "step": 43449
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.4079559384078055e-06,
+      "loss": 0.162,
+      "step": 43450
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.40762010310144e-06,
+      "loss": 0.2226,
+      "step": 43451
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.407284267795073e-06,
+      "loss": 0.2719,
+      "step": 43452
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.406948432488708e-06,
+      "loss": 0.1289,
+      "step": 43453
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.406612597182342e-06,
+      "loss": 0.2145,
+      "step": 43454
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.406276761875976e-06,
+      "loss": 0.4615,
+      "step": 43455
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.405940926569611e-06,
+      "loss": 0.2931,
+      "step": 43456
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.4056050912632455e-06,
+      "loss": 0.3567,
+      "step": 43457
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.405269255956878e-06,
+      "loss": 0.2559,
+      "step": 43458
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.404933420650513e-06,
+      "loss": 0.2188,
+      "step": 43459
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.404597585344148e-06,
+      "loss": 0.272,
+      "step": 43460
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.4042617500377815e-06,
+      "loss": 0.3348,
+      "step": 43461
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.403925914731416e-06,
+      "loss": 0.2309,
+      "step": 43462
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.403590079425051e-06,
+      "loss": 0.2843,
+      "step": 43463
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.403254244118684e-06,
+      "loss": 0.2317,
+      "step": 43464
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.402918408812318e-06,
+      "loss": 0.1504,
+      "step": 43465
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.402582573505953e-06,
+      "loss": 0.2982,
+      "step": 43466
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.402246738199587e-06,
+      "loss": 0.3333,
+      "step": 43467
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.4019109028932215e-06,
+      "loss": 0.2152,
+      "step": 43468
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.401575067586856e-06,
+      "loss": 0.3119,
+      "step": 43469
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.40123923228049e-06,
+      "loss": 0.1107,
+      "step": 43470
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.400903396974125e-06,
+      "loss": 0.3096,
+      "step": 43471
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.400567561667759e-06,
+      "loss": 0.4183,
+      "step": 43472
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.400231726361392e-06,
+      "loss": 0.2843,
+      "step": 43473
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.399895891055027e-06,
+      "loss": 0.2238,
+      "step": 43474
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.3995600557486615e-06,
+      "loss": 0.3544,
+      "step": 43475
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.399224220442295e-06,
+      "loss": 0.1822,
+      "step": 43476
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.39888838513593e-06,
+      "loss": 0.2723,
+      "step": 43477
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.398552549829565e-06,
+      "loss": 0.2784,
+      "step": 43478
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.3982167145231975e-06,
+      "loss": 0.3336,
+      "step": 43479
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.397880879216832e-06,
+      "loss": 0.2033,
+      "step": 43480
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.397545043910467e-06,
+      "loss": 0.2034,
+      "step": 43481
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.397209208604101e-06,
+      "loss": 0.247,
+      "step": 43482
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.396873373297735e-06,
+      "loss": 0.2522,
+      "step": 43483
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.39653753799137e-06,
+      "loss": 0.2881,
+      "step": 43484
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.396201702685004e-06,
+      "loss": 0.2933,
+      "step": 43485
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.3958658673786375e-06,
+      "loss": 0.166,
+      "step": 43486
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.395530032072272e-06,
+      "loss": 0.1601,
+      "step": 43487
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.395194196765906e-06,
+      "loss": 0.3319,
+      "step": 43488
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.394858361459541e-06,
+      "loss": 0.2011,
+      "step": 43489
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.394522526153175e-06,
+      "loss": 0.2903,
+      "step": 43490
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.394186690846809e-06,
+      "loss": 0.2963,
+      "step": 43491
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.393850855540444e-06,
+      "loss": 0.3101,
+      "step": 43492
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.393515020234078e-06,
+      "loss": 0.3133,
+      "step": 43493
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.393179184927711e-06,
+      "loss": 0.2403,
+      "step": 43494
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.392843349621346e-06,
+      "loss": 0.1755,
+      "step": 43495
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.392507514314981e-06,
+      "loss": 0.4338,
+      "step": 43496
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.392171679008614e-06,
+      "loss": 0.2455,
+      "step": 43497
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.391835843702249e-06,
+      "loss": 0.4666,
+      "step": 43498
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.391500008395884e-06,
+      "loss": 0.1997,
+      "step": 43499
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.391164173089517e-06,
+      "loss": 0.1521,
+      "step": 43500
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.390828337783151e-06,
+      "loss": 0.3491,
+      "step": 43501
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.390492502476786e-06,
+      "loss": 0.2357,
+      "step": 43502
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.39015666717042e-06,
+      "loss": 0.3436,
+      "step": 43503
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.389820831864054e-06,
+      "loss": 0.3156,
+      "step": 43504
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.389484996557689e-06,
+      "loss": 0.3295,
+      "step": 43505
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.389149161251323e-06,
+      "loss": 0.242,
+      "step": 43506
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.3888133259449575e-06,
+      "loss": 0.2636,
+      "step": 43507
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.388477490638591e-06,
+      "loss": 0.5058,
+      "step": 43508
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.388141655332225e-06,
+      "loss": 0.3264,
+      "step": 43509
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.38780582002586e-06,
+      "loss": 0.2735,
+      "step": 43510
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.387469984719494e-06,
+      "loss": 0.3472,
+      "step": 43511
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.387134149413128e-06,
+      "loss": 0.3092,
+      "step": 43512
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.386798314106763e-06,
+      "loss": 0.1865,
+      "step": 43513
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.3864624788003975e-06,
+      "loss": 0.1719,
+      "step": 43514
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.3861266434940304e-06,
+      "loss": 0.3625,
+      "step": 43515
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.385790808187665e-06,
+      "loss": 0.2967,
+      "step": 43516
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.3854549728813e-06,
+      "loss": 0.1352,
+      "step": 43517
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.3851191375749335e-06,
+      "loss": 0.2738,
+      "step": 43518
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.384783302268568e-06,
+      "loss": 0.2666,
+      "step": 43519
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.384447466962203e-06,
+      "loss": 0.284,
+      "step": 43520
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.384111631655836e-06,
+      "loss": 0.288,
+      "step": 43521
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.3837757963494704e-06,
+      "loss": 0.3132,
+      "step": 43522
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.383439961043105e-06,
+      "loss": 0.2186,
+      "step": 43523
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.383104125736739e-06,
+      "loss": 0.335,
+      "step": 43524
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.3827682904303735e-06,
+      "loss": 0.3295,
+      "step": 43525
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.382432455124008e-06,
+      "loss": 0.2744,
+      "step": 43526
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.382096619817642e-06,
+      "loss": 0.3628,
+      "step": 43527
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.381760784511277e-06,
+      "loss": 0.2162,
+      "step": 43528
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.381424949204911e-06,
+      "loss": 0.2907,
+      "step": 43529
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.381089113898544e-06,
+      "loss": 0.3106,
+      "step": 43530
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.380753278592179e-06,
+      "loss": 0.2802,
+      "step": 43531
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.3804174432858135e-06,
+      "loss": 0.3258,
+      "step": 43532
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.380081607979447e-06,
+      "loss": 0.273,
+      "step": 43533
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.379745772673082e-06,
+      "loss": 0.2556,
+      "step": 43534
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.379409937366717e-06,
+      "loss": 0.3376,
+      "step": 43535
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.3790741020603496e-06,
+      "loss": 0.2821,
+      "step": 43536
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.378738266753984e-06,
+      "loss": 0.2258,
+      "step": 43537
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.378402431447618e-06,
+      "loss": 0.3927,
+      "step": 43538
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.378066596141253e-06,
+      "loss": 0.375,
+      "step": 43539
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.377730760834887e-06,
+      "loss": 0.1995,
+      "step": 43540
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.377394925528521e-06,
+      "loss": 0.3005,
+      "step": 43541
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.377059090222155e-06,
+      "loss": 0.3581,
+      "step": 43542
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.3767232549157896e-06,
+      "loss": 0.3551,
+      "step": 43543
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.376387419609423e-06,
+      "loss": 0.1714,
+      "step": 43544
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.376051584303058e-06,
+      "loss": 0.2021,
+      "step": 43545
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.375715748996693e-06,
+      "loss": 0.3493,
+      "step": 43546
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.3753799136903265e-06,
+      "loss": 0.2576,
+      "step": 43547
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.375044078383961e-06,
+      "loss": 0.2268,
+      "step": 43548
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.374708243077596e-06,
+      "loss": 0.2361,
+      "step": 43549
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.374372407771229e-06,
+      "loss": 0.2896,
+      "step": 43550
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.374036572464863e-06,
+      "loss": 0.1839,
+      "step": 43551
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.373700737158498e-06,
+      "loss": 0.3984,
+      "step": 43552
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.373364901852132e-06,
+      "loss": 0.3664,
+      "step": 43553
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.3730290665457664e-06,
+      "loss": 0.1522,
+      "step": 43554
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.372693231239401e-06,
+      "loss": 0.1895,
+      "step": 43555
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.372357395933034e-06,
+      "loss": 0.2004,
+      "step": 43556
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.372021560626669e-06,
+      "loss": 0.2845,
+      "step": 43557
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.371685725320303e-06,
+      "loss": 0.1617,
+      "step": 43558
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.371349890013937e-06,
+      "loss": 0.2192,
+      "step": 43559
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.371014054707572e-06,
+      "loss": 0.2002,
+      "step": 43560
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.3706782194012064e-06,
+      "loss": 0.3625,
+      "step": 43561
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.37034238409484e-06,
+      "loss": 0.1987,
+      "step": 43562
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.370006548788475e-06,
+      "loss": 0.3855,
+      "step": 43563
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.3696707134821095e-06,
+      "loss": 0.4731,
+      "step": 43564
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.3693348781757425e-06,
+      "loss": 0.3332,
+      "step": 43565
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.368999042869377e-06,
+      "loss": 0.1704,
+      "step": 43566
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.368663207563012e-06,
+      "loss": 0.2244,
+      "step": 43567
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.368327372256646e-06,
+      "loss": 0.184,
+      "step": 43568
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.36799153695028e-06,
+      "loss": 0.1854,
+      "step": 43569
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.367655701643915e-06,
+      "loss": 0.3571,
+      "step": 43570
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.367319866337548e-06,
+      "loss": 0.17,
+      "step": 43571
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 5.3669840310311825e-06,
+      "loss": 0.3116,
+      "step": 43572
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.366648195724817e-06,
+      "loss": 0.1276,
+      "step": 43573
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.366312360418451e-06,
+      "loss": 0.3439,
+      "step": 43574
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3659765251120856e-06,
+      "loss": 0.2508,
+      "step": 43575
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.36564068980572e-06,
+      "loss": 0.4239,
+      "step": 43576
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.365304854499353e-06,
+      "loss": 0.2646,
+      "step": 43577
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.364969019192988e-06,
+      "loss": 0.333,
+      "step": 43578
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3646331838866225e-06,
+      "loss": 0.3428,
+      "step": 43579
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.364297348580256e-06,
+      "loss": 0.2294,
+      "step": 43580
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.363961513273891e-06,
+      "loss": 0.2258,
+      "step": 43581
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3636256779675256e-06,
+      "loss": 0.3366,
+      "step": 43582
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.363289842661159e-06,
+      "loss": 0.2845,
+      "step": 43583
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.362954007354794e-06,
+      "loss": 0.1619,
+      "step": 43584
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.362618172048429e-06,
+      "loss": 0.1352,
+      "step": 43585
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.362282336742062e-06,
+      "loss": 0.1744,
+      "step": 43586
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.361946501435696e-06,
+      "loss": 0.1622,
+      "step": 43587
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.361610666129331e-06,
+      "loss": 0.2675,
+      "step": 43588
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.361274830822965e-06,
+      "loss": 0.376,
+      "step": 43589
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.360938995516599e-06,
+      "loss": 0.3711,
+      "step": 43590
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.360603160210234e-06,
+      "loss": 0.2643,
+      "step": 43591
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.360267324903867e-06,
+      "loss": 0.1963,
+      "step": 43592
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.359931489597502e-06,
+      "loss": 0.2367,
+      "step": 43593
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.359595654291136e-06,
+      "loss": 0.2246,
+      "step": 43594
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.35925981898477e-06,
+      "loss": 0.1147,
+      "step": 43595
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.358923983678405e-06,
+      "loss": 0.2278,
+      "step": 43596
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.358588148372039e-06,
+      "loss": 0.429,
+      "step": 43597
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.358252313065673e-06,
+      "loss": 0.1264,
+      "step": 43598
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.357916477759307e-06,
+      "loss": 0.1625,
+      "step": 43599
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.357580642452942e-06,
+      "loss": 0.1994,
+      "step": 43600
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.357244807146575e-06,
+      "loss": 0.2307,
+      "step": 43601
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.35690897184021e-06,
+      "loss": 0.2248,
+      "step": 43602
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.356573136533845e-06,
+      "loss": 0.1524,
+      "step": 43603
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3562373012274785e-06,
+      "loss": 0.3558,
+      "step": 43604
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.355901465921113e-06,
+      "loss": 0.2146,
+      "step": 43605
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.355565630614748e-06,
+      "loss": 0.1117,
+      "step": 43606
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.355229795308381e-06,
+      "loss": 0.1595,
+      "step": 43607
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.354893960002015e-06,
+      "loss": 0.3439,
+      "step": 43608
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.35455812469565e-06,
+      "loss": 0.2643,
+      "step": 43609
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.354222289389284e-06,
+      "loss": 0.5353,
+      "step": 43610
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3538864540829185e-06,
+      "loss": 0.373,
+      "step": 43611
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.353550618776553e-06,
+      "loss": 0.3178,
+      "step": 43612
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.353214783470186e-06,
+      "loss": 0.2365,
+      "step": 43613
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.352878948163821e-06,
+      "loss": 0.2787,
+      "step": 43614
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.352543112857455e-06,
+      "loss": 0.1159,
+      "step": 43615
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.352207277551089e-06,
+      "loss": 0.3399,
+      "step": 43616
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.351871442244724e-06,
+      "loss": 0.15,
+      "step": 43617
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3515356069383585e-06,
+      "loss": 0.2923,
+      "step": 43618
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.351199771631992e-06,
+      "loss": 0.3838,
+      "step": 43619
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.350863936325627e-06,
+      "loss": 0.457,
+      "step": 43620
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.350528101019261e-06,
+      "loss": 0.3607,
+      "step": 43621
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3501922657128945e-06,
+      "loss": 0.1725,
+      "step": 43622
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.349856430406529e-06,
+      "loss": 0.4059,
+      "step": 43623
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.349520595100164e-06,
+      "loss": 0.1813,
+      "step": 43624
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.349184759793798e-06,
+      "loss": 0.24,
+      "step": 43625
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.348848924487432e-06,
+      "loss": 0.2072,
+      "step": 43626
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.348513089181067e-06,
+      "loss": 0.3382,
+      "step": 43627
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3481772538747e-06,
+      "loss": 0.1828,
+      "step": 43628
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3478414185683345e-06,
+      "loss": 0.1476,
+      "step": 43629
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.347505583261968e-06,
+      "loss": 0.2502,
+      "step": 43630
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.347169747955603e-06,
+      "loss": 0.3536,
+      "step": 43631
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.346833912649238e-06,
+      "loss": 0.2708,
+      "step": 43632
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3464980773428706e-06,
+      "loss": 0.1978,
+      "step": 43633
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.346162242036505e-06,
+      "loss": 0.2201,
+      "step": 43634
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.34582640673014e-06,
+      "loss": 0.1746,
+      "step": 43635
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.345490571423774e-06,
+      "loss": 0.2917,
+      "step": 43636
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.345154736117408e-06,
+      "loss": 0.3259,
+      "step": 43637
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.344818900811043e-06,
+      "loss": 0.3749,
+      "step": 43638
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.344483065504677e-06,
+      "loss": 0.2793,
+      "step": 43639
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.344147230198311e-06,
+      "loss": 0.273,
+      "step": 43640
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.343811394891946e-06,
+      "loss": 0.4362,
+      "step": 43641
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.343475559585579e-06,
+      "loss": 0.3073,
+      "step": 43642
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.343139724279214e-06,
+      "loss": 0.1664,
+      "step": 43643
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.342803888972848e-06,
+      "loss": 0.2873,
+      "step": 43644
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.342468053666482e-06,
+      "loss": 0.2095,
+      "step": 43645
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.342132218360117e-06,
+      "loss": 0.2492,
+      "step": 43646
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.341796383053751e-06,
+      "loss": 0.4585,
+      "step": 43647
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.341460547747384e-06,
+      "loss": 0.3142,
+      "step": 43648
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.341124712441019e-06,
+      "loss": 0.2372,
+      "step": 43649
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.340788877134654e-06,
+      "loss": 0.2161,
+      "step": 43650
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3404530418282874e-06,
+      "loss": 0.2329,
+      "step": 43651
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.340117206521922e-06,
+      "loss": 0.2117,
+      "step": 43652
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.339781371215557e-06,
+      "loss": 0.1618,
+      "step": 43653
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3394455359091905e-06,
+      "loss": 0.1671,
+      "step": 43654
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.339109700602824e-06,
+      "loss": 0.3239,
+      "step": 43655
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.338773865296459e-06,
+      "loss": 0.1547,
+      "step": 43656
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.338438029990093e-06,
+      "loss": 0.3616,
+      "step": 43657
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3381021946837274e-06,
+      "loss": 0.284,
+      "step": 43658
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.337766359377362e-06,
+      "loss": 0.1298,
+      "step": 43659
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.337430524070996e-06,
+      "loss": 0.2619,
+      "step": 43660
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3370946887646305e-06,
+      "loss": 0.2278,
+      "step": 43661
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.336758853458265e-06,
+      "loss": 0.4275,
+      "step": 43662
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.336423018151898e-06,
+      "loss": 0.3421,
+      "step": 43663
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.336087182845533e-06,
+      "loss": 0.1967,
+      "step": 43664
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.335751347539167e-06,
+      "loss": 0.2472,
+      "step": 43665
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.335415512232801e-06,
+      "loss": 0.317,
+      "step": 43666
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.335079676926436e-06,
+      "loss": 0.2948,
+      "step": 43667
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3347438416200705e-06,
+      "loss": 0.2561,
+      "step": 43668
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3344080063137035e-06,
+      "loss": 0.2331,
+      "step": 43669
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.334072171007338e-06,
+      "loss": 0.2388,
+      "step": 43670
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.333736335700973e-06,
+      "loss": 0.3397,
+      "step": 43671
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3334005003946066e-06,
+      "loss": 0.322,
+      "step": 43672
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.333064665088241e-06,
+      "loss": 0.2426,
+      "step": 43673
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.332728829781876e-06,
+      "loss": 0.2533,
+      "step": 43674
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.33239299447551e-06,
+      "loss": 0.2238,
+      "step": 43675
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.332057159169144e-06,
+      "loss": 0.2461,
+      "step": 43676
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.331721323862778e-06,
+      "loss": 0.2107,
+      "step": 43677
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.331385488556412e-06,
+      "loss": 0.2056,
+      "step": 43678
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3310496532500465e-06,
+      "loss": 0.3034,
+      "step": 43679
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.330713817943681e-06,
+      "loss": 0.1725,
+      "step": 43680
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.330377982637315e-06,
+      "loss": 0.328,
+      "step": 43681
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.33004214733095e-06,
+      "loss": 0.2833,
+      "step": 43682
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.329706312024584e-06,
+      "loss": 0.3614,
+      "step": 43683
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.329370476718217e-06,
+      "loss": 0.4114,
+      "step": 43684
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.329034641411852e-06,
+      "loss": 0.3591,
+      "step": 43685
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3286988061054865e-06,
+      "loss": 0.2251,
+      "step": 43686
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.32836297079912e-06,
+      "loss": 0.2558,
+      "step": 43687
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.328027135492755e-06,
+      "loss": 0.1631,
+      "step": 43688
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.32769130018639e-06,
+      "loss": 0.3318,
+      "step": 43689
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.327355464880023e-06,
+      "loss": 0.1823,
+      "step": 43690
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.327019629573657e-06,
+      "loss": 0.2482,
+      "step": 43691
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.326683794267292e-06,
+      "loss": 0.2069,
+      "step": 43692
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.326347958960926e-06,
+      "loss": 0.1932,
+      "step": 43693
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.32601212365456e-06,
+      "loss": 0.2171,
+      "step": 43694
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.325676288348195e-06,
+      "loss": 0.3451,
+      "step": 43695
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.325340453041829e-06,
+      "loss": 0.2366,
+      "step": 43696
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3250046177354634e-06,
+      "loss": 0.3212,
+      "step": 43697
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.324668782429098e-06,
+      "loss": 0.2735,
+      "step": 43698
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.324332947122731e-06,
+      "loss": 0.1951,
+      "step": 43699
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.323997111816366e-06,
+      "loss": 0.2164,
+      "step": 43700
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.32366127651e-06,
+      "loss": 0.1888,
+      "step": 43701
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.323325441203634e-06,
+      "loss": 0.2303,
+      "step": 43702
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.322989605897269e-06,
+      "loss": 0.389,
+      "step": 43703
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.322653770590903e-06,
+      "loss": 0.1343,
+      "step": 43704
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.322317935284536e-06,
+      "loss": 0.2606,
+      "step": 43705
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.321982099978171e-06,
+      "loss": 0.2052,
+      "step": 43706
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.321646264671806e-06,
+      "loss": 0.3451,
+      "step": 43707
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3213104293654395e-06,
+      "loss": 0.1381,
+      "step": 43708
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.320974594059074e-06,
+      "loss": 0.2832,
+      "step": 43709
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.320638758752709e-06,
+      "loss": 0.2323,
+      "step": 43710
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3203029234463426e-06,
+      "loss": 0.2729,
+      "step": 43711
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.319967088139976e-06,
+      "loss": 0.1779,
+      "step": 43712
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.319631252833611e-06,
+      "loss": 0.2217,
+      "step": 43713
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.319295417527245e-06,
+      "loss": 0.33,
+      "step": 43714
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3189595822208795e-06,
+      "loss": 0.2967,
+      "step": 43715
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.318623746914514e-06,
+      "loss": 0.1611,
+      "step": 43716
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.318287911608148e-06,
+      "loss": 0.2691,
+      "step": 43717
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3179520763017825e-06,
+      "loss": 0.2125,
+      "step": 43718
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3176162409954155e-06,
+      "loss": 0.3376,
+      "step": 43719
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.31728040568905e-06,
+      "loss": 0.1974,
+      "step": 43720
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.316944570382685e-06,
+      "loss": 0.0897,
+      "step": 43721
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.316608735076319e-06,
+      "loss": 0.3438,
+      "step": 43722
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.316272899769953e-06,
+      "loss": 0.3539,
+      "step": 43723
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.315937064463588e-06,
+      "loss": 0.3378,
+      "step": 43724
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.315601229157221e-06,
+      "loss": 0.1667,
+      "step": 43725
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3152653938508555e-06,
+      "loss": 0.3449,
+      "step": 43726
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.31492955854449e-06,
+      "loss": 0.3149,
+      "step": 43727
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.314593723238124e-06,
+      "loss": 0.1711,
+      "step": 43728
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.314257887931759e-06,
+      "loss": 0.3051,
+      "step": 43729
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.313922052625393e-06,
+      "loss": 0.173,
+      "step": 43730
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.313586217319027e-06,
+      "loss": 0.1593,
+      "step": 43731
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.313250382012662e-06,
+      "loss": 0.2428,
+      "step": 43732
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.312914546706296e-06,
+      "loss": 0.2323,
+      "step": 43733
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.312578711399929e-06,
+      "loss": 0.2907,
+      "step": 43734
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.312242876093564e-06,
+      "loss": 0.1895,
+      "step": 43735
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.311907040787199e-06,
+      "loss": 0.2358,
+      "step": 43736
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.311571205480832e-06,
+      "loss": 0.201,
+      "step": 43737
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.311235370174467e-06,
+      "loss": 0.3131,
+      "step": 43738
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.310899534868102e-06,
+      "loss": 0.2566,
+      "step": 43739
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.310563699561735e-06,
+      "loss": 0.1695,
+      "step": 43740
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.310227864255369e-06,
+      "loss": 0.2804,
+      "step": 43741
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.309892028949004e-06,
+      "loss": 0.3504,
+      "step": 43742
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.309556193642638e-06,
+      "loss": 0.3065,
+      "step": 43743
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.309220358336272e-06,
+      "loss": 0.4312,
+      "step": 43744
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.308884523029907e-06,
+      "loss": 0.2881,
+      "step": 43745
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.30854868772354e-06,
+      "loss": 0.3011,
+      "step": 43746
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.308212852417175e-06,
+      "loss": 0.1952,
+      "step": 43747
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.307877017110809e-06,
+      "loss": 0.2408,
+      "step": 43748
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.307541181804443e-06,
+      "loss": 0.164,
+      "step": 43749
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.307205346498078e-06,
+      "loss": 0.3597,
+      "step": 43750
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.306869511191712e-06,
+      "loss": 0.3106,
+      "step": 43751
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.306533675885346e-06,
+      "loss": 0.2334,
+      "step": 43752
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.306197840578981e-06,
+      "loss": 0.263,
+      "step": 43753
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3058620052726155e-06,
+      "loss": 0.3766,
+      "step": 43754
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.305526169966248e-06,
+      "loss": 0.1727,
+      "step": 43755
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.305190334659883e-06,
+      "loss": 0.2852,
+      "step": 43756
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.304854499353518e-06,
+      "loss": 0.2689,
+      "step": 43757
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3045186640471515e-06,
+      "loss": 0.2172,
+      "step": 43758
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.304182828740786e-06,
+      "loss": 0.3016,
+      "step": 43759
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.303846993434421e-06,
+      "loss": 0.4937,
+      "step": 43760
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.303511158128054e-06,
+      "loss": 0.2067,
+      "step": 43761
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.303175322821688e-06,
+      "loss": 0.1303,
+      "step": 43762
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.302839487515323e-06,
+      "loss": 0.3365,
+      "step": 43763
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.302503652208957e-06,
+      "loss": 0.3264,
+      "step": 43764
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.3021678169025915e-06,
+      "loss": 0.3348,
+      "step": 43765
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.301831981596226e-06,
+      "loss": 0.4039,
+      "step": 43766
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.30149614628986e-06,
+      "loss": 0.2756,
+      "step": 43767
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.301160310983494e-06,
+      "loss": 0.3595,
+      "step": 43768
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.300824475677128e-06,
+      "loss": 0.2787,
+      "step": 43769
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.300488640370762e-06,
+      "loss": 0.1805,
+      "step": 43770
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 5.300152805064397e-06,
+      "loss": 0.3869,
+      "step": 43771
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2998169697580315e-06,
+      "loss": 0.1762,
+      "step": 43772
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.299481134451665e-06,
+      "loss": 0.1336,
+      "step": 43773
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2991452991453e-06,
+      "loss": 0.1657,
+      "step": 43774
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.298809463838935e-06,
+      "loss": 0.2905,
+      "step": 43775
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2984736285325675e-06,
+      "loss": 0.312,
+      "step": 43776
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.298137793226202e-06,
+      "loss": 0.2643,
+      "step": 43777
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.297801957919837e-06,
+      "loss": 0.2983,
+      "step": 43778
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.297466122613471e-06,
+      "loss": 0.1522,
+      "step": 43779
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.297130287307105e-06,
+      "loss": 0.2448,
+      "step": 43780
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.29679445200074e-06,
+      "loss": 0.2954,
+      "step": 43781
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.296458616694373e-06,
+      "loss": 0.3356,
+      "step": 43782
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2961227813880075e-06,
+      "loss": 0.1745,
+      "step": 43783
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.295786946081642e-06,
+      "loss": 0.2589,
+      "step": 43784
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.295451110775276e-06,
+      "loss": 0.1979,
+      "step": 43785
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.295115275468911e-06,
+      "loss": 0.3364,
+      "step": 43786
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.294779440162545e-06,
+      "loss": 0.2061,
+      "step": 43787
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.294443604856179e-06,
+      "loss": 0.2828,
+      "step": 43788
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.294107769549814e-06,
+      "loss": 0.3309,
+      "step": 43789
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2937719342434475e-06,
+      "loss": 0.2588,
+      "step": 43790
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.293436098937081e-06,
+      "loss": 0.3729,
+      "step": 43791
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.293100263630716e-06,
+      "loss": 0.3396,
+      "step": 43792
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.292764428324351e-06,
+      "loss": 0.2953,
+      "step": 43793
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.292428593017984e-06,
+      "loss": 0.3052,
+      "step": 43794
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.292092757711619e-06,
+      "loss": 0.3754,
+      "step": 43795
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.291756922405254e-06,
+      "loss": 0.2673,
+      "step": 43796
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.291421087098887e-06,
+      "loss": 0.2201,
+      "step": 43797
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.291085251792521e-06,
+      "loss": 0.4723,
+      "step": 43798
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.290749416486156e-06,
+      "loss": 0.2841,
+      "step": 43799
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.29041358117979e-06,
+      "loss": 0.1792,
+      "step": 43800
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.290077745873424e-06,
+      "loss": 0.3473,
+      "step": 43801
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.289741910567059e-06,
+      "loss": 0.2404,
+      "step": 43802
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.289406075260692e-06,
+      "loss": 0.2179,
+      "step": 43803
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.289070239954327e-06,
+      "loss": 0.5071,
+      "step": 43804
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.288734404647961e-06,
+      "loss": 0.1844,
+      "step": 43805
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.288398569341595e-06,
+      "loss": 0.2634,
+      "step": 43806
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.28806273403523e-06,
+      "loss": 0.2193,
+      "step": 43807
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2877268987288635e-06,
+      "loss": 0.2618,
+      "step": 43808
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.287391063422498e-06,
+      "loss": 0.2181,
+      "step": 43809
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.287055228116133e-06,
+      "loss": 0.3833,
+      "step": 43810
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.286719392809766e-06,
+      "loss": 0.1931,
+      "step": 43811
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2863835575034004e-06,
+      "loss": 0.1461,
+      "step": 43812
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.286047722197035e-06,
+      "loss": 0.1167,
+      "step": 43813
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.285711886890669e-06,
+      "loss": 0.3043,
+      "step": 43814
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2853760515843035e-06,
+      "loss": 0.3439,
+      "step": 43815
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.285040216277938e-06,
+      "loss": 0.2759,
+      "step": 43816
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.284704380971571e-06,
+      "loss": 0.2518,
+      "step": 43817
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.284368545665206e-06,
+      "loss": 0.2496,
+      "step": 43818
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2840327103588404e-06,
+      "loss": 0.1525,
+      "step": 43819
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.283696875052474e-06,
+      "loss": 0.1876,
+      "step": 43820
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.283361039746109e-06,
+      "loss": 0.1824,
+      "step": 43821
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2830252044397435e-06,
+      "loss": 0.3321,
+      "step": 43822
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.282689369133377e-06,
+      "loss": 0.2485,
+      "step": 43823
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.282353533827011e-06,
+      "loss": 0.2812,
+      "step": 43824
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.282017698520646e-06,
+      "loss": 0.2429,
+      "step": 43825
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2816818632142796e-06,
+      "loss": 0.4125,
+      "step": 43826
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.281346027907914e-06,
+      "loss": 0.3153,
+      "step": 43827
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.281010192601549e-06,
+      "loss": 0.2465,
+      "step": 43828
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.280674357295183e-06,
+      "loss": 0.2655,
+      "step": 43829
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.280338521988817e-06,
+      "loss": 0.2725,
+      "step": 43830
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.280002686682452e-06,
+      "loss": 0.1702,
+      "step": 43831
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.279666851376085e-06,
+      "loss": 0.2582,
+      "step": 43832
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2793310160697196e-06,
+      "loss": 0.2383,
+      "step": 43833
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.278995180763354e-06,
+      "loss": 0.4292,
+      "step": 43834
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.278659345456988e-06,
+      "loss": 0.3049,
+      "step": 43835
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.278323510150623e-06,
+      "loss": 0.3117,
+      "step": 43836
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.277987674844257e-06,
+      "loss": 0.3497,
+      "step": 43837
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.27765183953789e-06,
+      "loss": 0.3289,
+      "step": 43838
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.277316004231525e-06,
+      "loss": 0.31,
+      "step": 43839
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2769801689251596e-06,
+      "loss": 0.2803,
+      "step": 43840
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.276644333618793e-06,
+      "loss": 0.3051,
+      "step": 43841
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.276308498312428e-06,
+      "loss": 0.2757,
+      "step": 43842
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.275972663006063e-06,
+      "loss": 0.3487,
+      "step": 43843
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2756368276996965e-06,
+      "loss": 0.3122,
+      "step": 43844
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.275300992393331e-06,
+      "loss": 0.386,
+      "step": 43845
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.274965157086966e-06,
+      "loss": 0.3579,
+      "step": 43846
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.274629321780599e-06,
+      "loss": 0.3124,
+      "step": 43847
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.274293486474233e-06,
+      "loss": 0.2214,
+      "step": 43848
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.273957651167868e-06,
+      "loss": 0.1895,
+      "step": 43849
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.273621815861502e-06,
+      "loss": 0.2605,
+      "step": 43850
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2732859805551364e-06,
+      "loss": 0.2667,
+      "step": 43851
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.272950145248771e-06,
+      "loss": 0.2008,
+      "step": 43852
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.272614309942404e-06,
+      "loss": 0.1434,
+      "step": 43853
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.272278474636039e-06,
+      "loss": 0.3972,
+      "step": 43854
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.271942639329673e-06,
+      "loss": 0.2142,
+      "step": 43855
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.271606804023307e-06,
+      "loss": 0.3365,
+      "step": 43856
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.271270968716942e-06,
+      "loss": 0.1662,
+      "step": 43857
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2709351334105764e-06,
+      "loss": 0.4024,
+      "step": 43858
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.270599298104209e-06,
+      "loss": 0.3408,
+      "step": 43859
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.270263462797844e-06,
+      "loss": 0.2044,
+      "step": 43860
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.269927627491479e-06,
+      "loss": 0.183,
+      "step": 43861
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2695917921851125e-06,
+      "loss": 0.3916,
+      "step": 43862
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.269255956878747e-06,
+      "loss": 0.1953,
+      "step": 43863
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.268920121572382e-06,
+      "loss": 0.2668,
+      "step": 43864
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2685842862660156e-06,
+      "loss": 0.3305,
+      "step": 43865
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.26824845095965e-06,
+      "loss": 0.2459,
+      "step": 43866
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.267912615653285e-06,
+      "loss": 0.1819,
+      "step": 43867
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.267576780346918e-06,
+      "loss": 0.2577,
+      "step": 43868
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2672409450405525e-06,
+      "loss": 0.268,
+      "step": 43869
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.266905109734187e-06,
+      "loss": 0.239,
+      "step": 43870
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.266569274427821e-06,
+      "loss": 0.2843,
+      "step": 43871
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2662334391214556e-06,
+      "loss": 0.1637,
+      "step": 43872
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.26589760381509e-06,
+      "loss": 0.3159,
+      "step": 43873
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.265561768508723e-06,
+      "loss": 0.2823,
+      "step": 43874
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.265225933202358e-06,
+      "loss": 0.1532,
+      "step": 43875
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2648900978959925e-06,
+      "loss": 0.1753,
+      "step": 43876
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.264554262589626e-06,
+      "loss": 0.2863,
+      "step": 43877
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.264218427283261e-06,
+      "loss": 0.3059,
+      "step": 43878
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2638825919768956e-06,
+      "loss": 0.343,
+      "step": 43879
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.263546756670529e-06,
+      "loss": 0.2155,
+      "step": 43880
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.263210921364163e-06,
+      "loss": 0.3584,
+      "step": 43881
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.262875086057798e-06,
+      "loss": 0.1101,
+      "step": 43882
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.262539250751432e-06,
+      "loss": 0.2504,
+      "step": 43883
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.262203415445066e-06,
+      "loss": 0.204,
+      "step": 43884
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.261867580138701e-06,
+      "loss": 0.1444,
+      "step": 43885
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.261531744832335e-06,
+      "loss": 0.474,
+      "step": 43886
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.261195909525969e-06,
+      "loss": 0.2068,
+      "step": 43887
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.260860074219604e-06,
+      "loss": 0.2479,
+      "step": 43888
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.260524238913237e-06,
+      "loss": 0.2347,
+      "step": 43889
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.260188403606872e-06,
+      "loss": 0.1367,
+      "step": 43890
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.259852568300506e-06,
+      "loss": 0.3047,
+      "step": 43891
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.25951673299414e-06,
+      "loss": 0.2998,
+      "step": 43892
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.259180897687775e-06,
+      "loss": 0.2636,
+      "step": 43893
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.258845062381409e-06,
+      "loss": 0.1658,
+      "step": 43894
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.258509227075042e-06,
+      "loss": 0.3595,
+      "step": 43895
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.258173391768677e-06,
+      "loss": 0.2046,
+      "step": 43896
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.257837556462312e-06,
+      "loss": 0.2135,
+      "step": 43897
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.257501721155945e-06,
+      "loss": 0.2282,
+      "step": 43898
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.25716588584958e-06,
+      "loss": 0.3406,
+      "step": 43899
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.256830050543214e-06,
+      "loss": 0.2599,
+      "step": 43900
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2564942152368485e-06,
+      "loss": 0.1787,
+      "step": 43901
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.256158379930483e-06,
+      "loss": 0.2959,
+      "step": 43902
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.255822544624116e-06,
+      "loss": 0.3384,
+      "step": 43903
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.255486709317751e-06,
+      "loss": 0.2702,
+      "step": 43904
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.255150874011385e-06,
+      "loss": 0.3276,
+      "step": 43905
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.254815038705019e-06,
+      "loss": 0.2098,
+      "step": 43906
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.254479203398654e-06,
+      "loss": 0.3851,
+      "step": 43907
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2541433680922885e-06,
+      "loss": 0.4272,
+      "step": 43908
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2538075327859214e-06,
+      "loss": 0.4449,
+      "step": 43909
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.253471697479556e-06,
+      "loss": 0.2458,
+      "step": 43910
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.253135862173191e-06,
+      "loss": 0.2508,
+      "step": 43911
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2528000268668245e-06,
+      "loss": 0.3402,
+      "step": 43912
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.252464191560459e-06,
+      "loss": 0.23,
+      "step": 43913
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.252128356254094e-06,
+      "loss": 0.2378,
+      "step": 43914
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.251792520947727e-06,
+      "loss": 0.0889,
+      "step": 43915
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.251456685641361e-06,
+      "loss": 0.2349,
+      "step": 43916
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.251120850334996e-06,
+      "loss": 0.2154,
+      "step": 43917
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.25078501502863e-06,
+      "loss": 0.3193,
+      "step": 43918
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2504491797222645e-06,
+      "loss": 0.3688,
+      "step": 43919
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.250113344415899e-06,
+      "loss": 0.3681,
+      "step": 43920
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.249777509109533e-06,
+      "loss": 0.375,
+      "step": 43921
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.249441673803168e-06,
+      "loss": 0.2044,
+      "step": 43922
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.249105838496802e-06,
+      "loss": 0.2424,
+      "step": 43923
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.248770003190435e-06,
+      "loss": 0.2753,
+      "step": 43924
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.24843416788407e-06,
+      "loss": 0.2471,
+      "step": 43925
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2480983325777045e-06,
+      "loss": 0.2332,
+      "step": 43926
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.247762497271338e-06,
+      "loss": 0.3355,
+      "step": 43927
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.247426661964973e-06,
+      "loss": 0.1419,
+      "step": 43928
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.247090826658608e-06,
+      "loss": 0.2407,
+      "step": 43929
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2467549913522406e-06,
+      "loss": 0.2485,
+      "step": 43930
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.246419156045875e-06,
+      "loss": 0.2646,
+      "step": 43931
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.24608332073951e-06,
+      "loss": 0.3162,
+      "step": 43932
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.245747485433144e-06,
+      "loss": 0.4296,
+      "step": 43933
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.245411650126778e-06,
+      "loss": 0.2941,
+      "step": 43934
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.245075814820413e-06,
+      "loss": 0.2547,
+      "step": 43935
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.244739979514047e-06,
+      "loss": 0.2278,
+      "step": 43936
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2444041442076805e-06,
+      "loss": 0.3503,
+      "step": 43937
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.244068308901315e-06,
+      "loss": 0.2633,
+      "step": 43938
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.243732473594949e-06,
+      "loss": 0.2409,
+      "step": 43939
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.243396638288584e-06,
+      "loss": 0.2676,
+      "step": 43940
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.243060802982218e-06,
+      "loss": 0.2725,
+      "step": 43941
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.242724967675852e-06,
+      "loss": 0.3124,
+      "step": 43942
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.242389132369487e-06,
+      "loss": 0.2256,
+      "step": 43943
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.242053297063121e-06,
+      "loss": 0.2332,
+      "step": 43944
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.241717461756754e-06,
+      "loss": 0.245,
+      "step": 43945
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.241381626450389e-06,
+      "loss": 0.3133,
+      "step": 43946
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.241045791144024e-06,
+      "loss": 0.3185,
+      "step": 43947
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2407099558376574e-06,
+      "loss": 0.228,
+      "step": 43948
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.240374120531292e-06,
+      "loss": 0.4468,
+      "step": 43949
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.240038285224927e-06,
+      "loss": 0.2626,
+      "step": 43950
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.23970244991856e-06,
+      "loss": 0.4353,
+      "step": 43951
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.239366614612194e-06,
+      "loss": 0.4619,
+      "step": 43952
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.239030779305829e-06,
+      "loss": 0.2536,
+      "step": 43953
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.238694943999463e-06,
+      "loss": 0.2701,
+      "step": 43954
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.238359108693097e-06,
+      "loss": 0.3172,
+      "step": 43955
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.238023273386732e-06,
+      "loss": 0.3332,
+      "step": 43956
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.237687438080366e-06,
+      "loss": 0.1529,
+      "step": 43957
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2373516027740005e-06,
+      "loss": 0.3723,
+      "step": 43958
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.237015767467634e-06,
+      "loss": 0.2475,
+      "step": 43959
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.236679932161268e-06,
+      "loss": 0.2018,
+      "step": 43960
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.236344096854903e-06,
+      "loss": 0.2007,
+      "step": 43961
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.236008261548537e-06,
+      "loss": 0.1982,
+      "step": 43962
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.235672426242171e-06,
+      "loss": 0.1896,
+      "step": 43963
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.235336590935806e-06,
+      "loss": 0.2301,
+      "step": 43964
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2350007556294405e-06,
+      "loss": 0.383,
+      "step": 43965
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2346649203230735e-06,
+      "loss": 0.227,
+      "step": 43966
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.234329085016708e-06,
+      "loss": 0.244,
+      "step": 43967
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.233993249710343e-06,
+      "loss": 0.1524,
+      "step": 43968
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 5.2336574144039766e-06,
+      "loss": 0.3753,
+      "step": 43969
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.233321579097611e-06,
+      "loss": 0.4446,
+      "step": 43970
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.232985743791246e-06,
+      "loss": 0.3037,
+      "step": 43971
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.232649908484879e-06,
+      "loss": 0.286,
+      "step": 43972
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.2323140731785134e-06,
+      "loss": 0.199,
+      "step": 43973
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.231978237872148e-06,
+      "loss": 0.2855,
+      "step": 43974
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.231642402565782e-06,
+      "loss": 0.325,
+      "step": 43975
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.2313065672594165e-06,
+      "loss": 0.1922,
+      "step": 43976
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.230970731953051e-06,
+      "loss": 0.3146,
+      "step": 43977
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.230634896646685e-06,
+      "loss": 0.222,
+      "step": 43978
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.23029906134032e-06,
+      "loss": 0.2981,
+      "step": 43979
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.229963226033954e-06,
+      "loss": 0.2566,
+      "step": 43980
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.229627390727587e-06,
+      "loss": 0.1436,
+      "step": 43981
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.229291555421222e-06,
+      "loss": 0.1185,
+      "step": 43982
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.2289557201148565e-06,
+      "loss": 0.1839,
+      "step": 43983
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.22861988480849e-06,
+      "loss": 0.1685,
+      "step": 43984
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.228284049502125e-06,
+      "loss": 0.219,
+      "step": 43985
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.22794821419576e-06,
+      "loss": 0.2768,
+      "step": 43986
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.227612378889393e-06,
+      "loss": 0.2488,
+      "step": 43987
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.227276543583027e-06,
+      "loss": 0.287,
+      "step": 43988
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.226940708276661e-06,
+      "loss": 0.2184,
+      "step": 43989
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.226604872970296e-06,
+      "loss": 0.2096,
+      "step": 43990
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.22626903766393e-06,
+      "loss": 0.2281,
+      "step": 43991
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.225933202357564e-06,
+      "loss": 0.4106,
+      "step": 43992
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.225597367051199e-06,
+      "loss": 0.2064,
+      "step": 43993
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.2252615317448326e-06,
+      "loss": 0.3004,
+      "step": 43994
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.224925696438466e-06,
+      "loss": 0.2444,
+      "step": 43995
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.224589861132101e-06,
+      "loss": 0.1877,
+      "step": 43996
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.224254025825736e-06,
+      "loss": 0.2491,
+      "step": 43997
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.2239181905193695e-06,
+      "loss": 0.1025,
+      "step": 43998
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.223582355213004e-06,
+      "loss": 0.2485,
+      "step": 43999
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.223246519906639e-06,
+      "loss": 0.2425,
+      "step": 44000
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.222910684600272e-06,
+      "loss": 0.1293,
+      "step": 44001
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.222574849293906e-06,
+      "loss": 0.2019,
+      "step": 44002
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.222239013987541e-06,
+      "loss": 0.1957,
+      "step": 44003
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.221903178681175e-06,
+      "loss": 0.2603,
+      "step": 44004
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.2215673433748095e-06,
+      "loss": 0.2669,
+      "step": 44005
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.221231508068444e-06,
+      "loss": 0.3253,
+      "step": 44006
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.220895672762077e-06,
+      "loss": 0.3543,
+      "step": 44007
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.220559837455712e-06,
+      "loss": 0.5423,
+      "step": 44008
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.220224002149346e-06,
+      "loss": 0.3914,
+      "step": 44009
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.21988816684298e-06,
+      "loss": 0.2006,
+      "step": 44010
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.219552331536615e-06,
+      "loss": 0.5025,
+      "step": 44011
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.2192164962302494e-06,
+      "loss": 0.4847,
+      "step": 44012
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.218880660923883e-06,
+      "loss": 0.3211,
+      "step": 44013
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.218544825617518e-06,
+      "loss": 0.2971,
+      "step": 44014
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.2182089903111525e-06,
+      "loss": 0.228,
+      "step": 44015
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.2178731550047855e-06,
+      "loss": 0.2609,
+      "step": 44016
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.21753731969842e-06,
+      "loss": 0.2697,
+      "step": 44017
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.217201484392055e-06,
+      "loss": 0.2499,
+      "step": 44018
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.216865649085689e-06,
+      "loss": 0.1923,
+      "step": 44019
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.216529813779323e-06,
+      "loss": 0.2907,
+      "step": 44020
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.216193978472958e-06,
+      "loss": 0.3061,
+      "step": 44021
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.215858143166591e-06,
+      "loss": 0.2693,
+      "step": 44022
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.2155223078602255e-06,
+      "loss": 0.301,
+      "step": 44023
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.21518647255386e-06,
+      "loss": 0.2021,
+      "step": 44024
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.214850637247494e-06,
+      "loss": 0.2626,
+      "step": 44025
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.214514801941129e-06,
+      "loss": 0.2586,
+      "step": 44026
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.214178966634763e-06,
+      "loss": 0.3591,
+      "step": 44027
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.213843131328396e-06,
+      "loss": 0.2341,
+      "step": 44028
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.213507296022031e-06,
+      "loss": 0.3127,
+      "step": 44029
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.2131714607156655e-06,
+      "loss": 0.3435,
+      "step": 44030
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.212835625409299e-06,
+      "loss": 0.3776,
+      "step": 44031
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.212499790102934e-06,
+      "loss": 0.2195,
+      "step": 44032
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.2121639547965686e-06,
+      "loss": 0.2585,
+      "step": 44033
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.211828119490202e-06,
+      "loss": 0.2625,
+      "step": 44034
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.211492284183837e-06,
+      "loss": 0.3244,
+      "step": 44035
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.211156448877472e-06,
+      "loss": 0.2099,
+      "step": 44036
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.210820613571105e-06,
+      "loss": 0.1207,
+      "step": 44037
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.210484778264739e-06,
+      "loss": 0.3605,
+      "step": 44038
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.210148942958374e-06,
+      "loss": 0.4137,
+      "step": 44039
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.209813107652008e-06,
+      "loss": 0.2672,
+      "step": 44040
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.209477272345642e-06,
+      "loss": 0.2179,
+      "step": 44041
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.209141437039277e-06,
+      "loss": 0.2107,
+      "step": 44042
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.20880560173291e-06,
+      "loss": 0.256,
+      "step": 44043
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.208469766426545e-06,
+      "loss": 0.3295,
+      "step": 44044
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.208133931120179e-06,
+      "loss": 0.1312,
+      "step": 44045
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.207798095813813e-06,
+      "loss": 0.4384,
+      "step": 44046
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.207462260507448e-06,
+      "loss": 0.3841,
+      "step": 44047
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.207126425201082e-06,
+      "loss": 0.1055,
+      "step": 44048
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.206790589894716e-06,
+      "loss": 0.2023,
+      "step": 44049
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.20645475458835e-06,
+      "loss": 0.1991,
+      "step": 44050
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.206118919281985e-06,
+      "loss": 0.2883,
+      "step": 44051
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.205783083975618e-06,
+      "loss": 0.2959,
+      "step": 44052
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.205447248669253e-06,
+      "loss": 0.3589,
+      "step": 44053
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.205111413362888e-06,
+      "loss": 0.3534,
+      "step": 44054
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.2047755780565215e-06,
+      "loss": 0.3021,
+      "step": 44055
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.204439742750156e-06,
+      "loss": 0.2273,
+      "step": 44056
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.204103907443791e-06,
+      "loss": 0.2343,
+      "step": 44057
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.203768072137424e-06,
+      "loss": 0.2463,
+      "step": 44058
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.203432236831058e-06,
+      "loss": 0.1996,
+      "step": 44059
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.203096401524693e-06,
+      "loss": 0.2771,
+      "step": 44060
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.202760566218327e-06,
+      "loss": 0.2935,
+      "step": 44061
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.2024247309119615e-06,
+      "loss": 0.2463,
+      "step": 44062
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.202088895605596e-06,
+      "loss": 0.4069,
+      "step": 44063
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.201753060299229e-06,
+      "loss": 0.3081,
+      "step": 44064
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.201417224992864e-06,
+      "loss": 0.1898,
+      "step": 44065
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.201081389686498e-06,
+      "loss": 0.204,
+      "step": 44066
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.200745554380132e-06,
+      "loss": 0.3192,
+      "step": 44067
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.200409719073767e-06,
+      "loss": 0.2291,
+      "step": 44068
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.2000738837674015e-06,
+      "loss": 0.3279,
+      "step": 44069
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.199738048461035e-06,
+      "loss": 0.5076,
+      "step": 44070
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.19940221315467e-06,
+      "loss": 0.4938,
+      "step": 44071
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.199066377848304e-06,
+      "loss": 0.2869,
+      "step": 44072
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1987305425419375e-06,
+      "loss": 0.1748,
+      "step": 44073
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.198394707235572e-06,
+      "loss": 0.3405,
+      "step": 44074
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.198058871929207e-06,
+      "loss": 0.1648,
+      "step": 44075
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.197723036622841e-06,
+      "loss": 0.2593,
+      "step": 44076
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.197387201316475e-06,
+      "loss": 0.3207,
+      "step": 44077
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.19705136601011e-06,
+      "loss": 0.1486,
+      "step": 44078
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.196715530703743e-06,
+      "loss": 0.4122,
+      "step": 44079
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1963796953973775e-06,
+      "loss": 0.1887,
+      "step": 44080
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.196043860091011e-06,
+      "loss": 0.3664,
+      "step": 44081
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.195708024784646e-06,
+      "loss": 0.1845,
+      "step": 44082
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.195372189478281e-06,
+      "loss": 0.3766,
+      "step": 44083
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1950363541719136e-06,
+      "loss": 0.31,
+      "step": 44084
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.194700518865548e-06,
+      "loss": 0.1747,
+      "step": 44085
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.194364683559183e-06,
+      "loss": 0.2508,
+      "step": 44086
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.194028848252817e-06,
+      "loss": 0.1712,
+      "step": 44087
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.193693012946451e-06,
+      "loss": 0.346,
+      "step": 44088
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.193357177640086e-06,
+      "loss": 0.3199,
+      "step": 44089
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.19302134233372e-06,
+      "loss": 0.1916,
+      "step": 44090
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.192685507027354e-06,
+      "loss": 0.4192,
+      "step": 44091
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.192349671720989e-06,
+      "loss": 0.1799,
+      "step": 44092
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.192013836414622e-06,
+      "loss": 0.2235,
+      "step": 44093
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.191678001108257e-06,
+      "loss": 0.3195,
+      "step": 44094
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.191342165801891e-06,
+      "loss": 0.2727,
+      "step": 44095
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.191006330495525e-06,
+      "loss": 0.2943,
+      "step": 44096
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.19067049518916e-06,
+      "loss": 0.1944,
+      "step": 44097
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.190334659882794e-06,
+      "loss": 0.3592,
+      "step": 44098
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.189998824576427e-06,
+      "loss": 0.2173,
+      "step": 44099
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.189662989270062e-06,
+      "loss": 0.2153,
+      "step": 44100
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.189327153963697e-06,
+      "loss": 0.1954,
+      "step": 44101
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1889913186573304e-06,
+      "loss": 0.3454,
+      "step": 44102
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.188655483350965e-06,
+      "loss": 0.2749,
+      "step": 44103
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1883196480446e-06,
+      "loss": 0.2909,
+      "step": 44104
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1879838127382335e-06,
+      "loss": 0.2313,
+      "step": 44105
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.187647977431868e-06,
+      "loss": 0.4115,
+      "step": 44106
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.187312142125502e-06,
+      "loss": 0.2712,
+      "step": 44107
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.186976306819136e-06,
+      "loss": 0.2353,
+      "step": 44108
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1866404715127704e-06,
+      "loss": 0.2277,
+      "step": 44109
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.186304636206405e-06,
+      "loss": 0.2281,
+      "step": 44110
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.185968800900039e-06,
+      "loss": 0.2531,
+      "step": 44111
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1856329655936735e-06,
+      "loss": 0.2171,
+      "step": 44112
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.185297130287308e-06,
+      "loss": 0.1105,
+      "step": 44113
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.184961294980941e-06,
+      "loss": 0.2718,
+      "step": 44114
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.184625459674576e-06,
+      "loss": 0.3332,
+      "step": 44115
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1842896243682104e-06,
+      "loss": 0.2989,
+      "step": 44116
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.183953789061844e-06,
+      "loss": 0.3313,
+      "step": 44117
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.183617953755479e-06,
+      "loss": 0.4289,
+      "step": 44118
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1832821184491135e-06,
+      "loss": 0.2588,
+      "step": 44119
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1829462831427465e-06,
+      "loss": 0.278,
+      "step": 44120
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.182610447836381e-06,
+      "loss": 0.1616,
+      "step": 44121
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.182274612530016e-06,
+      "loss": 0.2407,
+      "step": 44122
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1819387772236496e-06,
+      "loss": 0.2161,
+      "step": 44123
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.181602941917284e-06,
+      "loss": 0.3021,
+      "step": 44124
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.181267106610919e-06,
+      "loss": 0.3973,
+      "step": 44125
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.180931271304553e-06,
+      "loss": 0.2805,
+      "step": 44126
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.180595435998187e-06,
+      "loss": 0.2249,
+      "step": 44127
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.180259600691822e-06,
+      "loss": 0.3176,
+      "step": 44128
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.179923765385455e-06,
+      "loss": 0.2589,
+      "step": 44129
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1795879300790896e-06,
+      "loss": 0.4223,
+      "step": 44130
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.179252094772724e-06,
+      "loss": 0.3527,
+      "step": 44131
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.178916259466358e-06,
+      "loss": 0.2549,
+      "step": 44132
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.178580424159993e-06,
+      "loss": 0.3883,
+      "step": 44133
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.178244588853627e-06,
+      "loss": 0.1938,
+      "step": 44134
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.17790875354726e-06,
+      "loss": 0.1531,
+      "step": 44135
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.177572918240895e-06,
+      "loss": 0.1981,
+      "step": 44136
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1772370829345296e-06,
+      "loss": 0.2081,
+      "step": 44137
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.176901247628163e-06,
+      "loss": 0.2012,
+      "step": 44138
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.176565412321798e-06,
+      "loss": 0.1622,
+      "step": 44139
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.176229577015433e-06,
+      "loss": 0.2871,
+      "step": 44140
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.175893741709066e-06,
+      "loss": 0.3439,
+      "step": 44141
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1755579064027e-06,
+      "loss": 0.1822,
+      "step": 44142
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.175222071096335e-06,
+      "loss": 0.221,
+      "step": 44143
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.174886235789969e-06,
+      "loss": 0.243,
+      "step": 44144
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.174550400483603e-06,
+      "loss": 0.2479,
+      "step": 44145
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.174214565177238e-06,
+      "loss": 0.3309,
+      "step": 44146
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.173878729870872e-06,
+      "loss": 0.3377,
+      "step": 44147
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1735428945645064e-06,
+      "loss": 0.2112,
+      "step": 44148
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.173207059258141e-06,
+      "loss": 0.2699,
+      "step": 44149
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.172871223951774e-06,
+      "loss": 0.2966,
+      "step": 44150
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.172535388645409e-06,
+      "loss": 0.2562,
+      "step": 44151
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.172199553339043e-06,
+      "loss": 0.3271,
+      "step": 44152
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.171863718032677e-06,
+      "loss": 0.2179,
+      "step": 44153
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.171527882726312e-06,
+      "loss": 0.3811,
+      "step": 44154
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1711920474199464e-06,
+      "loss": 0.3105,
+      "step": 44155
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.170856212113579e-06,
+      "loss": 0.3462,
+      "step": 44156
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.170520376807214e-06,
+      "loss": 0.2356,
+      "step": 44157
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.170184541500849e-06,
+      "loss": 0.3457,
+      "step": 44158
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1698487061944825e-06,
+      "loss": 0.2079,
+      "step": 44159
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.169512870888117e-06,
+      "loss": 0.2291,
+      "step": 44160
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.169177035581752e-06,
+      "loss": 0.412,
+      "step": 44161
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1688412002753856e-06,
+      "loss": 0.4062,
+      "step": 44162
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.168505364969019e-06,
+      "loss": 0.3272,
+      "step": 44163
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.168169529662654e-06,
+      "loss": 0.4141,
+      "step": 44164
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.167833694356288e-06,
+      "loss": 0.2421,
+      "step": 44165
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.1674978590499225e-06,
+      "loss": 0.2015,
+      "step": 44166
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.167162023743557e-06,
+      "loss": 0.2048,
+      "step": 44167
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 5.166826188437191e-06,
+      "loss": 0.2325,
+      "step": 44168
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1664903531308256e-06,
+      "loss": 0.1173,
+      "step": 44169
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1661545178244585e-06,
+      "loss": 0.1457,
+      "step": 44170
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.165818682518093e-06,
+      "loss": 0.3085,
+      "step": 44171
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.165482847211728e-06,
+      "loss": 0.244,
+      "step": 44172
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.165147011905362e-06,
+      "loss": 0.3104,
+      "step": 44173
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.164811176598996e-06,
+      "loss": 0.2257,
+      "step": 44174
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.164475341292631e-06,
+      "loss": 0.3299,
+      "step": 44175
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.164139505986264e-06,
+      "loss": 0.271,
+      "step": 44176
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1638036706798985e-06,
+      "loss": 0.2961,
+      "step": 44177
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.163467835373533e-06,
+      "loss": 0.1957,
+      "step": 44178
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.163132000067167e-06,
+      "loss": 0.3708,
+      "step": 44179
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.162796164760802e-06,
+      "loss": 0.2285,
+      "step": 44180
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.162460329454436e-06,
+      "loss": 0.3423,
+      "step": 44181
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.16212449414807e-06,
+      "loss": 0.2234,
+      "step": 44182
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.161788658841705e-06,
+      "loss": 0.1122,
+      "step": 44183
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.161452823535339e-06,
+      "loss": 0.1854,
+      "step": 44184
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.161116988228972e-06,
+      "loss": 0.2135,
+      "step": 44185
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.160781152922607e-06,
+      "loss": 0.3468,
+      "step": 44186
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.160445317616242e-06,
+      "loss": 0.4327,
+      "step": 44187
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.160109482309875e-06,
+      "loss": 0.2814,
+      "step": 44188
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.15977364700351e-06,
+      "loss": 0.4039,
+      "step": 44189
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.159437811697145e-06,
+      "loss": 0.1752,
+      "step": 44190
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.159101976390778e-06,
+      "loss": 0.2251,
+      "step": 44191
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.158766141084412e-06,
+      "loss": 0.4016,
+      "step": 44192
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.158430305778047e-06,
+      "loss": 0.2515,
+      "step": 44193
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.158094470471681e-06,
+      "loss": 0.2175,
+      "step": 44194
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.157758635165315e-06,
+      "loss": 0.364,
+      "step": 44195
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.15742279985895e-06,
+      "loss": 0.4066,
+      "step": 44196
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.157086964552583e-06,
+      "loss": 0.2329,
+      "step": 44197
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.156751129246218e-06,
+      "loss": 0.2653,
+      "step": 44198
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.156415293939852e-06,
+      "loss": 0.2987,
+      "step": 44199
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.156079458633486e-06,
+      "loss": 0.3734,
+      "step": 44200
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.155743623327121e-06,
+      "loss": 0.3526,
+      "step": 44201
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.155407788020755e-06,
+      "loss": 0.2463,
+      "step": 44202
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.155071952714389e-06,
+      "loss": 0.3294,
+      "step": 44203
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.154736117408024e-06,
+      "loss": 0.3211,
+      "step": 44204
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1544002821016585e-06,
+      "loss": 0.3167,
+      "step": 44205
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.154064446795291e-06,
+      "loss": 0.2387,
+      "step": 44206
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.153728611488926e-06,
+      "loss": 0.3853,
+      "step": 44207
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.153392776182561e-06,
+      "loss": 0.304,
+      "step": 44208
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1530569408761945e-06,
+      "loss": 0.2377,
+      "step": 44209
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.152721105569829e-06,
+      "loss": 0.2336,
+      "step": 44210
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.152385270263464e-06,
+      "loss": 0.2729,
+      "step": 44211
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.152049434957097e-06,
+      "loss": 0.2894,
+      "step": 44212
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.151713599650731e-06,
+      "loss": 0.2971,
+      "step": 44213
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.151377764344366e-06,
+      "loss": 0.2926,
+      "step": 44214
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.151041929038e-06,
+      "loss": 0.1953,
+      "step": 44215
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1507060937316345e-06,
+      "loss": 0.2026,
+      "step": 44216
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.150370258425269e-06,
+      "loss": 0.2312,
+      "step": 44217
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.150034423118903e-06,
+      "loss": 0.1874,
+      "step": 44218
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.149698587812537e-06,
+      "loss": 0.2934,
+      "step": 44219
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.149362752506171e-06,
+      "loss": 0.1618,
+      "step": 44220
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.149026917199805e-06,
+      "loss": 0.1943,
+      "step": 44221
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.14869108189344e-06,
+      "loss": 0.3702,
+      "step": 44222
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1483552465870745e-06,
+      "loss": 0.337,
+      "step": 44223
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.148019411280708e-06,
+      "loss": 0.3186,
+      "step": 44224
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.147683575974343e-06,
+      "loss": 0.257,
+      "step": 44225
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.147347740667978e-06,
+      "loss": 0.3648,
+      "step": 44226
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1470119053616105e-06,
+      "loss": 0.3016,
+      "step": 44227
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.146676070055245e-06,
+      "loss": 0.2814,
+      "step": 44228
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.14634023474888e-06,
+      "loss": 0.3052,
+      "step": 44229
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.146004399442514e-06,
+      "loss": 0.1319,
+      "step": 44230
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.145668564136148e-06,
+      "loss": 0.2767,
+      "step": 44231
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.145332728829783e-06,
+      "loss": 0.1714,
+      "step": 44232
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.144996893523416e-06,
+      "loss": 0.2483,
+      "step": 44233
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1446610582170505e-06,
+      "loss": 0.3725,
+      "step": 44234
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.144325222910685e-06,
+      "loss": 0.3766,
+      "step": 44235
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.143989387604319e-06,
+      "loss": 0.2661,
+      "step": 44236
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.143653552297954e-06,
+      "loss": 0.3312,
+      "step": 44237
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.143317716991588e-06,
+      "loss": 0.1878,
+      "step": 44238
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.142981881685222e-06,
+      "loss": 0.1794,
+      "step": 44239
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.142646046378857e-06,
+      "loss": 0.3895,
+      "step": 44240
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.142310211072491e-06,
+      "loss": 0.1888,
+      "step": 44241
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.141974375766124e-06,
+      "loss": 0.1202,
+      "step": 44242
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.141638540459759e-06,
+      "loss": 0.4225,
+      "step": 44243
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.141302705153394e-06,
+      "loss": 0.3375,
+      "step": 44244
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.140966869847027e-06,
+      "loss": 0.341,
+      "step": 44245
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.140631034540662e-06,
+      "loss": 0.344,
+      "step": 44246
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.140295199234297e-06,
+      "loss": 0.2689,
+      "step": 44247
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.13995936392793e-06,
+      "loss": 0.1655,
+      "step": 44248
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.139623528621564e-06,
+      "loss": 0.3069,
+      "step": 44249
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.139287693315199e-06,
+      "loss": 0.2942,
+      "step": 44250
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.138951858008833e-06,
+      "loss": 0.1977,
+      "step": 44251
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.138616022702467e-06,
+      "loss": 0.534,
+      "step": 44252
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.138280187396102e-06,
+      "loss": 0.2318,
+      "step": 44253
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.137944352089735e-06,
+      "loss": 0.1216,
+      "step": 44254
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.13760851678337e-06,
+      "loss": 0.3476,
+      "step": 44255
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.137272681477004e-06,
+      "loss": 0.3287,
+      "step": 44256
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.136936846170638e-06,
+      "loss": 0.2994,
+      "step": 44257
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.136601010864273e-06,
+      "loss": 0.4319,
+      "step": 44258
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1362651755579066e-06,
+      "loss": 0.4623,
+      "step": 44259
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.135929340251541e-06,
+      "loss": 0.3973,
+      "step": 44260
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.135593504945176e-06,
+      "loss": 0.3046,
+      "step": 44261
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.135257669638809e-06,
+      "loss": 0.3082,
+      "step": 44262
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1349218343324435e-06,
+      "loss": 0.1386,
+      "step": 44263
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.134585999026078e-06,
+      "loss": 0.1803,
+      "step": 44264
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.134250163719712e-06,
+      "loss": 0.281,
+      "step": 44265
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1339143284133465e-06,
+      "loss": 0.2842,
+      "step": 44266
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.133578493106981e-06,
+      "loss": 0.1752,
+      "step": 44267
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.133242657800614e-06,
+      "loss": 0.2487,
+      "step": 44268
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.132906822494249e-06,
+      "loss": 0.2023,
+      "step": 44269
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1325709871878834e-06,
+      "loss": 0.2766,
+      "step": 44270
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.132235151881517e-06,
+      "loss": 0.3229,
+      "step": 44271
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.131899316575152e-06,
+      "loss": 0.2939,
+      "step": 44272
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1315634812687865e-06,
+      "loss": 0.3782,
+      "step": 44273
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.13122764596242e-06,
+      "loss": 0.1764,
+      "step": 44274
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.130891810656055e-06,
+      "loss": 0.3107,
+      "step": 44275
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.130555975349689e-06,
+      "loss": 0.2661,
+      "step": 44276
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.130220140043323e-06,
+      "loss": 0.2443,
+      "step": 44277
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.129884304736957e-06,
+      "loss": 0.1688,
+      "step": 44278
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.129548469430592e-06,
+      "loss": 0.4008,
+      "step": 44279
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.129212634124226e-06,
+      "loss": 0.2184,
+      "step": 44280
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.12887679881786e-06,
+      "loss": 0.319,
+      "step": 44281
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.128540963511495e-06,
+      "loss": 0.1944,
+      "step": 44282
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.128205128205128e-06,
+      "loss": 0.2673,
+      "step": 44283
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.127869292898763e-06,
+      "loss": 0.4222,
+      "step": 44284
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.127533457592397e-06,
+      "loss": 0.2504,
+      "step": 44285
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.127197622286031e-06,
+      "loss": 0.3624,
+      "step": 44286
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.126861786979666e-06,
+      "loss": 0.1403,
+      "step": 44287
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1265259516733e-06,
+      "loss": 0.2541,
+      "step": 44288
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.126190116366933e-06,
+      "loss": 0.3015,
+      "step": 44289
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.125854281060568e-06,
+      "loss": 0.1642,
+      "step": 44290
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1255184457542026e-06,
+      "loss": 0.252,
+      "step": 44291
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.125182610447836e-06,
+      "loss": 0.3836,
+      "step": 44292
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.124846775141471e-06,
+      "loss": 0.336,
+      "step": 44293
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.124510939835106e-06,
+      "loss": 0.1474,
+      "step": 44294
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1241751045287395e-06,
+      "loss": 0.3402,
+      "step": 44295
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.123839269222374e-06,
+      "loss": 0.3822,
+      "step": 44296
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.123503433916009e-06,
+      "loss": 0.2042,
+      "step": 44297
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.123167598609642e-06,
+      "loss": 0.2759,
+      "step": 44298
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.122831763303276e-06,
+      "loss": 0.2397,
+      "step": 44299
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.122495927996911e-06,
+      "loss": 0.2271,
+      "step": 44300
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.122160092690545e-06,
+      "loss": 0.3642,
+      "step": 44301
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1218242573841795e-06,
+      "loss": 0.2107,
+      "step": 44302
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.121488422077814e-06,
+      "loss": 0.3799,
+      "step": 44303
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.121152586771447e-06,
+      "loss": 0.3073,
+      "step": 44304
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.120816751465082e-06,
+      "loss": 0.2451,
+      "step": 44305
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.120480916158716e-06,
+      "loss": 0.2806,
+      "step": 44306
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.12014508085235e-06,
+      "loss": 0.2388,
+      "step": 44307
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.119809245545985e-06,
+      "loss": 0.4171,
+      "step": 44308
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1194734102396194e-06,
+      "loss": 0.1841,
+      "step": 44309
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.119137574933252e-06,
+      "loss": 0.1109,
+      "step": 44310
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.118801739626887e-06,
+      "loss": 0.3813,
+      "step": 44311
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.118465904320522e-06,
+      "loss": 0.3598,
+      "step": 44312
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1181300690141555e-06,
+      "loss": 0.216,
+      "step": 44313
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.11779423370779e-06,
+      "loss": 0.2607,
+      "step": 44314
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.117458398401425e-06,
+      "loss": 0.3232,
+      "step": 44315
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.117122563095059e-06,
+      "loss": 0.1988,
+      "step": 44316
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.116786727788693e-06,
+      "loss": 0.4127,
+      "step": 44317
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.116450892482328e-06,
+      "loss": 0.2614,
+      "step": 44318
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.116115057175961e-06,
+      "loss": 0.3713,
+      "step": 44319
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1157792218695955e-06,
+      "loss": 0.3018,
+      "step": 44320
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.11544338656323e-06,
+      "loss": 0.1902,
+      "step": 44321
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.115107551256864e-06,
+      "loss": 0.3562,
+      "step": 44322
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.114771715950499e-06,
+      "loss": 0.2998,
+      "step": 44323
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.114435880644133e-06,
+      "loss": 0.1899,
+      "step": 44324
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.114100045337766e-06,
+      "loss": 0.2122,
+      "step": 44325
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.113764210031401e-06,
+      "loss": 0.3013,
+      "step": 44326
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1134283747250355e-06,
+      "loss": 0.2344,
+      "step": 44327
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.113092539418669e-06,
+      "loss": 0.2396,
+      "step": 44328
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.112756704112304e-06,
+      "loss": 0.1782,
+      "step": 44329
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1124208688059386e-06,
+      "loss": 0.3702,
+      "step": 44330
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.112085033499572e-06,
+      "loss": 0.2918,
+      "step": 44331
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.111749198193206e-06,
+      "loss": 0.2162,
+      "step": 44332
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.111413362886841e-06,
+      "loss": 0.2055,
+      "step": 44333
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.111077527580475e-06,
+      "loss": 0.16,
+      "step": 44334
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.110741692274109e-06,
+      "loss": 0.182,
+      "step": 44335
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.110405856967744e-06,
+      "loss": 0.2506,
+      "step": 44336
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.110070021661378e-06,
+      "loss": 0.2874,
+      "step": 44337
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.109734186355012e-06,
+      "loss": 0.2613,
+      "step": 44338
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.109398351048647e-06,
+      "loss": 0.3267,
+      "step": 44339
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.10906251574228e-06,
+      "loss": 0.2172,
+      "step": 44340
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.108726680435915e-06,
+      "loss": 0.2334,
+      "step": 44341
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.108390845129549e-06,
+      "loss": 0.2463,
+      "step": 44342
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.108055009823183e-06,
+      "loss": 0.1591,
+      "step": 44343
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.107719174516818e-06,
+      "loss": 0.1557,
+      "step": 44344
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.107383339210452e-06,
+      "loss": 0.2809,
+      "step": 44345
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.107047503904085e-06,
+      "loss": 0.216,
+      "step": 44346
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.10671166859772e-06,
+      "loss": 0.2254,
+      "step": 44347
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.106375833291355e-06,
+      "loss": 0.258,
+      "step": 44348
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.106039997984988e-06,
+      "loss": 0.2001,
+      "step": 44349
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.105704162678623e-06,
+      "loss": 0.4078,
+      "step": 44350
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.105368327372257e-06,
+      "loss": 0.2923,
+      "step": 44351
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1050324920658915e-06,
+      "loss": 0.3179,
+      "step": 44352
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.104696656759526e-06,
+      "loss": 0.1955,
+      "step": 44353
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.104360821453159e-06,
+      "loss": 0.3782,
+      "step": 44354
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.104024986146794e-06,
+      "loss": 0.2446,
+      "step": 44355
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.103689150840428e-06,
+      "loss": 0.2428,
+      "step": 44356
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.103353315534062e-06,
+      "loss": 0.5145,
+      "step": 44357
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.103017480227697e-06,
+      "loss": 0.2881,
+      "step": 44358
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1026816449213315e-06,
+      "loss": 0.2837,
+      "step": 44359
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1023458096149644e-06,
+      "loss": 0.3124,
+      "step": 44360
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.102009974308599e-06,
+      "loss": 0.1991,
+      "step": 44361
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.101674139002234e-06,
+      "loss": 0.2826,
+      "step": 44362
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.1013383036958675e-06,
+      "loss": 0.2426,
+      "step": 44363
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.101002468389502e-06,
+      "loss": 0.1849,
+      "step": 44364
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.100666633083137e-06,
+      "loss": 0.1914,
+      "step": 44365
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 5.10033079777677e-06,
+      "loss": 0.2768,
+      "step": 44366
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0999949624704044e-06,
+      "loss": 0.4372,
+      "step": 44367
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.099659127164039e-06,
+      "loss": 0.3766,
+      "step": 44368
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.099323291857673e-06,
+      "loss": 0.3673,
+      "step": 44369
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0989874565513075e-06,
+      "loss": 0.2404,
+      "step": 44370
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.098651621244942e-06,
+      "loss": 0.276,
+      "step": 44371
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.098315785938576e-06,
+      "loss": 0.2119,
+      "step": 44372
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.097979950632211e-06,
+      "loss": 0.1917,
+      "step": 44373
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.097644115325845e-06,
+      "loss": 0.208,
+      "step": 44374
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.097308280019478e-06,
+      "loss": 0.2107,
+      "step": 44375
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.096972444713113e-06,
+      "loss": 0.2915,
+      "step": 44376
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0966366094067475e-06,
+      "loss": 0.2185,
+      "step": 44377
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.096300774100381e-06,
+      "loss": 0.4005,
+      "step": 44378
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.095964938794016e-06,
+      "loss": 0.2798,
+      "step": 44379
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.095629103487651e-06,
+      "loss": 0.292,
+      "step": 44380
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0952932681812836e-06,
+      "loss": 0.2813,
+      "step": 44381
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.094957432874918e-06,
+      "loss": 0.2831,
+      "step": 44382
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.094621597568553e-06,
+      "loss": 0.2444,
+      "step": 44383
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.094285762262187e-06,
+      "loss": 0.2034,
+      "step": 44384
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.093949926955821e-06,
+      "loss": 0.3202,
+      "step": 44385
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.093614091649456e-06,
+      "loss": 0.1693,
+      "step": 44386
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.09327825634309e-06,
+      "loss": 0.2332,
+      "step": 44387
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.092942421036724e-06,
+      "loss": 0.2987,
+      "step": 44388
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.092606585730358e-06,
+      "loss": 0.2164,
+      "step": 44389
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.092270750423992e-06,
+      "loss": 0.3111,
+      "step": 44390
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.091934915117627e-06,
+      "loss": 0.2982,
+      "step": 44391
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.091599079811261e-06,
+      "loss": 0.3002,
+      "step": 44392
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.091263244504895e-06,
+      "loss": 0.2287,
+      "step": 44393
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.09092740919853e-06,
+      "loss": 0.348,
+      "step": 44394
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.090591573892164e-06,
+      "loss": 0.2927,
+      "step": 44395
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.090255738585797e-06,
+      "loss": 0.3925,
+      "step": 44396
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.089919903279432e-06,
+      "loss": 0.3596,
+      "step": 44397
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.089584067973067e-06,
+      "loss": 0.2462,
+      "step": 44398
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0892482326667004e-06,
+      "loss": 0.4574,
+      "step": 44399
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.088912397360335e-06,
+      "loss": 0.2514,
+      "step": 44400
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.08857656205397e-06,
+      "loss": 0.33,
+      "step": 44401
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.088240726747603e-06,
+      "loss": 0.3195,
+      "step": 44402
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.087904891441237e-06,
+      "loss": 0.2999,
+      "step": 44403
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.087569056134872e-06,
+      "loss": 0.1425,
+      "step": 44404
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.087233220828506e-06,
+      "loss": 0.2438,
+      "step": 44405
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0868973855221404e-06,
+      "loss": 0.4401,
+      "step": 44406
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.086561550215775e-06,
+      "loss": 0.3074,
+      "step": 44407
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.086225714909409e-06,
+      "loss": 0.3243,
+      "step": 44408
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0858898796030435e-06,
+      "loss": 0.3423,
+      "step": 44409
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.085554044296678e-06,
+      "loss": 0.2237,
+      "step": 44410
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.085218208990311e-06,
+      "loss": 0.2296,
+      "step": 44411
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.084882373683946e-06,
+      "loss": 0.2158,
+      "step": 44412
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.08454653837758e-06,
+      "loss": 0.2683,
+      "step": 44413
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.084210703071214e-06,
+      "loss": 0.3513,
+      "step": 44414
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.083874867764849e-06,
+      "loss": 0.2238,
+      "step": 44415
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0835390324584835e-06,
+      "loss": 0.349,
+      "step": 44416
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0832031971521165e-06,
+      "loss": 0.3039,
+      "step": 44417
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.082867361845751e-06,
+      "loss": 0.1394,
+      "step": 44418
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.082531526539386e-06,
+      "loss": 0.2818,
+      "step": 44419
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0821956912330196e-06,
+      "loss": 0.3063,
+      "step": 44420
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.081859855926654e-06,
+      "loss": 0.2659,
+      "step": 44421
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.081524020620289e-06,
+      "loss": 0.2965,
+      "step": 44422
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.081188185313922e-06,
+      "loss": 0.1822,
+      "step": 44423
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0808523500075565e-06,
+      "loss": 0.2384,
+      "step": 44424
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.080516514701191e-06,
+      "loss": 0.2788,
+      "step": 44425
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.080180679394825e-06,
+      "loss": 0.2234,
+      "step": 44426
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0798448440884596e-06,
+      "loss": 0.2766,
+      "step": 44427
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.079509008782094e-06,
+      "loss": 0.3833,
+      "step": 44428
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.079173173475728e-06,
+      "loss": 0.268,
+      "step": 44429
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.078837338169363e-06,
+      "loss": 0.3093,
+      "step": 44430
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.078501502862997e-06,
+      "loss": 0.3201,
+      "step": 44431
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.07816566755663e-06,
+      "loss": 0.1981,
+      "step": 44432
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.077829832250265e-06,
+      "loss": 0.2263,
+      "step": 44433
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0774939969438995e-06,
+      "loss": 0.3323,
+      "step": 44434
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.077158161637533e-06,
+      "loss": 0.3839,
+      "step": 44435
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.076822326331168e-06,
+      "loss": 0.3178,
+      "step": 44436
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.076486491024803e-06,
+      "loss": 0.405,
+      "step": 44437
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.076150655718436e-06,
+      "loss": 0.2452,
+      "step": 44438
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.07581482041207e-06,
+      "loss": 0.1592,
+      "step": 44439
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.075478985105704e-06,
+      "loss": 0.1965,
+      "step": 44440
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.075143149799339e-06,
+      "loss": 0.4138,
+      "step": 44441
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.074807314492973e-06,
+      "loss": 0.3604,
+      "step": 44442
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.074471479186607e-06,
+      "loss": 0.2716,
+      "step": 44443
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.074135643880242e-06,
+      "loss": 0.3145,
+      "step": 44444
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.073799808573876e-06,
+      "loss": 0.274,
+      "step": 44445
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.073463973267509e-06,
+      "loss": 0.202,
+      "step": 44446
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.073128137961144e-06,
+      "loss": 0.3622,
+      "step": 44447
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.072792302654779e-06,
+      "loss": 0.2992,
+      "step": 44448
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0724564673484125e-06,
+      "loss": 0.2569,
+      "step": 44449
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.072120632042047e-06,
+      "loss": 0.2026,
+      "step": 44450
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.071784796735682e-06,
+      "loss": 0.257,
+      "step": 44451
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.071448961429315e-06,
+      "loss": 0.2109,
+      "step": 44452
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.071113126122949e-06,
+      "loss": 0.1846,
+      "step": 44453
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.070777290816584e-06,
+      "loss": 0.3736,
+      "step": 44454
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.070441455510218e-06,
+      "loss": 0.153,
+      "step": 44455
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0701056202038525e-06,
+      "loss": 0.2364,
+      "step": 44456
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.069769784897487e-06,
+      "loss": 0.2348,
+      "step": 44457
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.06943394959112e-06,
+      "loss": 0.3055,
+      "step": 44458
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.069098114284755e-06,
+      "loss": 0.3594,
+      "step": 44459
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.068762278978389e-06,
+      "loss": 0.4458,
+      "step": 44460
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.068426443672023e-06,
+      "loss": 0.203,
+      "step": 44461
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.068090608365658e-06,
+      "loss": 0.1736,
+      "step": 44462
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0677547730592925e-06,
+      "loss": 0.1083,
+      "step": 44463
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.067418937752926e-06,
+      "loss": 0.2835,
+      "step": 44464
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.067083102446561e-06,
+      "loss": 0.1769,
+      "step": 44465
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0667472671401956e-06,
+      "loss": 0.4662,
+      "step": 44466
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0664114318338285e-06,
+      "loss": 0.2805,
+      "step": 44467
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.066075596527463e-06,
+      "loss": 0.3325,
+      "step": 44468
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.065739761221098e-06,
+      "loss": 0.1246,
+      "step": 44469
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.065403925914732e-06,
+      "loss": 0.2378,
+      "step": 44470
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.065068090608366e-06,
+      "loss": 0.2059,
+      "step": 44471
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.064732255302001e-06,
+      "loss": 0.2355,
+      "step": 44472
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.064396419995634e-06,
+      "loss": 0.2987,
+      "step": 44473
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0640605846892685e-06,
+      "loss": 0.2529,
+      "step": 44474
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.063724749382903e-06,
+      "loss": 0.3303,
+      "step": 44475
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.063388914076537e-06,
+      "loss": 0.2877,
+      "step": 44476
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.063053078770172e-06,
+      "loss": 0.3587,
+      "step": 44477
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.062717243463806e-06,
+      "loss": 0.2323,
+      "step": 44478
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.062381408157439e-06,
+      "loss": 0.2432,
+      "step": 44479
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.062045572851074e-06,
+      "loss": 0.1821,
+      "step": 44480
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0617097375447085e-06,
+      "loss": 0.4313,
+      "step": 44481
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.061373902238342e-06,
+      "loss": 0.3627,
+      "step": 44482
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.061038066931977e-06,
+      "loss": 0.2535,
+      "step": 44483
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.060702231625612e-06,
+      "loss": 0.2995,
+      "step": 44484
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.060366396319245e-06,
+      "loss": 0.3632,
+      "step": 44485
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.06003056101288e-06,
+      "loss": 0.4702,
+      "step": 44486
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.059694725706515e-06,
+      "loss": 0.3942,
+      "step": 44487
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.059358890400148e-06,
+      "loss": 0.2371,
+      "step": 44488
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.059023055093782e-06,
+      "loss": 0.2362,
+      "step": 44489
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.058687219787417e-06,
+      "loss": 0.3271,
+      "step": 44490
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.058351384481051e-06,
+      "loss": 0.135,
+      "step": 44491
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.058015549174685e-06,
+      "loss": 0.2233,
+      "step": 44492
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.05767971386832e-06,
+      "loss": 0.3111,
+      "step": 44493
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.057343878561953e-06,
+      "loss": 0.3519,
+      "step": 44494
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.057008043255588e-06,
+      "loss": 0.2771,
+      "step": 44495
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.056672207949222e-06,
+      "loss": 0.3015,
+      "step": 44496
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.056336372642856e-06,
+      "loss": 0.1349,
+      "step": 44497
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.056000537336491e-06,
+      "loss": 0.136,
+      "step": 44498
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.055664702030125e-06,
+      "loss": 0.2979,
+      "step": 44499
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.055328866723759e-06,
+      "loss": 0.4114,
+      "step": 44500
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.054993031417394e-06,
+      "loss": 0.3625,
+      "step": 44501
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.054657196111028e-06,
+      "loss": 0.2477,
+      "step": 44502
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.054321360804661e-06,
+      "loss": 0.3492,
+      "step": 44503
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.053985525498296e-06,
+      "loss": 0.1876,
+      "step": 44504
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.053649690191931e-06,
+      "loss": 0.2338,
+      "step": 44505
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0533138548855645e-06,
+      "loss": 0.3868,
+      "step": 44506
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.052978019579199e-06,
+      "loss": 0.2523,
+      "step": 44507
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.052642184272834e-06,
+      "loss": 0.1749,
+      "step": 44508
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.052306348966467e-06,
+      "loss": 0.2125,
+      "step": 44509
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.051970513660101e-06,
+      "loss": 0.3592,
+      "step": 44510
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.051634678353736e-06,
+      "loss": 0.3446,
+      "step": 44511
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.05129884304737e-06,
+      "loss": 0.1722,
+      "step": 44512
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0509630077410045e-06,
+      "loss": 0.3518,
+      "step": 44513
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.050627172434639e-06,
+      "loss": 0.3017,
+      "step": 44514
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.050291337128272e-06,
+      "loss": 0.2293,
+      "step": 44515
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.049955501821907e-06,
+      "loss": 0.1844,
+      "step": 44516
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.049619666515541e-06,
+      "loss": 0.2837,
+      "step": 44517
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.049283831209175e-06,
+      "loss": 0.1154,
+      "step": 44518
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.04894799590281e-06,
+      "loss": 0.1726,
+      "step": 44519
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0486121605964445e-06,
+      "loss": 0.3227,
+      "step": 44520
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.048276325290078e-06,
+      "loss": 0.2445,
+      "step": 44521
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.047940489983713e-06,
+      "loss": 0.4806,
+      "step": 44522
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.047604654677348e-06,
+      "loss": 0.2285,
+      "step": 44523
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0472688193709805e-06,
+      "loss": 0.3708,
+      "step": 44524
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.046932984064615e-06,
+      "loss": 0.2323,
+      "step": 44525
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.04659714875825e-06,
+      "loss": 0.3163,
+      "step": 44526
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.046261313451884e-06,
+      "loss": 0.2463,
+      "step": 44527
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.045925478145518e-06,
+      "loss": 0.2513,
+      "step": 44528
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.045589642839153e-06,
+      "loss": 0.2891,
+      "step": 44529
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.045253807532786e-06,
+      "loss": 0.3445,
+      "step": 44530
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0449179722264205e-06,
+      "loss": 0.1931,
+      "step": 44531
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.044582136920054e-06,
+      "loss": 0.294,
+      "step": 44532
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.044246301613689e-06,
+      "loss": 0.2237,
+      "step": 44533
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.043910466307324e-06,
+      "loss": 0.3884,
+      "step": 44534
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0435746310009574e-06,
+      "loss": 0.2989,
+      "step": 44535
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.043238795694591e-06,
+      "loss": 0.1512,
+      "step": 44536
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.042902960388226e-06,
+      "loss": 0.2548,
+      "step": 44537
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.04256712508186e-06,
+      "loss": 0.251,
+      "step": 44538
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.042231289775494e-06,
+      "loss": 0.3284,
+      "step": 44539
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.041895454469129e-06,
+      "loss": 0.2504,
+      "step": 44540
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.041559619162763e-06,
+      "loss": 0.4187,
+      "step": 44541
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.041223783856397e-06,
+      "loss": 0.2331,
+      "step": 44542
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.040887948550032e-06,
+      "loss": 0.3902,
+      "step": 44543
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.040552113243665e-06,
+      "loss": 0.2273,
+      "step": 44544
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0402162779373e-06,
+      "loss": 0.2307,
+      "step": 44545
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.039880442630934e-06,
+      "loss": 0.3888,
+      "step": 44546
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.039544607324568e-06,
+      "loss": 0.3007,
+      "step": 44547
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.039208772018203e-06,
+      "loss": 0.2954,
+      "step": 44548
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.038872936711837e-06,
+      "loss": 0.2481,
+      "step": 44549
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.03853710140547e-06,
+      "loss": 0.2979,
+      "step": 44550
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.038201266099105e-06,
+      "loss": 0.3252,
+      "step": 44551
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.03786543079274e-06,
+      "loss": 0.2782,
+      "step": 44552
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0375295954863735e-06,
+      "loss": 0.1887,
+      "step": 44553
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.037193760180008e-06,
+      "loss": 0.2754,
+      "step": 44554
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.036857924873643e-06,
+      "loss": 0.374,
+      "step": 44555
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0365220895672766e-06,
+      "loss": 0.2237,
+      "step": 44556
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.036186254260911e-06,
+      "loss": 0.1907,
+      "step": 44557
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.035850418954545e-06,
+      "loss": 0.2855,
+      "step": 44558
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.035514583648179e-06,
+      "loss": 0.1833,
+      "step": 44559
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0351787483418134e-06,
+      "loss": 0.2345,
+      "step": 44560
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.034842913035448e-06,
+      "loss": 0.0763,
+      "step": 44561
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.034507077729082e-06,
+      "loss": 0.2058,
+      "step": 44562
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.0341712424227165e-06,
+      "loss": 0.2348,
+      "step": 44563
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.033835407116351e-06,
+      "loss": 0.2044,
+      "step": 44564
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 5.033499571809984e-06,
+      "loss": 0.2946,
+      "step": 44565
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.033163736503619e-06,
+      "loss": 0.2858,
+      "step": 44566
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.0328279011972534e-06,
+      "loss": 0.3331,
+      "step": 44567
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.032492065890887e-06,
+      "loss": 0.378,
+      "step": 44568
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.032156230584522e-06,
+      "loss": 0.2393,
+      "step": 44569
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.0318203952781565e-06,
+      "loss": 0.3052,
+      "step": 44570
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.0314845599717895e-06,
+      "loss": 0.1987,
+      "step": 44571
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.031148724665424e-06,
+      "loss": 0.2562,
+      "step": 44572
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.030812889359059e-06,
+      "loss": 0.2106,
+      "step": 44573
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.030477054052693e-06,
+      "loss": 0.2375,
+      "step": 44574
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.030141218746327e-06,
+      "loss": 0.207,
+      "step": 44575
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.029805383439962e-06,
+      "loss": 0.2421,
+      "step": 44576
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.029469548133596e-06,
+      "loss": 0.2603,
+      "step": 44577
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.02913371282723e-06,
+      "loss": 0.1795,
+      "step": 44578
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.028797877520865e-06,
+      "loss": 0.2992,
+      "step": 44579
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.028462042214498e-06,
+      "loss": 0.1984,
+      "step": 44580
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.0281262069081326e-06,
+      "loss": 0.151,
+      "step": 44581
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.027790371601767e-06,
+      "loss": 0.2743,
+      "step": 44582
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.027454536295401e-06,
+      "loss": 0.4802,
+      "step": 44583
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.027118700989036e-06,
+      "loss": 0.1899,
+      "step": 44584
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.02678286568267e-06,
+      "loss": 0.2864,
+      "step": 44585
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.026447030376303e-06,
+      "loss": 0.1453,
+      "step": 44586
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.026111195069938e-06,
+      "loss": 0.1872,
+      "step": 44587
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.0257753597635726e-06,
+      "loss": 0.233,
+      "step": 44588
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.025439524457206e-06,
+      "loss": 0.2793,
+      "step": 44589
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.025103689150841e-06,
+      "loss": 0.3121,
+      "step": 44590
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.024767853844476e-06,
+      "loss": 0.2385,
+      "step": 44591
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.024432018538109e-06,
+      "loss": 0.2385,
+      "step": 44592
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.024096183231743e-06,
+      "loss": 0.1626,
+      "step": 44593
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.023760347925378e-06,
+      "loss": 0.2933,
+      "step": 44594
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.023424512619012e-06,
+      "loss": 0.2659,
+      "step": 44595
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.023088677312646e-06,
+      "loss": 0.4132,
+      "step": 44596
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.022752842006281e-06,
+      "loss": 0.2144,
+      "step": 44597
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.022417006699915e-06,
+      "loss": 0.3679,
+      "step": 44598
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.0220811713935494e-06,
+      "loss": 0.2782,
+      "step": 44599
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.021745336087184e-06,
+      "loss": 0.378,
+      "step": 44600
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.021409500780817e-06,
+      "loss": 0.2379,
+      "step": 44601
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.021073665474452e-06,
+      "loss": 0.2661,
+      "step": 44602
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.020737830168086e-06,
+      "loss": 0.2563,
+      "step": 44603
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.02040199486172e-06,
+      "loss": 0.2272,
+      "step": 44604
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.020066159555355e-06,
+      "loss": 0.2325,
+      "step": 44605
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.0197303242489894e-06,
+      "loss": 0.2825,
+      "step": 44606
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.019394488942622e-06,
+      "loss": 0.2815,
+      "step": 44607
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.019058653636257e-06,
+      "loss": 0.2294,
+      "step": 44608
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.018722818329892e-06,
+      "loss": 0.3124,
+      "step": 44609
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.0183869830235255e-06,
+      "loss": 0.2203,
+      "step": 44610
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.01805114771716e-06,
+      "loss": 0.193,
+      "step": 44611
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.017715312410795e-06,
+      "loss": 0.2635,
+      "step": 44612
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.017379477104429e-06,
+      "loss": 0.1868,
+      "step": 44613
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.017043641798062e-06,
+      "loss": 0.4286,
+      "step": 44614
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.016707806491697e-06,
+      "loss": 0.3196,
+      "step": 44615
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.016371971185331e-06,
+      "loss": 0.1254,
+      "step": 44616
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.0160361358789655e-06,
+      "loss": 0.1878,
+      "step": 44617
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.0157003005726e-06,
+      "loss": 0.2929,
+      "step": 44618
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.015364465266234e-06,
+      "loss": 0.187,
+      "step": 44619
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.0150286299598686e-06,
+      "loss": 0.4382,
+      "step": 44620
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.0146927946535015e-06,
+      "loss": 0.3558,
+      "step": 44621
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.014356959347136e-06,
+      "loss": 0.2681,
+      "step": 44622
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.014021124040771e-06,
+      "loss": 0.4522,
+      "step": 44623
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.013685288734405e-06,
+      "loss": 0.1931,
+      "step": 44624
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.013349453428039e-06,
+      "loss": 0.2179,
+      "step": 44625
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.013013618121674e-06,
+      "loss": 0.2011,
+      "step": 44626
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.012677782815307e-06,
+      "loss": 0.2818,
+      "step": 44627
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.0123419475089415e-06,
+      "loss": 0.2272,
+      "step": 44628
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.012006112202576e-06,
+      "loss": 0.344,
+      "step": 44629
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.01167027689621e-06,
+      "loss": 0.408,
+      "step": 44630
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.011334441589845e-06,
+      "loss": 0.1531,
+      "step": 44631
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.010998606283479e-06,
+      "loss": 0.2771,
+      "step": 44632
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.010662770977113e-06,
+      "loss": 0.3615,
+      "step": 44633
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.010326935670748e-06,
+      "loss": 0.1991,
+      "step": 44634
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.009991100364382e-06,
+      "loss": 0.2011,
+      "step": 44635
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.009655265058015e-06,
+      "loss": 0.3691,
+      "step": 44636
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.00931942975165e-06,
+      "loss": 0.4003,
+      "step": 44637
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.008983594445285e-06,
+      "loss": 0.3821,
+      "step": 44638
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.008647759138918e-06,
+      "loss": 0.3515,
+      "step": 44639
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.008311923832553e-06,
+      "loss": 0.3843,
+      "step": 44640
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.007976088526188e-06,
+      "loss": 0.2869,
+      "step": 44641
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.007640253219821e-06,
+      "loss": 0.2742,
+      "step": 44642
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.007304417913455e-06,
+      "loss": 0.2681,
+      "step": 44643
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.00696858260709e-06,
+      "loss": 0.1883,
+      "step": 44644
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.006632747300724e-06,
+      "loss": 0.2411,
+      "step": 44645
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.006296911994358e-06,
+      "loss": 0.1819,
+      "step": 44646
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.005961076687993e-06,
+      "loss": 0.3023,
+      "step": 44647
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.005625241381627e-06,
+      "loss": 0.1568,
+      "step": 44648
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.005289406075261e-06,
+      "loss": 0.2015,
+      "step": 44649
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.004953570768895e-06,
+      "loss": 0.1855,
+      "step": 44650
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.004617735462529e-06,
+      "loss": 0.3388,
+      "step": 44651
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.004281900156164e-06,
+      "loss": 0.4157,
+      "step": 44652
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.003946064849798e-06,
+      "loss": 0.2496,
+      "step": 44653
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.003610229543432e-06,
+      "loss": 0.1577,
+      "step": 44654
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.003274394237067e-06,
+      "loss": 0.3616,
+      "step": 44655
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.0029385589307015e-06,
+      "loss": 0.283,
+      "step": 44656
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.0026027236243344e-06,
+      "loss": 0.1837,
+      "step": 44657
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.002266888317969e-06,
+      "loss": 0.2402,
+      "step": 44658
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.001931053011604e-06,
+      "loss": 0.3444,
+      "step": 44659
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.0015952177052375e-06,
+      "loss": 0.3078,
+      "step": 44660
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.001259382398872e-06,
+      "loss": 0.16,
+      "step": 44661
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.000923547092507e-06,
+      "loss": 0.3886,
+      "step": 44662
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.00058771178614e-06,
+      "loss": 0.2166,
+      "step": 44663
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 5.000251876479774e-06,
+      "loss": 0.1488,
+      "step": 44664
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.999916041173409e-06,
+      "loss": 0.0893,
+      "step": 44665
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.999580205867044e-06,
+      "loss": 0.1674,
+      "step": 44666
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9992443705606775e-06,
+      "loss": 0.3784,
+      "step": 44667
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.998908535254311e-06,
+      "loss": 0.1238,
+      "step": 44668
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.998572699947946e-06,
+      "loss": 0.2914,
+      "step": 44669
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.998236864641581e-06,
+      "loss": 0.2052,
+      "step": 44670
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.997901029335214e-06,
+      "loss": 0.3772,
+      "step": 44671
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.997565194028849e-06,
+      "loss": 0.3077,
+      "step": 44672
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.997229358722483e-06,
+      "loss": 0.1643,
+      "step": 44673
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.996893523416117e-06,
+      "loss": 0.2007,
+      "step": 44674
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.996557688109751e-06,
+      "loss": 0.2303,
+      "step": 44675
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.996221852803386e-06,
+      "loss": 0.2694,
+      "step": 44676
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.99588601749702e-06,
+      "loss": 0.3053,
+      "step": 44677
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.995550182190654e-06,
+      "loss": 0.2202,
+      "step": 44678
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.995214346884288e-06,
+      "loss": 0.3021,
+      "step": 44679
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.994878511577923e-06,
+      "loss": 0.2607,
+      "step": 44680
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9945426762715575e-06,
+      "loss": 0.2961,
+      "step": 44681
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.994206840965191e-06,
+      "loss": 0.3308,
+      "step": 44682
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.993871005658825e-06,
+      "loss": 0.412,
+      "step": 44683
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.99353517035246e-06,
+      "loss": 0.359,
+      "step": 44684
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9931993350460935e-06,
+      "loss": 0.2213,
+      "step": 44685
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.992863499739728e-06,
+      "loss": 0.2187,
+      "step": 44686
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.992527664433363e-06,
+      "loss": 0.2272,
+      "step": 44687
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.992191829126997e-06,
+      "loss": 0.3393,
+      "step": 44688
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9918559938206304e-06,
+      "loss": 0.3467,
+      "step": 44689
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.991520158514265e-06,
+      "loss": 0.3238,
+      "step": 44690
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9911843232079e-06,
+      "loss": 0.2568,
+      "step": 44691
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9908484879015335e-06,
+      "loss": 0.0971,
+      "step": 44692
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.990512652595167e-06,
+      "loss": 0.2394,
+      "step": 44693
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.990176817288802e-06,
+      "loss": 0.1471,
+      "step": 44694
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.989840981982436e-06,
+      "loss": 0.1886,
+      "step": 44695
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9895051466760704e-06,
+      "loss": 0.2086,
+      "step": 44696
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.989169311369705e-06,
+      "loss": 0.2469,
+      "step": 44697
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.988833476063339e-06,
+      "loss": 0.4075,
+      "step": 44698
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.988497640756973e-06,
+      "loss": 0.3121,
+      "step": 44699
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.988161805450607e-06,
+      "loss": 0.2383,
+      "step": 44700
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.987825970144242e-06,
+      "loss": 0.3488,
+      "step": 44701
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.987490134837876e-06,
+      "loss": 0.2776,
+      "step": 44702
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9871542995315104e-06,
+      "loss": 0.2938,
+      "step": 44703
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.986818464225144e-06,
+      "loss": 0.2789,
+      "step": 44704
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.986482628918778e-06,
+      "loss": 0.3293,
+      "step": 44705
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.986146793612413e-06,
+      "loss": 0.1832,
+      "step": 44706
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.985810958306047e-06,
+      "loss": 0.1737,
+      "step": 44707
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.985475122999681e-06,
+      "loss": 0.2634,
+      "step": 44708
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.985139287693316e-06,
+      "loss": 0.2844,
+      "step": 44709
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9848034523869496e-06,
+      "loss": 0.2457,
+      "step": 44710
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.984467617080584e-06,
+      "loss": 0.3619,
+      "step": 44711
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.984131781774219e-06,
+      "loss": 0.3617,
+      "step": 44712
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.983795946467853e-06,
+      "loss": 0.177,
+      "step": 44713
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9834601111614865e-06,
+      "loss": 0.289,
+      "step": 44714
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.983124275855121e-06,
+      "loss": 0.3716,
+      "step": 44715
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.982788440548755e-06,
+      "loss": 0.4305,
+      "step": 44716
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9824526052423896e-06,
+      "loss": 0.3258,
+      "step": 44717
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.982116769936024e-06,
+      "loss": 0.2937,
+      "step": 44718
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.981780934629658e-06,
+      "loss": 0.1479,
+      "step": 44719
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.981445099323292e-06,
+      "loss": 0.3038,
+      "step": 44720
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9811092640169265e-06,
+      "loss": 0.2545,
+      "step": 44721
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.980773428710561e-06,
+      "loss": 0.2254,
+      "step": 44722
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.980437593404195e-06,
+      "loss": 0.3109,
+      "step": 44723
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9801017580978295e-06,
+      "loss": 0.0773,
+      "step": 44724
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.979765922791463e-06,
+      "loss": 0.3325,
+      "step": 44725
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.979430087485098e-06,
+      "loss": 0.3312,
+      "step": 44726
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.979094252178732e-06,
+      "loss": 0.4275,
+      "step": 44727
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9787584168723664e-06,
+      "loss": 0.3786,
+      "step": 44728
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.978422581566e-06,
+      "loss": 0.2619,
+      "step": 44729
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.978086746259635e-06,
+      "loss": 0.3477,
+      "step": 44730
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.977750910953269e-06,
+      "loss": 0.2587,
+      "step": 44731
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.977415075646903e-06,
+      "loss": 0.336,
+      "step": 44732
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.977079240340537e-06,
+      "loss": 0.2709,
+      "step": 44733
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.976743405034172e-06,
+      "loss": 0.3122,
+      "step": 44734
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.976407569727806e-06,
+      "loss": 0.2435,
+      "step": 44735
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.97607173442144e-06,
+      "loss": 0.2586,
+      "step": 44736
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.975735899115075e-06,
+      "loss": 0.2012,
+      "step": 44737
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.975400063808709e-06,
+      "loss": 0.2688,
+      "step": 44738
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9750642285023425e-06,
+      "loss": 0.3086,
+      "step": 44739
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.974728393195977e-06,
+      "loss": 0.2345,
+      "step": 44740
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.974392557889611e-06,
+      "loss": 0.3044,
+      "step": 44741
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.974056722583246e-06,
+      "loss": 0.3808,
+      "step": 44742
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.97372088727688e-06,
+      "loss": 0.1977,
+      "step": 44743
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.973385051970514e-06,
+      "loss": 0.2479,
+      "step": 44744
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.973049216664148e-06,
+      "loss": 0.2552,
+      "step": 44745
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9727133813577825e-06,
+      "loss": 0.3614,
+      "step": 44746
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.972377546051417e-06,
+      "loss": 0.2226,
+      "step": 44747
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.972041710745051e-06,
+      "loss": 0.2265,
+      "step": 44748
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9717058754386856e-06,
+      "loss": 0.3143,
+      "step": 44749
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.971370040132319e-06,
+      "loss": 0.2819,
+      "step": 44750
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.971034204825953e-06,
+      "loss": 0.1601,
+      "step": 44751
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.970698369519588e-06,
+      "loss": 0.2216,
+      "step": 44752
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.9703625342132225e-06,
+      "loss": 0.2378,
+      "step": 44753
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.970026698906856e-06,
+      "loss": 0.2882,
+      "step": 44754
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.969690863600491e-06,
+      "loss": 0.1621,
+      "step": 44755
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.969355028294125e-06,
+      "loss": 0.3104,
+      "step": 44756
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.969019192987759e-06,
+      "loss": 0.1308,
+      "step": 44757
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.968683357681394e-06,
+      "loss": 0.4439,
+      "step": 44758
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.968347522375028e-06,
+      "loss": 0.2839,
+      "step": 44759
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.968011687068662e-06,
+      "loss": 0.2258,
+      "step": 44760
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.967675851762296e-06,
+      "loss": 0.1322,
+      "step": 44761
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.96734001645593e-06,
+      "loss": 0.2557,
+      "step": 44762
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.967004181149565e-06,
+      "loss": 0.3801,
+      "step": 44763
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 4.966668345843199e-06,
+      "loss": 0.3719,
+      "step": 44764
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.966332510536833e-06,
+      "loss": 0.2731,
+      "step": 44765
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.965996675230467e-06,
+      "loss": 0.2324,
+      "step": 44766
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.965660839924102e-06,
+      "loss": 0.2579,
+      "step": 44767
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.965325004617736e-06,
+      "loss": 0.3551,
+      "step": 44768
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.96498916931137e-06,
+      "loss": 0.3128,
+      "step": 44769
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.964653334005005e-06,
+      "loss": 0.3742,
+      "step": 44770
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9643174986986385e-06,
+      "loss": 0.3594,
+      "step": 44771
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.963981663392273e-06,
+      "loss": 0.3348,
+      "step": 44772
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.963645828085907e-06,
+      "loss": 0.2412,
+      "step": 44773
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.963309992779542e-06,
+      "loss": 0.351,
+      "step": 44774
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.962974157473175e-06,
+      "loss": 0.2248,
+      "step": 44775
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.96263832216681e-06,
+      "loss": 0.1968,
+      "step": 44776
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.962302486860444e-06,
+      "loss": 0.1985,
+      "step": 44777
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9619666515540785e-06,
+      "loss": 0.2375,
+      "step": 44778
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.961630816247712e-06,
+      "loss": 0.3994,
+      "step": 44779
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.961294980941347e-06,
+      "loss": 0.251,
+      "step": 44780
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.960959145634981e-06,
+      "loss": 0.1303,
+      "step": 44781
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.960623310328615e-06,
+      "loss": 0.1808,
+      "step": 44782
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.96028747502225e-06,
+      "loss": 0.3043,
+      "step": 44783
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.959951639715884e-06,
+      "loss": 0.2748,
+      "step": 44784
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.959615804409518e-06,
+      "loss": 0.2636,
+      "step": 44785
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.959279969103152e-06,
+      "loss": 0.2583,
+      "step": 44786
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.958944133796786e-06,
+      "loss": 0.2103,
+      "step": 44787
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.958608298490421e-06,
+      "loss": 0.2716,
+      "step": 44788
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.958272463184055e-06,
+      "loss": 0.3277,
+      "step": 44789
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.957936627877689e-06,
+      "loss": 0.3412,
+      "step": 44790
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.957600792571323e-06,
+      "loss": 0.2731,
+      "step": 44791
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.957264957264958e-06,
+      "loss": 0.2264,
+      "step": 44792
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.956929121958592e-06,
+      "loss": 0.2364,
+      "step": 44793
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.956593286652226e-06,
+      "loss": 0.1945,
+      "step": 44794
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.956257451345861e-06,
+      "loss": 0.1808,
+      "step": 44795
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9559216160394945e-06,
+      "loss": 0.2693,
+      "step": 44796
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.955585780733128e-06,
+      "loss": 0.3851,
+      "step": 44797
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.955249945426763e-06,
+      "loss": 0.3783,
+      "step": 44798
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.954914110120398e-06,
+      "loss": 0.396,
+      "step": 44799
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.954578274814031e-06,
+      "loss": 0.1778,
+      "step": 44800
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.954242439507666e-06,
+      "loss": 0.343,
+      "step": 44801
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9539066042013e-06,
+      "loss": 0.2532,
+      "step": 44802
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9535707688949345e-06,
+      "loss": 0.3161,
+      "step": 44803
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.953234933588569e-06,
+      "loss": 0.267,
+      "step": 44804
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.952899098282203e-06,
+      "loss": 0.3381,
+      "step": 44805
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.952563262975837e-06,
+      "loss": 0.1278,
+      "step": 44806
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.952227427669471e-06,
+      "loss": 0.2875,
+      "step": 44807
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.951891592363105e-06,
+      "loss": 0.3047,
+      "step": 44808
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.95155575705674e-06,
+      "loss": 0.1631,
+      "step": 44809
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9512199217503745e-06,
+      "loss": 0.3169,
+      "step": 44810
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.950884086444008e-06,
+      "loss": 0.1408,
+      "step": 44811
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.950548251137642e-06,
+      "loss": 0.1543,
+      "step": 44812
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.950212415831277e-06,
+      "loss": 0.2362,
+      "step": 44813
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.949876580524911e-06,
+      "loss": 0.3808,
+      "step": 44814
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.949540745218545e-06,
+      "loss": 0.2819,
+      "step": 44815
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.94920490991218e-06,
+      "loss": 0.3063,
+      "step": 44816
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.948869074605814e-06,
+      "loss": 0.2899,
+      "step": 44817
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9485332392994474e-06,
+      "loss": 0.237,
+      "step": 44818
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.948197403993082e-06,
+      "loss": 0.2164,
+      "step": 44819
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.947861568686717e-06,
+      "loss": 0.1912,
+      "step": 44820
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9475257333803505e-06,
+      "loss": 0.3763,
+      "step": 44821
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.947189898073985e-06,
+      "loss": 0.2534,
+      "step": 44822
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.946854062767619e-06,
+      "loss": 0.2524,
+      "step": 44823
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.946518227461254e-06,
+      "loss": 0.3119,
+      "step": 44824
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9461823921548874e-06,
+      "loss": 0.1027,
+      "step": 44825
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.945846556848522e-06,
+      "loss": 0.2337,
+      "step": 44826
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.945510721542156e-06,
+      "loss": 0.2926,
+      "step": 44827
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9451748862357905e-06,
+      "loss": 0.3313,
+      "step": 44828
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.944839050929424e-06,
+      "loss": 0.5055,
+      "step": 44829
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.944503215623059e-06,
+      "loss": 0.2171,
+      "step": 44830
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.944167380316693e-06,
+      "loss": 0.331,
+      "step": 44831
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.943831545010327e-06,
+      "loss": 0.2236,
+      "step": 44832
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.943495709703961e-06,
+      "loss": 0.2916,
+      "step": 44833
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.943159874397596e-06,
+      "loss": 0.2485,
+      "step": 44834
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9428240390912305e-06,
+      "loss": 0.2765,
+      "step": 44835
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.942488203784864e-06,
+      "loss": 0.4122,
+      "step": 44836
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.942152368478498e-06,
+      "loss": 0.12,
+      "step": 44837
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.941816533172133e-06,
+      "loss": 0.153,
+      "step": 44838
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.941480697865767e-06,
+      "loss": 0.119,
+      "step": 44839
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.941144862559401e-06,
+      "loss": 0.3277,
+      "step": 44840
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.940809027253036e-06,
+      "loss": 0.2995,
+      "step": 44841
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.94047319194667e-06,
+      "loss": 0.189,
+      "step": 44842
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9401373566403035e-06,
+      "loss": 0.1989,
+      "step": 44843
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.939801521333938e-06,
+      "loss": 0.2632,
+      "step": 44844
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.939465686027573e-06,
+      "loss": 0.4969,
+      "step": 44845
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9391298507212066e-06,
+      "loss": 0.3091,
+      "step": 44846
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.938794015414841e-06,
+      "loss": 0.2698,
+      "step": 44847
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.938458180108475e-06,
+      "loss": 0.3716,
+      "step": 44848
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.93812234480211e-06,
+      "loss": 0.1479,
+      "step": 44849
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.937786509495744e-06,
+      "loss": 0.1211,
+      "step": 44850
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.937450674189378e-06,
+      "loss": 0.2041,
+      "step": 44851
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.937114838883012e-06,
+      "loss": 0.3661,
+      "step": 44852
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9367790035766465e-06,
+      "loss": 0.2082,
+      "step": 44853
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.93644316827028e-06,
+      "loss": 0.2035,
+      "step": 44854
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.936107332963915e-06,
+      "loss": 0.3393,
+      "step": 44855
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.93577149765755e-06,
+      "loss": 0.2805,
+      "step": 44856
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9354356623511834e-06,
+      "loss": 0.207,
+      "step": 44857
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.935099827044817e-06,
+      "loss": 0.2009,
+      "step": 44858
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.934763991738452e-06,
+      "loss": 0.219,
+      "step": 44859
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9344281564320865e-06,
+      "loss": 0.209,
+      "step": 44860
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.93409232112572e-06,
+      "loss": 0.3293,
+      "step": 44861
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.933756485819355e-06,
+      "loss": 0.4273,
+      "step": 44862
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.933420650512989e-06,
+      "loss": 0.3541,
+      "step": 44863
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.933084815206623e-06,
+      "loss": 0.1978,
+      "step": 44864
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.932748979900257e-06,
+      "loss": 0.2709,
+      "step": 44865
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.932413144593892e-06,
+      "loss": 0.2802,
+      "step": 44866
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.932077309287526e-06,
+      "loss": 0.2166,
+      "step": 44867
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9317414739811595e-06,
+      "loss": 0.6029,
+      "step": 44868
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.931405638674794e-06,
+      "loss": 0.1481,
+      "step": 44869
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.931069803368429e-06,
+      "loss": 0.2491,
+      "step": 44870
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9307339680620626e-06,
+      "loss": 0.3145,
+      "step": 44871
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.930398132755697e-06,
+      "loss": 0.1579,
+      "step": 44872
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.930062297449331e-06,
+      "loss": 0.2032,
+      "step": 44873
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.929726462142965e-06,
+      "loss": 0.1361,
+      "step": 44874
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9293906268365995e-06,
+      "loss": 0.124,
+      "step": 44875
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.929054791530234e-06,
+      "loss": 0.3332,
+      "step": 44876
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.928718956223868e-06,
+      "loss": 0.2785,
+      "step": 44877
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9283831209175026e-06,
+      "loss": 0.3698,
+      "step": 44878
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.928047285611136e-06,
+      "loss": 0.3751,
+      "step": 44879
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.927711450304771e-06,
+      "loss": 0.2361,
+      "step": 44880
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.927375614998406e-06,
+      "loss": 0.3081,
+      "step": 44881
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9270397796920395e-06,
+      "loss": 0.2171,
+      "step": 44882
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.926703944385673e-06,
+      "loss": 0.2152,
+      "step": 44883
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.926368109079308e-06,
+      "loss": 0.2054,
+      "step": 44884
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.926032273772942e-06,
+      "loss": 0.321,
+      "step": 44885
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.925696438466576e-06,
+      "loss": 0.247,
+      "step": 44886
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.925360603160211e-06,
+      "loss": 0.2213,
+      "step": 44887
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.925024767853845e-06,
+      "loss": 0.3478,
+      "step": 44888
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.924688932547479e-06,
+      "loss": 0.2377,
+      "step": 44889
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.924353097241113e-06,
+      "loss": 0.2533,
+      "step": 44890
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.924017261934748e-06,
+      "loss": 0.3115,
+      "step": 44891
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.923681426628382e-06,
+      "loss": 0.3364,
+      "step": 44892
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.923345591322016e-06,
+      "loss": 0.3368,
+      "step": 44893
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.92300975601565e-06,
+      "loss": 0.2052,
+      "step": 44894
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.922673920709285e-06,
+      "loss": 0.2273,
+      "step": 44895
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9223380854029194e-06,
+      "loss": 0.4478,
+      "step": 44896
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.922002250096553e-06,
+      "loss": 0.3318,
+      "step": 44897
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.921666414790187e-06,
+      "loss": 0.3288,
+      "step": 44898
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.921330579483822e-06,
+      "loss": 0.1215,
+      "step": 44899
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9209947441774555e-06,
+      "loss": 0.2786,
+      "step": 44900
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.92065890887109e-06,
+      "loss": 0.1785,
+      "step": 44901
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.920323073564725e-06,
+      "loss": 0.4598,
+      "step": 44902
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.919987238258359e-06,
+      "loss": 0.4137,
+      "step": 44903
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.919651402951992e-06,
+      "loss": 0.2954,
+      "step": 44904
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.919315567645627e-06,
+      "loss": 0.2367,
+      "step": 44905
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.918979732339262e-06,
+      "loss": 0.3768,
+      "step": 44906
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9186438970328955e-06,
+      "loss": 0.3826,
+      "step": 44907
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.91830806172653e-06,
+      "loss": 0.2191,
+      "step": 44908
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.917972226420164e-06,
+      "loss": 0.2896,
+      "step": 44909
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.917636391113798e-06,
+      "loss": 0.2838,
+      "step": 44910
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.917300555807432e-06,
+      "loss": 0.2285,
+      "step": 44911
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.916964720501067e-06,
+      "loss": 0.4656,
+      "step": 44912
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.916628885194701e-06,
+      "loss": 0.22,
+      "step": 44913
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.916293049888335e-06,
+      "loss": 0.2338,
+      "step": 44914
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.915957214581969e-06,
+      "loss": 0.3857,
+      "step": 44915
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.915621379275604e-06,
+      "loss": 0.3649,
+      "step": 44916
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.915285543969238e-06,
+      "loss": 0.3424,
+      "step": 44917
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.914949708662872e-06,
+      "loss": 0.2308,
+      "step": 44918
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.914613873356506e-06,
+      "loss": 0.3033,
+      "step": 44919
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.91427803805014e-06,
+      "loss": 0.2758,
+      "step": 44920
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.913942202743775e-06,
+      "loss": 0.2167,
+      "step": 44921
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.913606367437409e-06,
+      "loss": 0.3459,
+      "step": 44922
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.913270532131043e-06,
+      "loss": 0.2082,
+      "step": 44923
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.912934696824678e-06,
+      "loss": 0.263,
+      "step": 44924
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9125988615183115e-06,
+      "loss": 0.4598,
+      "step": 44925
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.912263026211946e-06,
+      "loss": 0.3114,
+      "step": 44926
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.911927190905581e-06,
+      "loss": 0.3863,
+      "step": 44927
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.911591355599215e-06,
+      "loss": 0.3698,
+      "step": 44928
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.911255520292848e-06,
+      "loss": 0.284,
+      "step": 44929
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.910919684986483e-06,
+      "loss": 0.3467,
+      "step": 44930
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.910583849680117e-06,
+      "loss": 0.2309,
+      "step": 44931
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9102480143737515e-06,
+      "loss": 0.181,
+      "step": 44932
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.909912179067386e-06,
+      "loss": 0.2338,
+      "step": 44933
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.90957634376102e-06,
+      "loss": 0.2951,
+      "step": 44934
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.909240508454654e-06,
+      "loss": 0.2652,
+      "step": 44935
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.908904673148288e-06,
+      "loss": 0.2486,
+      "step": 44936
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.908568837841923e-06,
+      "loss": 0.2724,
+      "step": 44937
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.908233002535557e-06,
+      "loss": 0.3087,
+      "step": 44938
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9078971672291915e-06,
+      "loss": 0.2794,
+      "step": 44939
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.907561331922825e-06,
+      "loss": 0.2116,
+      "step": 44940
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.90722549661646e-06,
+      "loss": 0.2023,
+      "step": 44941
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.906889661310094e-06,
+      "loss": 0.436,
+      "step": 44942
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.906553826003728e-06,
+      "loss": 0.1818,
+      "step": 44943
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.906217990697362e-06,
+      "loss": 0.2542,
+      "step": 44944
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.905882155390997e-06,
+      "loss": 0.2791,
+      "step": 44945
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.905546320084631e-06,
+      "loss": 0.2978,
+      "step": 44946
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.905210484778265e-06,
+      "loss": 0.2166,
+      "step": 44947
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9048746494719e-06,
+      "loss": 0.2378,
+      "step": 44948
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.904538814165534e-06,
+      "loss": 0.1389,
+      "step": 44949
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9042029788591675e-06,
+      "loss": 0.2873,
+      "step": 44950
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.903867143552802e-06,
+      "loss": 0.2576,
+      "step": 44951
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.903531308246437e-06,
+      "loss": 0.31,
+      "step": 44952
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.903195472940071e-06,
+      "loss": 0.3761,
+      "step": 44953
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.902859637633705e-06,
+      "loss": 0.278,
+      "step": 44954
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.902523802327339e-06,
+      "loss": 0.4824,
+      "step": 44955
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.902187967020973e-06,
+      "loss": 0.3138,
+      "step": 44956
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.9018521317146075e-06,
+      "loss": 0.1829,
+      "step": 44957
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.901516296408242e-06,
+      "loss": 0.2392,
+      "step": 44958
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.901180461101876e-06,
+      "loss": 0.2868,
+      "step": 44959
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.90084462579551e-06,
+      "loss": 0.3257,
+      "step": 44960
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.900508790489144e-06,
+      "loss": 0.4286,
+      "step": 44961
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 4.900172955182779e-06,
+      "loss": 0.3124,
+      "step": 44962
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.899837119876413e-06,
+      "loss": 0.286,
+      "step": 44963
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8995012845700475e-06,
+      "loss": 0.3617,
+      "step": 44964
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.899165449263681e-06,
+      "loss": 0.3959,
+      "step": 44965
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.898829613957315e-06,
+      "loss": 0.2623,
+      "step": 44966
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.89849377865095e-06,
+      "loss": 0.3522,
+      "step": 44967
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.898157943344584e-06,
+      "loss": 0.2959,
+      "step": 44968
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.897822108038218e-06,
+      "loss": 0.3117,
+      "step": 44969
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.897486272731853e-06,
+      "loss": 0.1279,
+      "step": 44970
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.897150437425487e-06,
+      "loss": 0.3244,
+      "step": 44971
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.896814602119121e-06,
+      "loss": 0.2691,
+      "step": 44972
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.896478766812756e-06,
+      "loss": 0.2578,
+      "step": 44973
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.89614293150639e-06,
+      "loss": 0.2913,
+      "step": 44974
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8958070962000236e-06,
+      "loss": 0.2951,
+      "step": 44975
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.895471260893658e-06,
+      "loss": 0.3203,
+      "step": 44976
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.895135425587292e-06,
+      "loss": 0.2893,
+      "step": 44977
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.894799590280927e-06,
+      "loss": 0.2152,
+      "step": 44978
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.894463754974561e-06,
+      "loss": 0.2831,
+      "step": 44979
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.894127919668195e-06,
+      "loss": 0.3415,
+      "step": 44980
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.893792084361829e-06,
+      "loss": 0.2793,
+      "step": 44981
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8934562490554635e-06,
+      "loss": 0.2412,
+      "step": 44982
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.893120413749098e-06,
+      "loss": 0.2069,
+      "step": 44983
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.892784578442732e-06,
+      "loss": 0.3301,
+      "step": 44984
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.892448743136367e-06,
+      "loss": 0.3059,
+      "step": 44985
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8921129078300004e-06,
+      "loss": 0.2294,
+      "step": 44986
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.891777072523634e-06,
+      "loss": 0.2252,
+      "step": 44987
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.891441237217269e-06,
+      "loss": 0.2818,
+      "step": 44988
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8911054019109035e-06,
+      "loss": 0.2952,
+      "step": 44989
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.890769566604537e-06,
+      "loss": 0.2428,
+      "step": 44990
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.890433731298172e-06,
+      "loss": 0.0943,
+      "step": 44991
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.890097895991806e-06,
+      "loss": 0.3045,
+      "step": 44992
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8897620606854404e-06,
+      "loss": 0.2528,
+      "step": 44993
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.889426225379075e-06,
+      "loss": 0.1866,
+      "step": 44994
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.889090390072709e-06,
+      "loss": 0.1616,
+      "step": 44995
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.888754554766343e-06,
+      "loss": 0.1787,
+      "step": 44996
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.888418719459977e-06,
+      "loss": 0.1515,
+      "step": 44997
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.888082884153611e-06,
+      "loss": 0.2346,
+      "step": 44998
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.887747048847246e-06,
+      "loss": 0.1812,
+      "step": 44999
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.88741121354088e-06,
+      "loss": 0.2143,
+      "step": 45000
+    },
+    {
+      "epoch": 2.27,
+      "eval_accuracy": 0.8846487264819753,
+      "eval_f1_microaverage": 0.8846487264819753,
+      "eval_loss": 0.30961233377456665,
+      "eval_precision_macroaverage": 0.7648626263586411,
+      "eval_precision_microaverage": 0.8846487264819753,
+      "eval_recall_macroaverage": 0.6863613299189268,
+      "eval_recall_microaverage": 0.8846487264819753,
+      "eval_runtime": 1173.9643,
+      "eval_samples_per_second": 231.896,
+      "eval_steps_per_second": 3.624,
+      "step": 45000
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.887075378234514e-06,
+      "loss": 0.218,
+      "step": 45001
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.886739542928148e-06,
+      "loss": 0.147,
+      "step": 45002
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.886403707621783e-06,
+      "loss": 0.1629,
+      "step": 45003
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.886067872315417e-06,
+      "loss": 0.3834,
+      "step": 45004
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.885732037009051e-06,
+      "loss": 0.229,
+      "step": 45005
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.885396201702685e-06,
+      "loss": 0.2122,
+      "step": 45006
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8850603663963196e-06,
+      "loss": 0.3884,
+      "step": 45007
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.884724531089954e-06,
+      "loss": 0.2788,
+      "step": 45008
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.884388695783588e-06,
+      "loss": 0.2882,
+      "step": 45009
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.884052860477223e-06,
+      "loss": 0.4881,
+      "step": 45010
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8837170251708565e-06,
+      "loss": 0.3541,
+      "step": 45011
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.88338118986449e-06,
+      "loss": 0.3514,
+      "step": 45012
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.883045354558125e-06,
+      "loss": 0.2543,
+      "step": 45013
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8827095192517596e-06,
+      "loss": 0.2881,
+      "step": 45014
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.882373683945393e-06,
+      "loss": 0.3494,
+      "step": 45015
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.882037848639028e-06,
+      "loss": 0.287,
+      "step": 45016
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.881702013332662e-06,
+      "loss": 0.2513,
+      "step": 45017
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8813661780262964e-06,
+      "loss": 0.18,
+      "step": 45018
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.881030342719931e-06,
+      "loss": 0.2204,
+      "step": 45019
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.880694507413565e-06,
+      "loss": 0.349,
+      "step": 45020
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.880358672107199e-06,
+      "loss": 0.1902,
+      "step": 45021
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.880022836800833e-06,
+      "loss": 0.1643,
+      "step": 45022
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.879687001494467e-06,
+      "loss": 0.3906,
+      "step": 45023
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.879351166188102e-06,
+      "loss": 0.35,
+      "step": 45024
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8790153308817364e-06,
+      "loss": 0.2338,
+      "step": 45025
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.87867949557537e-06,
+      "loss": 0.3281,
+      "step": 45026
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.878343660269004e-06,
+      "loss": 0.3694,
+      "step": 45027
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.878007824962639e-06,
+      "loss": 0.24,
+      "step": 45028
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.877671989656273e-06,
+      "loss": 0.4483,
+      "step": 45029
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.877336154349907e-06,
+      "loss": 0.2316,
+      "step": 45030
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.877000319043542e-06,
+      "loss": 0.2788,
+      "step": 45031
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.876664483737176e-06,
+      "loss": 0.473,
+      "step": 45032
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.876328648430809e-06,
+      "loss": 0.2374,
+      "step": 45033
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.875992813124444e-06,
+      "loss": 0.3555,
+      "step": 45034
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.875656977818079e-06,
+      "loss": 0.2672,
+      "step": 45035
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8753211425117125e-06,
+      "loss": 0.2381,
+      "step": 45036
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.874985307205347e-06,
+      "loss": 0.1789,
+      "step": 45037
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.874649471898981e-06,
+      "loss": 0.1846,
+      "step": 45038
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8743136365926156e-06,
+      "loss": 0.1982,
+      "step": 45039
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.87397780128625e-06,
+      "loss": 0.2432,
+      "step": 45040
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.873641965979884e-06,
+      "loss": 0.276,
+      "step": 45041
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.873306130673518e-06,
+      "loss": 0.3971,
+      "step": 45042
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8729702953671525e-06,
+      "loss": 0.2442,
+      "step": 45043
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.872634460060786e-06,
+      "loss": 0.295,
+      "step": 45044
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.872298624754421e-06,
+      "loss": 0.1601,
+      "step": 45045
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8719627894480556e-06,
+      "loss": 0.2022,
+      "step": 45046
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.871626954141689e-06,
+      "loss": 0.2092,
+      "step": 45047
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.871291118835323e-06,
+      "loss": 0.2713,
+      "step": 45048
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.870955283528958e-06,
+      "loss": 0.2156,
+      "step": 45049
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8706194482225925e-06,
+      "loss": 0.3349,
+      "step": 45050
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.870283612916226e-06,
+      "loss": 0.2892,
+      "step": 45051
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.86994777760986e-06,
+      "loss": 0.1908,
+      "step": 45052
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.869611942303495e-06,
+      "loss": 0.4872,
+      "step": 45053
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.869276106997129e-06,
+      "loss": 0.3074,
+      "step": 45054
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.868940271690763e-06,
+      "loss": 0.3845,
+      "step": 45055
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.868604436384398e-06,
+      "loss": 0.2412,
+      "step": 45056
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.868268601078032e-06,
+      "loss": 0.3497,
+      "step": 45057
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.867932765771665e-06,
+      "loss": 0.1363,
+      "step": 45058
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8675969304653e-06,
+      "loss": 0.2316,
+      "step": 45059
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.867261095158935e-06,
+      "loss": 0.2127,
+      "step": 45060
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8669252598525685e-06,
+      "loss": 0.1968,
+      "step": 45061
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.866589424546203e-06,
+      "loss": 0.363,
+      "step": 45062
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.866253589239837e-06,
+      "loss": 0.1092,
+      "step": 45063
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.865917753933472e-06,
+      "loss": 0.1907,
+      "step": 45064
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.865581918627106e-06,
+      "loss": 0.2963,
+      "step": 45065
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.86524608332074e-06,
+      "loss": 0.1414,
+      "step": 45066
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.864910248014374e-06,
+      "loss": 0.3522,
+      "step": 45067
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8645744127080085e-06,
+      "loss": 0.3578,
+      "step": 45068
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.864238577401642e-06,
+      "loss": 0.2479,
+      "step": 45069
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.863902742095277e-06,
+      "loss": 0.305,
+      "step": 45070
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.863566906788912e-06,
+      "loss": 0.2556,
+      "step": 45071
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.863231071482545e-06,
+      "loss": 0.206,
+      "step": 45072
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.862895236176179e-06,
+      "loss": 0.4767,
+      "step": 45073
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.862559400869814e-06,
+      "loss": 0.2842,
+      "step": 45074
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8622235655634485e-06,
+      "loss": 0.3551,
+      "step": 45075
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.861887730257082e-06,
+      "loss": 0.2713,
+      "step": 45076
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.861551894950717e-06,
+      "loss": 0.1492,
+      "step": 45077
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.861216059644351e-06,
+      "loss": 0.2278,
+      "step": 45078
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8608802243379845e-06,
+      "loss": 0.2759,
+      "step": 45079
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.860544389031619e-06,
+      "loss": 0.353,
+      "step": 45080
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.860208553725254e-06,
+      "loss": 0.3164,
+      "step": 45081
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.859872718418888e-06,
+      "loss": 0.1626,
+      "step": 45082
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.859536883112522e-06,
+      "loss": 0.2589,
+      "step": 45083
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.859201047806156e-06,
+      "loss": 0.2622,
+      "step": 45084
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.858865212499791e-06,
+      "loss": 0.0875,
+      "step": 45085
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.858529377193425e-06,
+      "loss": 0.4161,
+      "step": 45086
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.858193541887059e-06,
+      "loss": 0.3268,
+      "step": 45087
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.857857706580693e-06,
+      "loss": 0.3181,
+      "step": 45088
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.857521871274328e-06,
+      "loss": 0.3534,
+      "step": 45089
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.857186035967961e-06,
+      "loss": 0.1891,
+      "step": 45090
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.856850200661596e-06,
+      "loss": 0.2673,
+      "step": 45091
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.856514365355231e-06,
+      "loss": 0.2314,
+      "step": 45092
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8561785300488645e-06,
+      "loss": 0.1022,
+      "step": 45093
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.855842694742498e-06,
+      "loss": 0.2305,
+      "step": 45094
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.855506859436133e-06,
+      "loss": 0.1879,
+      "step": 45095
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.855171024129768e-06,
+      "loss": 0.2491,
+      "step": 45096
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.854835188823401e-06,
+      "loss": 0.3023,
+      "step": 45097
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.854499353517035e-06,
+      "loss": 0.175,
+      "step": 45098
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.85416351821067e-06,
+      "loss": 0.3604,
+      "step": 45099
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.853827682904304e-06,
+      "loss": 0.4308,
+      "step": 45100
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.853491847597938e-06,
+      "loss": 0.3511,
+      "step": 45101
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.853156012291573e-06,
+      "loss": 0.2097,
+      "step": 45102
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.852820176985207e-06,
+      "loss": 0.164,
+      "step": 45103
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8524843416788405e-06,
+      "loss": 0.1404,
+      "step": 45104
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.852148506372475e-06,
+      "loss": 0.269,
+      "step": 45105
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.85181267106611e-06,
+      "loss": 0.2602,
+      "step": 45106
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.851476835759744e-06,
+      "loss": 0.2455,
+      "step": 45107
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.851141000453378e-06,
+      "loss": 0.3063,
+      "step": 45108
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.850805165147012e-06,
+      "loss": 0.2647,
+      "step": 45109
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.850469329840647e-06,
+      "loss": 0.3561,
+      "step": 45110
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8501334945342805e-06,
+      "loss": 0.2442,
+      "step": 45111
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.849797659227915e-06,
+      "loss": 0.2193,
+      "step": 45112
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.849461823921549e-06,
+      "loss": 0.1197,
+      "step": 45113
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.849125988615184e-06,
+      "loss": 0.1144,
+      "step": 45114
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8487901533088174e-06,
+      "loss": 0.3499,
+      "step": 45115
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.848454318002452e-06,
+      "loss": 0.3422,
+      "step": 45116
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.848118482696087e-06,
+      "loss": 0.2624,
+      "step": 45117
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8477826473897205e-06,
+      "loss": 0.2487,
+      "step": 45118
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.847446812083354e-06,
+      "loss": 0.2285,
+      "step": 45119
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.847110976776989e-06,
+      "loss": 0.2994,
+      "step": 45120
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.846775141470624e-06,
+      "loss": 0.4271,
+      "step": 45121
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8464393061642574e-06,
+      "loss": 0.2934,
+      "step": 45122
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.846103470857892e-06,
+      "loss": 0.3954,
+      "step": 45123
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.845767635551526e-06,
+      "loss": 0.1566,
+      "step": 45124
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.84543180024516e-06,
+      "loss": 0.3344,
+      "step": 45125
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.845095964938794e-06,
+      "loss": 0.2245,
+      "step": 45126
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.844760129632429e-06,
+      "loss": 0.3375,
+      "step": 45127
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.844424294326063e-06,
+      "loss": 0.2751,
+      "step": 45128
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.844088459019697e-06,
+      "loss": 0.144,
+      "step": 45129
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.843752623713331e-06,
+      "loss": 0.3788,
+      "step": 45130
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.843416788406966e-06,
+      "loss": 0.5109,
+      "step": 45131
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8430809531006005e-06,
+      "loss": 0.2567,
+      "step": 45132
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.842745117794234e-06,
+      "loss": 0.4141,
+      "step": 45133
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.842409282487868e-06,
+      "loss": 0.2595,
+      "step": 45134
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.842073447181503e-06,
+      "loss": 0.282,
+      "step": 45135
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8417376118751366e-06,
+      "loss": 0.2963,
+      "step": 45136
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.841401776568771e-06,
+      "loss": 0.1916,
+      "step": 45137
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.841065941262406e-06,
+      "loss": 0.4352,
+      "step": 45138
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.84073010595604e-06,
+      "loss": 0.19,
+      "step": 45139
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8403942706496735e-06,
+      "loss": 0.2653,
+      "step": 45140
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.840058435343308e-06,
+      "loss": 0.2715,
+      "step": 45141
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.839722600036943e-06,
+      "loss": 0.2537,
+      "step": 45142
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8393867647305766e-06,
+      "loss": 0.2955,
+      "step": 45143
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.83905092942421e-06,
+      "loss": 0.2909,
+      "step": 45144
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.838715094117845e-06,
+      "loss": 0.3252,
+      "step": 45145
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.838379258811479e-06,
+      "loss": 0.2183,
+      "step": 45146
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8380434235051134e-06,
+      "loss": 0.28,
+      "step": 45147
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.837707588198748e-06,
+      "loss": 0.2863,
+      "step": 45148
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.837371752892382e-06,
+      "loss": 0.2475,
+      "step": 45149
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.837035917586016e-06,
+      "loss": 0.245,
+      "step": 45150
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.83670008227965e-06,
+      "loss": 0.2844,
+      "step": 45151
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.836364246973285e-06,
+      "loss": 0.2108,
+      "step": 45152
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.836028411666919e-06,
+      "loss": 0.1888,
+      "step": 45153
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.8356925763605534e-06,
+      "loss": 0.2903,
+      "step": 45154
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.835356741054187e-06,
+      "loss": 0.1369,
+      "step": 45155
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.835020905747821e-06,
+      "loss": 0.2458,
+      "step": 45156
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.834685070441456e-06,
+      "loss": 0.2175,
+      "step": 45157
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.83434923513509e-06,
+      "loss": 0.3313,
+      "step": 45158
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.834013399828724e-06,
+      "loss": 0.2692,
+      "step": 45159
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.833677564522359e-06,
+      "loss": 0.3588,
+      "step": 45160
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 4.833341729215993e-06,
+      "loss": 0.1404,
+      "step": 45161
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.833005893909627e-06,
+      "loss": 0.4339,
+      "step": 45162
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.832670058603262e-06,
+      "loss": 0.2447,
+      "step": 45163
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.832334223296896e-06,
+      "loss": 0.3425,
+      "step": 45164
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.8319983879905295e-06,
+      "loss": 0.2326,
+      "step": 45165
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.831662552684164e-06,
+      "loss": 0.5705,
+      "step": 45166
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.831326717377799e-06,
+      "loss": 0.1422,
+      "step": 45167
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.8309908820714326e-06,
+      "loss": 0.2349,
+      "step": 45168
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.830655046765067e-06,
+      "loss": 0.2041,
+      "step": 45169
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.830319211458701e-06,
+      "loss": 0.2532,
+      "step": 45170
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.829983376152335e-06,
+      "loss": 0.3365,
+      "step": 45171
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.8296475408459695e-06,
+      "loss": 0.2036,
+      "step": 45172
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.829311705539604e-06,
+      "loss": 0.2779,
+      "step": 45173
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.828975870233238e-06,
+      "loss": 0.1812,
+      "step": 45174
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.8286400349268726e-06,
+      "loss": 0.2436,
+      "step": 45175
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.828304199620506e-06,
+      "loss": 0.2334,
+      "step": 45176
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.827968364314141e-06,
+      "loss": 0.3428,
+      "step": 45177
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.827632529007776e-06,
+      "loss": 0.2799,
+      "step": 45178
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.8272966937014095e-06,
+      "loss": 0.2909,
+      "step": 45179
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.826960858395043e-06,
+      "loss": 0.2589,
+      "step": 45180
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.826625023088678e-06,
+      "loss": 0.1805,
+      "step": 45181
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.826289187782312e-06,
+      "loss": 0.2099,
+      "step": 45182
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.825953352475946e-06,
+      "loss": 0.2554,
+      "step": 45183
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.82561751716958e-06,
+      "loss": 0.1962,
+      "step": 45184
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.825281681863215e-06,
+      "loss": 0.3025,
+      "step": 45185
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.824945846556849e-06,
+      "loss": 0.1722,
+      "step": 45186
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.824610011250483e-06,
+      "loss": 0.3691,
+      "step": 45187
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.824274175944118e-06,
+      "loss": 0.1755,
+      "step": 45188
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.823938340637752e-06,
+      "loss": 0.1711,
+      "step": 45189
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.8236025053313855e-06,
+      "loss": 0.2247,
+      "step": 45190
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.82326667002502e-06,
+      "loss": 0.3052,
+      "step": 45191
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.822930834718654e-06,
+      "loss": 0.3928,
+      "step": 45192
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.822594999412289e-06,
+      "loss": 0.1806,
+      "step": 45193
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.822259164105923e-06,
+      "loss": 0.4027,
+      "step": 45194
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.821923328799557e-06,
+      "loss": 0.2747,
+      "step": 45195
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.821587493493191e-06,
+      "loss": 0.3694,
+      "step": 45196
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.8212516581868255e-06,
+      "loss": 0.233,
+      "step": 45197
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.82091582288046e-06,
+      "loss": 0.2713,
+      "step": 45198
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.820579987574094e-06,
+      "loss": 0.2891,
+      "step": 45199
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.820244152267729e-06,
+      "loss": 0.2087,
+      "step": 45200
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.819908316961362e-06,
+      "loss": 0.2663,
+      "step": 45201
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.819572481654996e-06,
+      "loss": 0.243,
+      "step": 45202
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.819236646348631e-06,
+      "loss": 0.1657,
+      "step": 45203
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.8189008110422655e-06,
+      "loss": 0.519,
+      "step": 45204
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.818564975735899e-06,
+      "loss": 0.2765,
+      "step": 45205
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.818229140429534e-06,
+      "loss": 0.192,
+      "step": 45206
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.817893305123168e-06,
+      "loss": 0.1764,
+      "step": 45207
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.817557469816802e-06,
+      "loss": 0.2848,
+      "step": 45208
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.817221634510437e-06,
+      "loss": 0.366,
+      "step": 45209
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.816885799204071e-06,
+      "loss": 0.2161,
+      "step": 45210
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.816549963897705e-06,
+      "loss": 0.3524,
+      "step": 45211
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.816214128591339e-06,
+      "loss": 0.2451,
+      "step": 45212
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.815878293284973e-06,
+      "loss": 0.2438,
+      "step": 45213
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.815542457978608e-06,
+      "loss": 0.1624,
+      "step": 45214
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.815206622672242e-06,
+      "loss": 0.1508,
+      "step": 45215
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.814870787365876e-06,
+      "loss": 0.3193,
+      "step": 45216
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.81453495205951e-06,
+      "loss": 0.2744,
+      "step": 45217
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.814199116753145e-06,
+      "loss": 0.3161,
+      "step": 45218
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.813863281446779e-06,
+      "loss": 0.2897,
+      "step": 45219
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.813527446140413e-06,
+      "loss": 0.2721,
+      "step": 45220
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.813191610834048e-06,
+      "loss": 0.2136,
+      "step": 45221
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.8128557755276815e-06,
+      "loss": 0.2988,
+      "step": 45222
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.812519940221316e-06,
+      "loss": 0.1491,
+      "step": 45223
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.81218410491495e-06,
+      "loss": 0.4687,
+      "step": 45224
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.811848269608585e-06,
+      "loss": 0.3518,
+      "step": 45225
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.811512434302218e-06,
+      "loss": 0.1715,
+      "step": 45226
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.811176598995853e-06,
+      "loss": 0.3332,
+      "step": 45227
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.810840763689487e-06,
+      "loss": 0.324,
+      "step": 45228
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.8105049283831215e-06,
+      "loss": 0.3968,
+      "step": 45229
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.810169093076755e-06,
+      "loss": 0.3612,
+      "step": 45230
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.80983325777039e-06,
+      "loss": 0.246,
+      "step": 45231
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.809497422464024e-06,
+      "loss": 0.1799,
+      "step": 45232
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.809161587157658e-06,
+      "loss": 0.2817,
+      "step": 45233
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.808825751851293e-06,
+      "loss": 0.3501,
+      "step": 45234
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.808489916544927e-06,
+      "loss": 0.3333,
+      "step": 45235
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.808154081238561e-06,
+      "loss": 0.1146,
+      "step": 45236
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.807818245932195e-06,
+      "loss": 0.1758,
+      "step": 45237
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.807482410625829e-06,
+      "loss": 0.2527,
+      "step": 45238
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.807146575319464e-06,
+      "loss": 0.1722,
+      "step": 45239
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.806810740013098e-06,
+      "loss": 0.2084,
+      "step": 45240
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.806474904706732e-06,
+      "loss": 0.3823,
+      "step": 45241
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.806139069400366e-06,
+      "loss": 0.2967,
+      "step": 45242
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.805803234094001e-06,
+      "loss": 0.1699,
+      "step": 45243
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.805467398787635e-06,
+      "loss": 0.3603,
+      "step": 45244
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.805131563481269e-06,
+      "loss": 0.2861,
+      "step": 45245
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.804795728174904e-06,
+      "loss": 0.2438,
+      "step": 45246
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.8044598928685375e-06,
+      "loss": 0.3947,
+      "step": 45247
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.804124057562171e-06,
+      "loss": 0.2189,
+      "step": 45248
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.803788222255806e-06,
+      "loss": 0.2974,
+      "step": 45249
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.803452386949441e-06,
+      "loss": 0.302,
+      "step": 45250
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.803116551643074e-06,
+      "loss": 0.3928,
+      "step": 45251
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.802780716336709e-06,
+      "loss": 0.2393,
+      "step": 45252
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.802444881030343e-06,
+      "loss": 0.3854,
+      "step": 45253
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.8021090457239775e-06,
+      "loss": 0.293,
+      "step": 45254
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.801773210417612e-06,
+      "loss": 0.3188,
+      "step": 45255
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.801437375111246e-06,
+      "loss": 0.2052,
+      "step": 45256
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.80110153980488e-06,
+      "loss": 0.2258,
+      "step": 45257
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.800765704498514e-06,
+      "loss": 0.3227,
+      "step": 45258
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.800429869192148e-06,
+      "loss": 0.1687,
+      "step": 45259
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.800094033885783e-06,
+      "loss": 0.1874,
+      "step": 45260
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7997581985794175e-06,
+      "loss": 0.2663,
+      "step": 45261
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.799422363273051e-06,
+      "loss": 0.2499,
+      "step": 45262
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.799086527966685e-06,
+      "loss": 0.322,
+      "step": 45263
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.79875069266032e-06,
+      "loss": 0.1436,
+      "step": 45264
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.798414857353954e-06,
+      "loss": 0.4258,
+      "step": 45265
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.798079022047588e-06,
+      "loss": 0.3983,
+      "step": 45266
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.797743186741223e-06,
+      "loss": 0.2344,
+      "step": 45267
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.797407351434857e-06,
+      "loss": 0.3411,
+      "step": 45268
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7970715161284905e-06,
+      "loss": 0.4365,
+      "step": 45269
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.796735680822125e-06,
+      "loss": 0.185,
+      "step": 45270
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.79639984551576e-06,
+      "loss": 0.28,
+      "step": 45271
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7960640102093935e-06,
+      "loss": 0.3583,
+      "step": 45272
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.795728174903028e-06,
+      "loss": 0.143,
+      "step": 45273
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.795392339596662e-06,
+      "loss": 0.2531,
+      "step": 45274
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.795056504290297e-06,
+      "loss": 0.2377,
+      "step": 45275
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7947206689839304e-06,
+      "loss": 0.287,
+      "step": 45276
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.794384833677565e-06,
+      "loss": 0.4374,
+      "step": 45277
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.794048998371199e-06,
+      "loss": 0.3351,
+      "step": 45278
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7937131630648335e-06,
+      "loss": 0.2794,
+      "step": 45279
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.793377327758467e-06,
+      "loss": 0.1285,
+      "step": 45280
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.793041492452102e-06,
+      "loss": 0.3434,
+      "step": 45281
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.792705657145736e-06,
+      "loss": 0.4645,
+      "step": 45282
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7923698218393704e-06,
+      "loss": 0.2234,
+      "step": 45283
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.792033986533004e-06,
+      "loss": 0.1709,
+      "step": 45284
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.791698151226639e-06,
+      "loss": 0.2751,
+      "step": 45285
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7913623159202735e-06,
+      "loss": 0.1409,
+      "step": 45286
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.791026480613907e-06,
+      "loss": 0.2056,
+      "step": 45287
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.790690645307541e-06,
+      "loss": 0.2737,
+      "step": 45288
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.790354810001176e-06,
+      "loss": 0.2135,
+      "step": 45289
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7900189746948104e-06,
+      "loss": 0.2399,
+      "step": 45290
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.789683139388444e-06,
+      "loss": 0.2856,
+      "step": 45291
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.789347304082079e-06,
+      "loss": 0.3665,
+      "step": 45292
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.789011468775713e-06,
+      "loss": 0.2294,
+      "step": 45293
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7886756334693465e-06,
+      "loss": 0.3129,
+      "step": 45294
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.788339798162981e-06,
+      "loss": 0.2311,
+      "step": 45295
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.788003962856616e-06,
+      "loss": 0.389,
+      "step": 45296
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7876681275502496e-06,
+      "loss": 0.2805,
+      "step": 45297
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.787332292243884e-06,
+      "loss": 0.2575,
+      "step": 45298
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.786996456937518e-06,
+      "loss": 0.1989,
+      "step": 45299
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.786660621631153e-06,
+      "loss": 0.2725,
+      "step": 45300
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.786324786324787e-06,
+      "loss": 0.2383,
+      "step": 45301
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.785988951018421e-06,
+      "loss": 0.2873,
+      "step": 45302
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.785653115712055e-06,
+      "loss": 0.1755,
+      "step": 45303
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7853172804056896e-06,
+      "loss": 0.3394,
+      "step": 45304
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.784981445099323e-06,
+      "loss": 0.1952,
+      "step": 45305
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.784645609792958e-06,
+      "loss": 0.3609,
+      "step": 45306
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.784309774486593e-06,
+      "loss": 0.2011,
+      "step": 45307
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7839739391802265e-06,
+      "loss": 0.3316,
+      "step": 45308
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.78363810387386e-06,
+      "loss": 0.3813,
+      "step": 45309
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.783302268567495e-06,
+      "loss": 0.1731,
+      "step": 45310
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7829664332611295e-06,
+      "loss": 0.7283,
+      "step": 45311
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.782630597954763e-06,
+      "loss": 0.42,
+      "step": 45312
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.782294762648398e-06,
+      "loss": 0.2511,
+      "step": 45313
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.781958927342032e-06,
+      "loss": 0.1781,
+      "step": 45314
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.781623092035666e-06,
+      "loss": 0.2285,
+      "step": 45315
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7812872567293e-06,
+      "loss": 0.3012,
+      "step": 45316
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.780951421422935e-06,
+      "loss": 0.3419,
+      "step": 45317
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.780615586116569e-06,
+      "loss": 0.2679,
+      "step": 45318
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7802797508102025e-06,
+      "loss": 0.447,
+      "step": 45319
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.779943915503837e-06,
+      "loss": 0.2687,
+      "step": 45320
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.779608080197472e-06,
+      "loss": 0.2644,
+      "step": 45321
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.779272244891106e-06,
+      "loss": 0.3261,
+      "step": 45322
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.77893640958474e-06,
+      "loss": 0.1921,
+      "step": 45323
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.778600574278374e-06,
+      "loss": 0.3475,
+      "step": 45324
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.778264738972009e-06,
+      "loss": 0.1837,
+      "step": 45325
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7779289036656425e-06,
+      "loss": 0.2793,
+      "step": 45326
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.777593068359277e-06,
+      "loss": 0.3518,
+      "step": 45327
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.777257233052911e-06,
+      "loss": 0.1901,
+      "step": 45328
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.776921397746546e-06,
+      "loss": 0.3417,
+      "step": 45329
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.776585562440179e-06,
+      "loss": 0.2065,
+      "step": 45330
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.776249727133814e-06,
+      "loss": 0.2261,
+      "step": 45331
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.775913891827449e-06,
+      "loss": 0.2953,
+      "step": 45332
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7755780565210825e-06,
+      "loss": 0.3236,
+      "step": 45333
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.775242221214716e-06,
+      "loss": 0.251,
+      "step": 45334
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.774906385908351e-06,
+      "loss": 0.332,
+      "step": 45335
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7745705506019856e-06,
+      "loss": 0.2866,
+      "step": 45336
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.774234715295619e-06,
+      "loss": 0.1876,
+      "step": 45337
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.773898879989254e-06,
+      "loss": 0.1403,
+      "step": 45338
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.773563044682888e-06,
+      "loss": 0.3979,
+      "step": 45339
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.773227209376522e-06,
+      "loss": 0.2043,
+      "step": 45340
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.772891374070156e-06,
+      "loss": 0.3698,
+      "step": 45341
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.772555538763791e-06,
+      "loss": 0.2353,
+      "step": 45342
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.772219703457425e-06,
+      "loss": 0.2618,
+      "step": 45343
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.771883868151059e-06,
+      "loss": 0.366,
+      "step": 45344
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.771548032844693e-06,
+      "loss": 0.2502,
+      "step": 45345
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.771212197538328e-06,
+      "loss": 0.2331,
+      "step": 45346
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7708763622319625e-06,
+      "loss": 0.2834,
+      "step": 45347
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.770540526925596e-06,
+      "loss": 0.272,
+      "step": 45348
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.77020469161923e-06,
+      "loss": 0.3799,
+      "step": 45349
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.769868856312865e-06,
+      "loss": 0.1478,
+      "step": 45350
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7695330210064985e-06,
+      "loss": 0.2001,
+      "step": 45351
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.769197185700133e-06,
+      "loss": 0.2362,
+      "step": 45352
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.768861350393768e-06,
+      "loss": 0.3576,
+      "step": 45353
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.768525515087402e-06,
+      "loss": 0.2865,
+      "step": 45354
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.768189679781035e-06,
+      "loss": 0.1829,
+      "step": 45355
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.76785384447467e-06,
+      "loss": 0.325,
+      "step": 45356
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.767518009168305e-06,
+      "loss": 0.2183,
+      "step": 45357
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.7671821738619385e-06,
+      "loss": 0.2364,
+      "step": 45358
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 4.766846338555573e-06,
+      "loss": 0.3689,
+      "step": 45359
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.766510503249207e-06,
+      "loss": 0.2601,
+      "step": 45360
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.766174667942841e-06,
+      "loss": 0.3283,
+      "step": 45361
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.765838832636475e-06,
+      "loss": 0.2477,
+      "step": 45362
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.76550299733011e-06,
+      "loss": 0.3647,
+      "step": 45363
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.765167162023744e-06,
+      "loss": 0.3379,
+      "step": 45364
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.764831326717378e-06,
+      "loss": 0.2017,
+      "step": 45365
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.764495491411012e-06,
+      "loss": 0.2804,
+      "step": 45366
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.764159656104647e-06,
+      "loss": 0.2508,
+      "step": 45367
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.763823820798281e-06,
+      "loss": 0.2087,
+      "step": 45368
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.763487985491915e-06,
+      "loss": 0.2258,
+      "step": 45369
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.763152150185549e-06,
+      "loss": 0.3504,
+      "step": 45370
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.762816314879183e-06,
+      "loss": 0.2049,
+      "step": 45371
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.762480479572818e-06,
+      "loss": 0.1744,
+      "step": 45372
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.762144644266452e-06,
+      "loss": 0.4238,
+      "step": 45373
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.761808808960086e-06,
+      "loss": 0.301,
+      "step": 45374
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.761472973653721e-06,
+      "loss": 0.2016,
+      "step": 45375
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7611371383473545e-06,
+      "loss": 0.2451,
+      "step": 45376
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.760801303040989e-06,
+      "loss": 0.1777,
+      "step": 45377
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.760465467734624e-06,
+      "loss": 0.3564,
+      "step": 45378
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.760129632428258e-06,
+      "loss": 0.5073,
+      "step": 45379
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.759793797121891e-06,
+      "loss": 0.2072,
+      "step": 45380
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.759457961815526e-06,
+      "loss": 0.1737,
+      "step": 45381
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.75912212650916e-06,
+      "loss": 0.6698,
+      "step": 45382
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7587862912027945e-06,
+      "loss": 0.2729,
+      "step": 45383
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.758450455896429e-06,
+      "loss": 0.1279,
+      "step": 45384
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.758114620590063e-06,
+      "loss": 0.2939,
+      "step": 45385
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.757778785283697e-06,
+      "loss": 0.1879,
+      "step": 45386
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.757442949977331e-06,
+      "loss": 0.1583,
+      "step": 45387
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.757107114670966e-06,
+      "loss": 0.164,
+      "step": 45388
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7567712793646e-06,
+      "loss": 0.2339,
+      "step": 45389
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7564354440582345e-06,
+      "loss": 0.4507,
+      "step": 45390
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.756099608751868e-06,
+      "loss": 0.2798,
+      "step": 45391
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.755763773445503e-06,
+      "loss": 0.2317,
+      "step": 45392
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.755427938139137e-06,
+      "loss": 0.4791,
+      "step": 45393
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.755092102832771e-06,
+      "loss": 0.5318,
+      "step": 45394
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.754756267526405e-06,
+      "loss": 0.3021,
+      "step": 45395
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.75442043222004e-06,
+      "loss": 0.2701,
+      "step": 45396
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.754084596913674e-06,
+      "loss": 0.2565,
+      "step": 45397
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.753748761607308e-06,
+      "loss": 0.2169,
+      "step": 45398
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.753412926300943e-06,
+      "loss": 0.3403,
+      "step": 45399
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.753077090994577e-06,
+      "loss": 0.2319,
+      "step": 45400
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7527412556882105e-06,
+      "loss": 0.2151,
+      "step": 45401
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.752405420381845e-06,
+      "loss": 0.1693,
+      "step": 45402
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.75206958507548e-06,
+      "loss": 0.3991,
+      "step": 45403
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.751733749769114e-06,
+      "loss": 0.1666,
+      "step": 45404
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.751397914462748e-06,
+      "loss": 0.3713,
+      "step": 45405
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.751062079156382e-06,
+      "loss": 0.2896,
+      "step": 45406
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.750726243850016e-06,
+      "loss": 0.2709,
+      "step": 45407
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7503904085436505e-06,
+      "loss": 0.4092,
+      "step": 45408
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.750054573237285e-06,
+      "loss": 0.2269,
+      "step": 45409
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.749718737930919e-06,
+      "loss": 0.1603,
+      "step": 45410
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.749382902624553e-06,
+      "loss": 0.374,
+      "step": 45411
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7490470673181874e-06,
+      "loss": 0.2038,
+      "step": 45412
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.748711232011822e-06,
+      "loss": 0.1936,
+      "step": 45413
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.748375396705456e-06,
+      "loss": 0.2816,
+      "step": 45414
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7480395613990905e-06,
+      "loss": 0.2508,
+      "step": 45415
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.747703726092724e-06,
+      "loss": 0.2517,
+      "step": 45416
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.747367890786358e-06,
+      "loss": 0.2816,
+      "step": 45417
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.747032055479993e-06,
+      "loss": 0.2571,
+      "step": 45418
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.746696220173627e-06,
+      "loss": 0.2283,
+      "step": 45419
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.746360384867261e-06,
+      "loss": 0.2267,
+      "step": 45420
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.746024549560896e-06,
+      "loss": 0.3027,
+      "step": 45421
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.74568871425453e-06,
+      "loss": 0.3571,
+      "step": 45422
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.745352878948164e-06,
+      "loss": 0.1246,
+      "step": 45423
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.745017043641799e-06,
+      "loss": 0.4245,
+      "step": 45424
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.744681208335433e-06,
+      "loss": 0.2998,
+      "step": 45425
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7443453730290666e-06,
+      "loss": 0.3197,
+      "step": 45426
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.744009537722701e-06,
+      "loss": 0.3509,
+      "step": 45427
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.743673702416335e-06,
+      "loss": 0.2558,
+      "step": 45428
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.74333786710997e-06,
+      "loss": 0.2232,
+      "step": 45429
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.743002031803604e-06,
+      "loss": 0.2807,
+      "step": 45430
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.742666196497238e-06,
+      "loss": 0.371,
+      "step": 45431
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.742330361190872e-06,
+      "loss": 0.2221,
+      "step": 45432
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7419945258845066e-06,
+      "loss": 0.2135,
+      "step": 45433
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.741658690578141e-06,
+      "loss": 0.2622,
+      "step": 45434
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.741322855271775e-06,
+      "loss": 0.2653,
+      "step": 45435
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.74098701996541e-06,
+      "loss": 0.2215,
+      "step": 45436
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7406511846590435e-06,
+      "loss": 0.288,
+      "step": 45437
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.740315349352678e-06,
+      "loss": 0.2761,
+      "step": 45438
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.739979514046312e-06,
+      "loss": 0.2673,
+      "step": 45439
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7396436787399465e-06,
+      "loss": 0.2481,
+      "step": 45440
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.73930784343358e-06,
+      "loss": 0.2574,
+      "step": 45441
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.738972008127215e-06,
+      "loss": 0.1684,
+      "step": 45442
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.738636172820849e-06,
+      "loss": 0.2567,
+      "step": 45443
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7383003375144834e-06,
+      "loss": 0.2444,
+      "step": 45444
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.737964502208118e-06,
+      "loss": 0.3259,
+      "step": 45445
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.737628666901752e-06,
+      "loss": 0.2608,
+      "step": 45446
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.737292831595386e-06,
+      "loss": 0.195,
+      "step": 45447
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.73695699628902e-06,
+      "loss": 0.2155,
+      "step": 45448
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.736621160982655e-06,
+      "loss": 0.191,
+      "step": 45449
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.736285325676289e-06,
+      "loss": 0.2201,
+      "step": 45450
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7359494903699234e-06,
+      "loss": 0.4247,
+      "step": 45451
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.735613655063557e-06,
+      "loss": 0.1795,
+      "step": 45452
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.735277819757191e-06,
+      "loss": 0.2266,
+      "step": 45453
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.734941984450826e-06,
+      "loss": 0.3423,
+      "step": 45454
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.73460614914446e-06,
+      "loss": 0.3562,
+      "step": 45455
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.734270313838094e-06,
+      "loss": 0.3083,
+      "step": 45456
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.733934478531728e-06,
+      "loss": 0.4715,
+      "step": 45457
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7335986432253626e-06,
+      "loss": 0.4552,
+      "step": 45458
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.733262807918997e-06,
+      "loss": 0.4825,
+      "step": 45459
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.732926972612631e-06,
+      "loss": 0.4581,
+      "step": 45460
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.732591137306266e-06,
+      "loss": 0.1717,
+      "step": 45461
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7322553019998995e-06,
+      "loss": 0.2507,
+      "step": 45462
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.731919466693533e-06,
+      "loss": 0.3962,
+      "step": 45463
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.731583631387168e-06,
+      "loss": 0.3618,
+      "step": 45464
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7312477960808026e-06,
+      "loss": 0.2167,
+      "step": 45465
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.730911960774436e-06,
+      "loss": 0.2532,
+      "step": 45466
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.730576125468071e-06,
+      "loss": 0.2474,
+      "step": 45467
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.730240290161705e-06,
+      "loss": 0.2776,
+      "step": 45468
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7299044548553395e-06,
+      "loss": 0.4271,
+      "step": 45469
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.729568619548974e-06,
+      "loss": 0.3415,
+      "step": 45470
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.729232784242608e-06,
+      "loss": 0.2009,
+      "step": 45471
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.728896948936242e-06,
+      "loss": 0.2418,
+      "step": 45472
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.728561113629876e-06,
+      "loss": 0.1875,
+      "step": 45473
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.72822527832351e-06,
+      "loss": 0.1697,
+      "step": 45474
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.727889443017145e-06,
+      "loss": 0.3721,
+      "step": 45475
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7275536077107795e-06,
+      "loss": 0.1674,
+      "step": 45476
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.727217772404413e-06,
+      "loss": 0.4537,
+      "step": 45477
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.726881937098047e-06,
+      "loss": 0.2092,
+      "step": 45478
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.726546101791682e-06,
+      "loss": 0.264,
+      "step": 45479
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.726210266485316e-06,
+      "loss": 0.2592,
+      "step": 45480
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.72587443117895e-06,
+      "loss": 0.2959,
+      "step": 45481
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.725538595872585e-06,
+      "loss": 0.2415,
+      "step": 45482
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.725202760566219e-06,
+      "loss": 0.3372,
+      "step": 45483
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.724866925259852e-06,
+      "loss": 0.2555,
+      "step": 45484
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.724531089953487e-06,
+      "loss": 0.3808,
+      "step": 45485
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.724195254647122e-06,
+      "loss": 0.3125,
+      "step": 45486
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7238594193407555e-06,
+      "loss": 0.3494,
+      "step": 45487
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.72352358403439e-06,
+      "loss": 0.4684,
+      "step": 45488
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.723187748728024e-06,
+      "loss": 0.4389,
+      "step": 45489
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.722851913421659e-06,
+      "loss": 0.2566,
+      "step": 45490
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.722516078115293e-06,
+      "loss": 0.1747,
+      "step": 45491
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.722180242808927e-06,
+      "loss": 0.336,
+      "step": 45492
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.721844407502561e-06,
+      "loss": 0.2535,
+      "step": 45493
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7215085721961955e-06,
+      "loss": 0.202,
+      "step": 45494
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.721172736889829e-06,
+      "loss": 0.144,
+      "step": 45495
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.720836901583464e-06,
+      "loss": 0.1531,
+      "step": 45496
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7205010662770986e-06,
+      "loss": 0.3194,
+      "step": 45497
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.720165230970732e-06,
+      "loss": 0.3843,
+      "step": 45498
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.719829395664366e-06,
+      "loss": 0.2453,
+      "step": 45499
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.719493560358001e-06,
+      "loss": 0.2506,
+      "step": 45500
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7191577250516355e-06,
+      "loss": 0.2961,
+      "step": 45501
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.718821889745269e-06,
+      "loss": 0.3101,
+      "step": 45502
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.718486054438903e-06,
+      "loss": 0.3168,
+      "step": 45503
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.718150219132538e-06,
+      "loss": 0.248,
+      "step": 45504
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.717814383826172e-06,
+      "loss": 0.1551,
+      "step": 45505
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.717478548519806e-06,
+      "loss": 0.2256,
+      "step": 45506
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.717142713213441e-06,
+      "loss": 0.2464,
+      "step": 45507
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.716806877907075e-06,
+      "loss": 0.4495,
+      "step": 45508
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.716471042600708e-06,
+      "loss": 0.2657,
+      "step": 45509
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.716135207294343e-06,
+      "loss": 0.2782,
+      "step": 45510
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.715799371987978e-06,
+      "loss": 0.2725,
+      "step": 45511
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7154635366816115e-06,
+      "loss": 0.169,
+      "step": 45512
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.715127701375246e-06,
+      "loss": 0.2335,
+      "step": 45513
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.71479186606888e-06,
+      "loss": 0.2193,
+      "step": 45514
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.714456030762515e-06,
+      "loss": 0.429,
+      "step": 45515
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.714120195456149e-06,
+      "loss": 0.291,
+      "step": 45516
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.713784360149783e-06,
+      "loss": 0.2648,
+      "step": 45517
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.713448524843417e-06,
+      "loss": 0.246,
+      "step": 45518
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7131126895370515e-06,
+      "loss": 0.2249,
+      "step": 45519
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.712776854230685e-06,
+      "loss": 0.2064,
+      "step": 45520
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.71244101892432e-06,
+      "loss": 0.3143,
+      "step": 45521
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.712105183617955e-06,
+      "loss": 0.3112,
+      "step": 45522
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.711769348311588e-06,
+      "loss": 0.2552,
+      "step": 45523
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.711433513005222e-06,
+      "loss": 0.2538,
+      "step": 45524
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.711097677698857e-06,
+      "loss": 0.164,
+      "step": 45525
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7107618423924915e-06,
+      "loss": 0.3073,
+      "step": 45526
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.710426007086125e-06,
+      "loss": 0.2634,
+      "step": 45527
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.71009017177976e-06,
+      "loss": 0.163,
+      "step": 45528
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.709754336473394e-06,
+      "loss": 0.2172,
+      "step": 45529
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7094185011670275e-06,
+      "loss": 0.249,
+      "step": 45530
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.709082665860662e-06,
+      "loss": 0.2145,
+      "step": 45531
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.708746830554297e-06,
+      "loss": 0.2623,
+      "step": 45532
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.708410995247931e-06,
+      "loss": 0.3405,
+      "step": 45533
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.708075159941565e-06,
+      "loss": 0.1653,
+      "step": 45534
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.707739324635199e-06,
+      "loss": 0.2496,
+      "step": 45535
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.707403489328834e-06,
+      "loss": 0.3316,
+      "step": 45536
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.707067654022468e-06,
+      "loss": 0.3234,
+      "step": 45537
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.706731818716102e-06,
+      "loss": 0.2567,
+      "step": 45538
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.706395983409736e-06,
+      "loss": 0.1062,
+      "step": 45539
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.706060148103371e-06,
+      "loss": 0.45,
+      "step": 45540
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7057243127970044e-06,
+      "loss": 0.3189,
+      "step": 45541
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.705388477490639e-06,
+      "loss": 0.1945,
+      "step": 45542
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.705052642184274e-06,
+      "loss": 0.2134,
+      "step": 45543
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7047168068779075e-06,
+      "loss": 0.4506,
+      "step": 45544
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.704380971571541e-06,
+      "loss": 0.3172,
+      "step": 45545
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.704045136265176e-06,
+      "loss": 0.1448,
+      "step": 45546
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.703709300958811e-06,
+      "loss": 0.2608,
+      "step": 45547
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.703373465652444e-06,
+      "loss": 0.2309,
+      "step": 45548
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.703037630346078e-06,
+      "loss": 0.3015,
+      "step": 45549
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.702701795039713e-06,
+      "loss": 0.25,
+      "step": 45550
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.702365959733347e-06,
+      "loss": 0.2039,
+      "step": 45551
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.702030124426981e-06,
+      "loss": 0.0982,
+      "step": 45552
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.701694289120616e-06,
+      "loss": 0.2612,
+      "step": 45553
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.70135845381425e-06,
+      "loss": 0.175,
+      "step": 45554
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.7010226185078836e-06,
+      "loss": 0.2461,
+      "step": 45555
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.700686783201518e-06,
+      "loss": 0.2826,
+      "step": 45556
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.700350947895153e-06,
+      "loss": 0.2534,
+      "step": 45557
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 4.700015112588787e-06,
+      "loss": 0.325,
+      "step": 45558
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.699679277282421e-06,
+      "loss": 0.3291,
+      "step": 45559
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.699343441976055e-06,
+      "loss": 0.2887,
+      "step": 45560
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.69900760666969e-06,
+      "loss": 0.2158,
+      "step": 45561
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6986717713633236e-06,
+      "loss": 0.1581,
+      "step": 45562
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.698335936056958e-06,
+      "loss": 0.1665,
+      "step": 45563
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.698000100750592e-06,
+      "loss": 0.2064,
+      "step": 45564
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.697664265444227e-06,
+      "loss": 0.1898,
+      "step": 45565
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6973284301378604e-06,
+      "loss": 0.2629,
+      "step": 45566
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.696992594831495e-06,
+      "loss": 0.2287,
+      "step": 45567
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.69665675952513e-06,
+      "loss": 0.3369,
+      "step": 45568
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6963209242187635e-06,
+      "loss": 0.383,
+      "step": 45569
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.695985088912397e-06,
+      "loss": 0.2574,
+      "step": 45570
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.695649253606032e-06,
+      "loss": 0.3925,
+      "step": 45571
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.695313418299667e-06,
+      "loss": 0.232,
+      "step": 45572
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6949775829933004e-06,
+      "loss": 0.2417,
+      "step": 45573
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.694641747686935e-06,
+      "loss": 0.3909,
+      "step": 45574
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.694305912380569e-06,
+      "loss": 0.2133,
+      "step": 45575
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.693970077074203e-06,
+      "loss": 0.2714,
+      "step": 45576
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.693634241767837e-06,
+      "loss": 0.2988,
+      "step": 45577
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.693298406461472e-06,
+      "loss": 0.3134,
+      "step": 45578
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.692962571155106e-06,
+      "loss": 0.2542,
+      "step": 45579
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6926267358487404e-06,
+      "loss": 0.3526,
+      "step": 45580
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.692290900542374e-06,
+      "loss": 0.2509,
+      "step": 45581
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.691955065236009e-06,
+      "loss": 0.2889,
+      "step": 45582
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6916192299296435e-06,
+      "loss": 0.3771,
+      "step": 45583
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.691283394623277e-06,
+      "loss": 0.2569,
+      "step": 45584
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.690947559316911e-06,
+      "loss": 0.1694,
+      "step": 45585
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.690611724010546e-06,
+      "loss": 0.1982,
+      "step": 45586
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6902758887041796e-06,
+      "loss": 0.2246,
+      "step": 45587
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.689940053397814e-06,
+      "loss": 0.113,
+      "step": 45588
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.689604218091449e-06,
+      "loss": 0.1903,
+      "step": 45589
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.689268382785083e-06,
+      "loss": 0.3757,
+      "step": 45590
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6889325474787165e-06,
+      "loss": 0.4521,
+      "step": 45591
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.688596712172351e-06,
+      "loss": 0.2222,
+      "step": 45592
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.688260876865986e-06,
+      "loss": 0.201,
+      "step": 45593
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6879250415596196e-06,
+      "loss": 0.172,
+      "step": 45594
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.687589206253253e-06,
+      "loss": 0.3895,
+      "step": 45595
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.687253370946888e-06,
+      "loss": 0.3612,
+      "step": 45596
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.686917535640522e-06,
+      "loss": 0.3281,
+      "step": 45597
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6865817003341565e-06,
+      "loss": 0.3284,
+      "step": 45598
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.686245865027791e-06,
+      "loss": 0.3831,
+      "step": 45599
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.685910029721425e-06,
+      "loss": 0.2912,
+      "step": 45600
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.685574194415059e-06,
+      "loss": 0.5747,
+      "step": 45601
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.685238359108693e-06,
+      "loss": 0.2979,
+      "step": 45602
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.684902523802328e-06,
+      "loss": 0.2789,
+      "step": 45603
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.684566688495962e-06,
+      "loss": 0.3866,
+      "step": 45604
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6842308531895964e-06,
+      "loss": 0.2538,
+      "step": 45605
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.68389501788323e-06,
+      "loss": 0.2051,
+      "step": 45606
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.683559182576865e-06,
+      "loss": 0.1801,
+      "step": 45607
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.683223347270499e-06,
+      "loss": 0.2835,
+      "step": 45608
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.682887511964133e-06,
+      "loss": 0.2154,
+      "step": 45609
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.682551676657767e-06,
+      "loss": 0.3514,
+      "step": 45610
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.682215841351402e-06,
+      "loss": 0.3978,
+      "step": 45611
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.681880006045036e-06,
+      "loss": 0.3564,
+      "step": 45612
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.68154417073867e-06,
+      "loss": 0.1309,
+      "step": 45613
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.681208335432305e-06,
+      "loss": 0.2502,
+      "step": 45614
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.680872500125939e-06,
+      "loss": 0.2462,
+      "step": 45615
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6805366648195725e-06,
+      "loss": 0.1706,
+      "step": 45616
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.680200829513207e-06,
+      "loss": 0.2093,
+      "step": 45617
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.679864994206842e-06,
+      "loss": 0.2133,
+      "step": 45618
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.679529158900476e-06,
+      "loss": 0.2885,
+      "step": 45619
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.67919332359411e-06,
+      "loss": 0.5958,
+      "step": 45620
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.678857488287744e-06,
+      "loss": 0.345,
+      "step": 45621
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.678521652981378e-06,
+      "loss": 0.2304,
+      "step": 45622
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6781858176750125e-06,
+      "loss": 0.3309,
+      "step": 45623
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.677849982368647e-06,
+      "loss": 0.1818,
+      "step": 45624
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.677514147062281e-06,
+      "loss": 0.3503,
+      "step": 45625
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6771783117559156e-06,
+      "loss": 0.3665,
+      "step": 45626
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.676842476449549e-06,
+      "loss": 0.2705,
+      "step": 45627
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.676506641143184e-06,
+      "loss": 0.2616,
+      "step": 45628
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.676170805836819e-06,
+      "loss": 0.2354,
+      "step": 45629
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6758349705304525e-06,
+      "loss": 0.3158,
+      "step": 45630
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.675499135224086e-06,
+      "loss": 0.181,
+      "step": 45631
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.675163299917721e-06,
+      "loss": 0.1177,
+      "step": 45632
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.674827464611355e-06,
+      "loss": 0.1946,
+      "step": 45633
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.674491629304989e-06,
+      "loss": 0.2478,
+      "step": 45634
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.674155793998623e-06,
+      "loss": 0.3166,
+      "step": 45635
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.673819958692258e-06,
+      "loss": 0.2754,
+      "step": 45636
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.673484123385892e-06,
+      "loss": 0.3275,
+      "step": 45637
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.673148288079526e-06,
+      "loss": 0.1771,
+      "step": 45638
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.672812452773161e-06,
+      "loss": 0.3416,
+      "step": 45639
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.672476617466795e-06,
+      "loss": 0.119,
+      "step": 45640
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6721407821604285e-06,
+      "loss": 0.3524,
+      "step": 45641
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.671804946854063e-06,
+      "loss": 0.2951,
+      "step": 45642
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.671469111547697e-06,
+      "loss": 0.1745,
+      "step": 45643
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.671133276241332e-06,
+      "loss": 0.2252,
+      "step": 45644
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.670797440934966e-06,
+      "loss": 0.1899,
+      "step": 45645
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6704616056286e-06,
+      "loss": 0.3754,
+      "step": 45646
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.670125770322234e-06,
+      "loss": 0.2149,
+      "step": 45647
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6697899350158685e-06,
+      "loss": 0.1599,
+      "step": 45648
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.669454099709503e-06,
+      "loss": 0.1594,
+      "step": 45649
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.669118264403137e-06,
+      "loss": 0.2043,
+      "step": 45650
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.668782429096772e-06,
+      "loss": 0.3044,
+      "step": 45651
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.668446593790405e-06,
+      "loss": 0.3045,
+      "step": 45652
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.668110758484039e-06,
+      "loss": 0.1649,
+      "step": 45653
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.667774923177674e-06,
+      "loss": 0.2943,
+      "step": 45654
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6674390878713085e-06,
+      "loss": 0.3082,
+      "step": 45655
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.667103252564942e-06,
+      "loss": 0.2251,
+      "step": 45656
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.666767417258577e-06,
+      "loss": 0.2394,
+      "step": 45657
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.666431581952211e-06,
+      "loss": 0.2106,
+      "step": 45658
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.666095746645845e-06,
+      "loss": 0.2315,
+      "step": 45659
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.66575991133948e-06,
+      "loss": 0.1773,
+      "step": 45660
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.665424076033114e-06,
+      "loss": 0.4338,
+      "step": 45661
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.665088240726748e-06,
+      "loss": 0.1436,
+      "step": 45662
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.664752405420382e-06,
+      "loss": 0.1877,
+      "step": 45663
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.664416570114016e-06,
+      "loss": 0.2485,
+      "step": 45664
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.664080734807651e-06,
+      "loss": 0.2661,
+      "step": 45665
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.663744899501285e-06,
+      "loss": 0.2992,
+      "step": 45666
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.663409064194919e-06,
+      "loss": 0.461,
+      "step": 45667
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.663073228888553e-06,
+      "loss": 0.1058,
+      "step": 45668
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.662737393582188e-06,
+      "loss": 0.2521,
+      "step": 45669
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.662401558275822e-06,
+      "loss": 0.2939,
+      "step": 45670
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.662065722969456e-06,
+      "loss": 0.2937,
+      "step": 45671
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.661729887663091e-06,
+      "loss": 0.1872,
+      "step": 45672
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6613940523567245e-06,
+      "loss": 0.1688,
+      "step": 45673
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.661058217050359e-06,
+      "loss": 0.4161,
+      "step": 45674
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.660722381743993e-06,
+      "loss": 0.3447,
+      "step": 45675
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.660386546437628e-06,
+      "loss": 0.2587,
+      "step": 45676
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.660050711131261e-06,
+      "loss": 0.4508,
+      "step": 45677
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.659714875824896e-06,
+      "loss": 0.2846,
+      "step": 45678
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.65937904051853e-06,
+      "loss": 0.2639,
+      "step": 45679
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6590432052121645e-06,
+      "loss": 0.3292,
+      "step": 45680
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.658707369905798e-06,
+      "loss": 0.3681,
+      "step": 45681
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.658371534599433e-06,
+      "loss": 0.281,
+      "step": 45682
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.658035699293067e-06,
+      "loss": 0.1513,
+      "step": 45683
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.657699863986701e-06,
+      "loss": 0.2989,
+      "step": 45684
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.657364028680336e-06,
+      "loss": 0.306,
+      "step": 45685
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.65702819337397e-06,
+      "loss": 0.188,
+      "step": 45686
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.656692358067604e-06,
+      "loss": 0.2735,
+      "step": 45687
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.656356522761238e-06,
+      "loss": 0.2252,
+      "step": 45688
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.656020687454872e-06,
+      "loss": 0.1847,
+      "step": 45689
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.655684852148507e-06,
+      "loss": 0.2688,
+      "step": 45690
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.655349016842141e-06,
+      "loss": 0.3349,
+      "step": 45691
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.655013181535775e-06,
+      "loss": 0.2465,
+      "step": 45692
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.654677346229409e-06,
+      "loss": 0.3149,
+      "step": 45693
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.654341510923044e-06,
+      "loss": 0.2753,
+      "step": 45694
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.654005675616678e-06,
+      "loss": 0.1964,
+      "step": 45695
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.653669840310312e-06,
+      "loss": 0.1868,
+      "step": 45696
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.653334005003947e-06,
+      "loss": 0.1924,
+      "step": 45697
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6529981696975805e-06,
+      "loss": 0.4518,
+      "step": 45698
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.652662334391214e-06,
+      "loss": 0.3156,
+      "step": 45699
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.652326499084849e-06,
+      "loss": 0.3615,
+      "step": 45700
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.651990663778484e-06,
+      "loss": 0.1778,
+      "step": 45701
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6516548284721174e-06,
+      "loss": 0.1929,
+      "step": 45702
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.651318993165752e-06,
+      "loss": 0.2501,
+      "step": 45703
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.650983157859386e-06,
+      "loss": 0.1646,
+      "step": 45704
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6506473225530205e-06,
+      "loss": 0.4873,
+      "step": 45705
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.650311487246655e-06,
+      "loss": 0.3118,
+      "step": 45706
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.649975651940289e-06,
+      "loss": 0.3715,
+      "step": 45707
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.649639816633923e-06,
+      "loss": 0.2275,
+      "step": 45708
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6493039813275574e-06,
+      "loss": 0.1167,
+      "step": 45709
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.648968146021191e-06,
+      "loss": 0.2572,
+      "step": 45710
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.648632310714826e-06,
+      "loss": 0.1886,
+      "step": 45711
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6482964754084605e-06,
+      "loss": 0.3043,
+      "step": 45712
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.647960640102094e-06,
+      "loss": 0.3497,
+      "step": 45713
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.647624804795728e-06,
+      "loss": 0.3242,
+      "step": 45714
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.647288969489363e-06,
+      "loss": 0.2976,
+      "step": 45715
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.646953134182997e-06,
+      "loss": 0.1803,
+      "step": 45716
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.646617298876631e-06,
+      "loss": 0.3693,
+      "step": 45717
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.646281463570266e-06,
+      "loss": 0.3392,
+      "step": 45718
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6459456282639e-06,
+      "loss": 0.1489,
+      "step": 45719
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.645609792957534e-06,
+      "loss": 0.1903,
+      "step": 45720
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.645273957651168e-06,
+      "loss": 0.4321,
+      "step": 45721
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.644938122344803e-06,
+      "loss": 0.2871,
+      "step": 45722
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6446022870384366e-06,
+      "loss": 0.2811,
+      "step": 45723
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.644266451732071e-06,
+      "loss": 0.2654,
+      "step": 45724
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.643930616425705e-06,
+      "loss": 0.1904,
+      "step": 45725
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.64359478111934e-06,
+      "loss": 0.4526,
+      "step": 45726
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6432589458129735e-06,
+      "loss": 0.2659,
+      "step": 45727
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.642923110506608e-06,
+      "loss": 0.191,
+      "step": 45728
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.642587275200242e-06,
+      "loss": 0.2193,
+      "step": 45729
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6422514398938765e-06,
+      "loss": 0.2665,
+      "step": 45730
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.641915604587511e-06,
+      "loss": 0.2992,
+      "step": 45731
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.641579769281145e-06,
+      "loss": 0.2553,
+      "step": 45732
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.641243933974779e-06,
+      "loss": 0.3174,
+      "step": 45733
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6409080986684134e-06,
+      "loss": 0.3095,
+      "step": 45734
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.640572263362047e-06,
+      "loss": 0.2803,
+      "step": 45735
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.640236428055682e-06,
+      "loss": 0.3014,
+      "step": 45736
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6399005927493165e-06,
+      "loss": 0.1862,
+      "step": 45737
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.63956475744295e-06,
+      "loss": 0.1763,
+      "step": 45738
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.639228922136584e-06,
+      "loss": 0.3432,
+      "step": 45739
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.638893086830219e-06,
+      "loss": 0.3593,
+      "step": 45740
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6385572515238534e-06,
+      "loss": 0.2164,
+      "step": 45741
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.638221416217487e-06,
+      "loss": 0.3217,
+      "step": 45742
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.637885580911122e-06,
+      "loss": 0.2501,
+      "step": 45743
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.637549745604756e-06,
+      "loss": 0.2367,
+      "step": 45744
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6372139102983895e-06,
+      "loss": 0.3005,
+      "step": 45745
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.636878074992024e-06,
+      "loss": 0.2702,
+      "step": 45746
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.636542239685659e-06,
+      "loss": 0.2642,
+      "step": 45747
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.636206404379293e-06,
+      "loss": 0.1485,
+      "step": 45748
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.635870569072927e-06,
+      "loss": 0.2272,
+      "step": 45749
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.635534733766561e-06,
+      "loss": 0.2694,
+      "step": 45750
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.635198898460196e-06,
+      "loss": 0.2114,
+      "step": 45751
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.63486306315383e-06,
+      "loss": 0.2659,
+      "step": 45752
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.634527227847464e-06,
+      "loss": 0.3125,
+      "step": 45753
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.634191392541098e-06,
+      "loss": 0.3987,
+      "step": 45754
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.6338555572347326e-06,
+      "loss": 0.2519,
+      "step": 45755
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 4.633519721928366e-06,
+      "loss": 0.3117,
+      "step": 45756
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.633183886622001e-06,
+      "loss": 0.1005,
+      "step": 45757
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.632848051315636e-06,
+      "loss": 0.2563,
+      "step": 45758
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.6325122160092695e-06,
+      "loss": 0.2907,
+      "step": 45759
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.632176380702903e-06,
+      "loss": 0.2073,
+      "step": 45760
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.631840545396538e-06,
+      "loss": 0.3862,
+      "step": 45761
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.6315047100901726e-06,
+      "loss": 0.178,
+      "step": 45762
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.631168874783806e-06,
+      "loss": 0.1656,
+      "step": 45763
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.630833039477441e-06,
+      "loss": 0.2983,
+      "step": 45764
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.630497204171075e-06,
+      "loss": 0.243,
+      "step": 45765
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.630161368864709e-06,
+      "loss": 0.2982,
+      "step": 45766
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.629825533558343e-06,
+      "loss": 0.2604,
+      "step": 45767
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.629489698251978e-06,
+      "loss": 0.2608,
+      "step": 45768
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.629153862945612e-06,
+      "loss": 0.2106,
+      "step": 45769
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.628818027639246e-06,
+      "loss": 0.348,
+      "step": 45770
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.62848219233288e-06,
+      "loss": 0.2107,
+      "step": 45771
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.628146357026515e-06,
+      "loss": 0.1712,
+      "step": 45772
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.627810521720149e-06,
+      "loss": 0.3487,
+      "step": 45773
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.627474686413783e-06,
+      "loss": 0.1751,
+      "step": 45774
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.627138851107417e-06,
+      "loss": 0.2889,
+      "step": 45775
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.626803015801052e-06,
+      "loss": 0.3048,
+      "step": 45776
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.6264671804946855e-06,
+      "loss": 0.2338,
+      "step": 45777
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.62613134518832e-06,
+      "loss": 0.2352,
+      "step": 45778
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.625795509881954e-06,
+      "loss": 0.2308,
+      "step": 45779
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.625459674575589e-06,
+      "loss": 0.3496,
+      "step": 45780
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.625123839269222e-06,
+      "loss": 0.2936,
+      "step": 45781
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.624788003962857e-06,
+      "loss": 0.2207,
+      "step": 45782
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.624452168656492e-06,
+      "loss": 0.1754,
+      "step": 45783
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.6241163333501255e-06,
+      "loss": 0.1719,
+      "step": 45784
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.623780498043759e-06,
+      "loss": 0.2326,
+      "step": 45785
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.623444662737394e-06,
+      "loss": 0.1337,
+      "step": 45786
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.623108827431029e-06,
+      "loss": 0.2564,
+      "step": 45787
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.622772992124662e-06,
+      "loss": 0.2094,
+      "step": 45788
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.622437156818297e-06,
+      "loss": 0.2192,
+      "step": 45789
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.622101321511931e-06,
+      "loss": 0.3557,
+      "step": 45790
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.621765486205565e-06,
+      "loss": 0.3098,
+      "step": 45791
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.621429650899199e-06,
+      "loss": 0.2364,
+      "step": 45792
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.621093815592834e-06,
+      "loss": 0.2988,
+      "step": 45793
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.620757980286468e-06,
+      "loss": 0.2436,
+      "step": 45794
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.620422144980102e-06,
+      "loss": 0.2128,
+      "step": 45795
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.620086309673736e-06,
+      "loss": 0.3248,
+      "step": 45796
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.619750474367371e-06,
+      "loss": 0.2072,
+      "step": 45797
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.6194146390610055e-06,
+      "loss": 0.1658,
+      "step": 45798
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.619078803754639e-06,
+      "loss": 0.3956,
+      "step": 45799
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.618742968448273e-06,
+      "loss": 0.2256,
+      "step": 45800
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.618407133141908e-06,
+      "loss": 0.2079,
+      "step": 45801
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.6180712978355415e-06,
+      "loss": 0.4029,
+      "step": 45802
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.617735462529176e-06,
+      "loss": 0.2494,
+      "step": 45803
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.617399627222811e-06,
+      "loss": 0.2925,
+      "step": 45804
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.617063791916445e-06,
+      "loss": 0.3142,
+      "step": 45805
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.616727956610078e-06,
+      "loss": 0.1587,
+      "step": 45806
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.616392121303713e-06,
+      "loss": 0.2579,
+      "step": 45807
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.616056285997348e-06,
+      "loss": 0.2176,
+      "step": 45808
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.6157204506909815e-06,
+      "loss": 0.2202,
+      "step": 45809
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.615384615384616e-06,
+      "loss": 0.2107,
+      "step": 45810
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.61504878007825e-06,
+      "loss": 0.3041,
+      "step": 45811
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.614712944771884e-06,
+      "loss": 0.2019,
+      "step": 45812
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.614377109465518e-06,
+      "loss": 0.2919,
+      "step": 45813
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.614041274159153e-06,
+      "loss": 0.1465,
+      "step": 45814
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.613705438852787e-06,
+      "loss": 0.3711,
+      "step": 45815
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.613369603546421e-06,
+      "loss": 0.3076,
+      "step": 45816
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.613033768240055e-06,
+      "loss": 0.2183,
+      "step": 45817
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.61269793293369e-06,
+      "loss": 0.3029,
+      "step": 45818
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.612362097627324e-06,
+      "loss": 0.4647,
+      "step": 45819
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.612026262320958e-06,
+      "loss": 0.3383,
+      "step": 45820
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.611690427014592e-06,
+      "loss": 0.1953,
+      "step": 45821
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.611354591708226e-06,
+      "loss": 0.2994,
+      "step": 45822
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.611018756401861e-06,
+      "loss": 0.1998,
+      "step": 45823
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.610682921095495e-06,
+      "loss": 0.3354,
+      "step": 45824
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.610347085789129e-06,
+      "loss": 0.1689,
+      "step": 45825
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.610011250482764e-06,
+      "loss": 0.2877,
+      "step": 45826
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.6096754151763975e-06,
+      "loss": 0.1428,
+      "step": 45827
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.609339579870032e-06,
+      "loss": 0.246,
+      "step": 45828
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.609003744563667e-06,
+      "loss": 0.2745,
+      "step": 45829
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.608667909257301e-06,
+      "loss": 0.2613,
+      "step": 45830
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.6083320739509344e-06,
+      "loss": 0.122,
+      "step": 45831
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.607996238644569e-06,
+      "loss": 0.3768,
+      "step": 45832
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.607660403338204e-06,
+      "loss": 0.3136,
+      "step": 45833
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.6073245680318375e-06,
+      "loss": 0.3772,
+      "step": 45834
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.606988732725472e-06,
+      "loss": 0.1516,
+      "step": 45835
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.606652897419106e-06,
+      "loss": 0.3532,
+      "step": 45836
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.60631706211274e-06,
+      "loss": 0.224,
+      "step": 45837
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.605981226806374e-06,
+      "loss": 0.2951,
+      "step": 45838
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.605645391500009e-06,
+      "loss": 0.147,
+      "step": 45839
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.605309556193643e-06,
+      "loss": 0.1962,
+      "step": 45840
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.6049737208872775e-06,
+      "loss": 0.2815,
+      "step": 45841
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.604637885580911e-06,
+      "loss": 0.4586,
+      "step": 45842
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.604302050274546e-06,
+      "loss": 0.1884,
+      "step": 45843
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.603966214968181e-06,
+      "loss": 0.3437,
+      "step": 45844
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.603630379661814e-06,
+      "loss": 0.2326,
+      "step": 45845
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.603294544355448e-06,
+      "loss": 0.1758,
+      "step": 45846
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.602958709049083e-06,
+      "loss": 0.2882,
+      "step": 45847
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.602622873742717e-06,
+      "loss": 0.1906,
+      "step": 45848
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.602287038436351e-06,
+      "loss": 0.177,
+      "step": 45849
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.601951203129986e-06,
+      "loss": 0.1487,
+      "step": 45850
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.60161536782362e-06,
+      "loss": 0.3641,
+      "step": 45851
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.6012795325172536e-06,
+      "loss": 0.2058,
+      "step": 45852
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.600943697210888e-06,
+      "loss": 0.2488,
+      "step": 45853
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.600607861904523e-06,
+      "loss": 0.2816,
+      "step": 45854
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.600272026598157e-06,
+      "loss": 0.2667,
+      "step": 45855
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.599936191291791e-06,
+      "loss": 0.2496,
+      "step": 45856
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.599600355985425e-06,
+      "loss": 0.2725,
+      "step": 45857
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.599264520679059e-06,
+      "loss": 0.2065,
+      "step": 45858
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.5989286853726935e-06,
+      "loss": 0.323,
+      "step": 45859
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.598592850066328e-06,
+      "loss": 0.2141,
+      "step": 45860
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.598257014759962e-06,
+      "loss": 0.3794,
+      "step": 45861
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.597921179453596e-06,
+      "loss": 0.1925,
+      "step": 45862
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.5975853441472304e-06,
+      "loss": 0.2385,
+      "step": 45863
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.597249508840865e-06,
+      "loss": 0.2003,
+      "step": 45864
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.596913673534499e-06,
+      "loss": 0.145,
+      "step": 45865
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.5965778382281335e-06,
+      "loss": 0.4204,
+      "step": 45866
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.596242002921767e-06,
+      "loss": 0.2563,
+      "step": 45867
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.595906167615401e-06,
+      "loss": 0.2216,
+      "step": 45868
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.595570332309036e-06,
+      "loss": 0.3788,
+      "step": 45869
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.5952344970026704e-06,
+      "loss": 0.3417,
+      "step": 45870
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.594898661696304e-06,
+      "loss": 0.2297,
+      "step": 45871
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.594562826389939e-06,
+      "loss": 0.3061,
+      "step": 45872
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.594226991083573e-06,
+      "loss": 0.1966,
+      "step": 45873
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.593891155777207e-06,
+      "loss": 0.2972,
+      "step": 45874
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.593555320470842e-06,
+      "loss": 0.2231,
+      "step": 45875
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.593219485164476e-06,
+      "loss": 0.2572,
+      "step": 45876
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.5928836498581096e-06,
+      "loss": 0.3246,
+      "step": 45877
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.592547814551744e-06,
+      "loss": 0.3204,
+      "step": 45878
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.592211979245378e-06,
+      "loss": 0.2124,
+      "step": 45879
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.591876143939013e-06,
+      "loss": 0.1336,
+      "step": 45880
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.591540308632647e-06,
+      "loss": 0.2565,
+      "step": 45881
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.591204473326281e-06,
+      "loss": 0.1284,
+      "step": 45882
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.590868638019915e-06,
+      "loss": 0.1408,
+      "step": 45883
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.5905328027135496e-06,
+      "loss": 0.1604,
+      "step": 45884
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.590196967407184e-06,
+      "loss": 0.3439,
+      "step": 45885
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.589861132100818e-06,
+      "loss": 0.2363,
+      "step": 45886
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.589525296794453e-06,
+      "loss": 0.2957,
+      "step": 45887
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.5891894614880865e-06,
+      "loss": 0.2128,
+      "step": 45888
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.588853626181721e-06,
+      "loss": 0.1675,
+      "step": 45889
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.588517790875355e-06,
+      "loss": 0.2434,
+      "step": 45890
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.5881819555689896e-06,
+      "loss": 0.2423,
+      "step": 45891
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.587846120262623e-06,
+      "loss": 0.2843,
+      "step": 45892
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.587510284956258e-06,
+      "loss": 0.2837,
+      "step": 45893
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.587174449649892e-06,
+      "loss": 0.1829,
+      "step": 45894
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.5868386143435265e-06,
+      "loss": 0.2018,
+      "step": 45895
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.586502779037161e-06,
+      "loss": 0.2017,
+      "step": 45896
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.586166943730795e-06,
+      "loss": 0.2934,
+      "step": 45897
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.585831108424429e-06,
+      "loss": 0.5935,
+      "step": 45898
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.585495273118063e-06,
+      "loss": 0.2711,
+      "step": 45899
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.585159437811698e-06,
+      "loss": 0.198,
+      "step": 45900
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.584823602505332e-06,
+      "loss": 0.2765,
+      "step": 45901
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.5844877671989664e-06,
+      "loss": 0.3513,
+      "step": 45902
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.5841519318926e-06,
+      "loss": 0.3794,
+      "step": 45903
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.583816096586234e-06,
+      "loss": 0.1875,
+      "step": 45904
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.583480261279869e-06,
+      "loss": 0.2402,
+      "step": 45905
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.583144425973503e-06,
+      "loss": 0.2928,
+      "step": 45906
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.582808590667137e-06,
+      "loss": 0.2022,
+      "step": 45907
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.582472755360771e-06,
+      "loss": 0.1497,
+      "step": 45908
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.582136920054406e-06,
+      "loss": 0.2593,
+      "step": 45909
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.58180108474804e-06,
+      "loss": 0.4338,
+      "step": 45910
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.581465249441674e-06,
+      "loss": 0.2992,
+      "step": 45911
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.581129414135309e-06,
+      "loss": 0.2367,
+      "step": 45912
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.5807935788289425e-06,
+      "loss": 0.2327,
+      "step": 45913
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.580457743522576e-06,
+      "loss": 0.3308,
+      "step": 45914
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.580121908216211e-06,
+      "loss": 0.3607,
+      "step": 45915
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.579786072909846e-06,
+      "loss": 0.3479,
+      "step": 45916
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.579450237603479e-06,
+      "loss": 0.2374,
+      "step": 45917
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.579114402297114e-06,
+      "loss": 0.235,
+      "step": 45918
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.578778566990748e-06,
+      "loss": 0.351,
+      "step": 45919
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.5784427316843825e-06,
+      "loss": 0.1375,
+      "step": 45920
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.578106896378017e-06,
+      "loss": 0.2684,
+      "step": 45921
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.577771061071651e-06,
+      "loss": 0.2314,
+      "step": 45922
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.577435225765285e-06,
+      "loss": 0.1534,
+      "step": 45923
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.577099390458919e-06,
+      "loss": 0.2227,
+      "step": 45924
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.576763555152553e-06,
+      "loss": 0.3058,
+      "step": 45925
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.576427719846188e-06,
+      "loss": 0.2815,
+      "step": 45926
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.5760918845398225e-06,
+      "loss": 0.2138,
+      "step": 45927
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.575756049233456e-06,
+      "loss": 0.2892,
+      "step": 45928
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.57542021392709e-06,
+      "loss": 0.2364,
+      "step": 45929
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.575084378620725e-06,
+      "loss": 0.503,
+      "step": 45930
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.574748543314359e-06,
+      "loss": 0.1115,
+      "step": 45931
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.574412708007993e-06,
+      "loss": 0.0903,
+      "step": 45932
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.574076872701628e-06,
+      "loss": 0.2034,
+      "step": 45933
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.573741037395262e-06,
+      "loss": 0.2852,
+      "step": 45934
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.573405202088895e-06,
+      "loss": 0.5045,
+      "step": 45935
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.57306936678253e-06,
+      "loss": 0.1737,
+      "step": 45936
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.572733531476165e-06,
+      "loss": 0.3551,
+      "step": 45937
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.5723976961697985e-06,
+      "loss": 0.2682,
+      "step": 45938
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.572061860863433e-06,
+      "loss": 0.3624,
+      "step": 45939
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.571726025557067e-06,
+      "loss": 0.2905,
+      "step": 45940
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.571390190250702e-06,
+      "loss": 0.3182,
+      "step": 45941
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.571054354944336e-06,
+      "loss": 0.2475,
+      "step": 45942
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.57071851963797e-06,
+      "loss": 0.1706,
+      "step": 45943
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.570382684331604e-06,
+      "loss": 0.3131,
+      "step": 45944
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.5700468490252385e-06,
+      "loss": 0.2832,
+      "step": 45945
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.569711013718872e-06,
+      "loss": 0.2013,
+      "step": 45946
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.569375178412507e-06,
+      "loss": 0.2841,
+      "step": 45947
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.569039343106142e-06,
+      "loss": 0.1841,
+      "step": 45948
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.568703507799775e-06,
+      "loss": 0.1971,
+      "step": 45949
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.568367672493409e-06,
+      "loss": 0.5525,
+      "step": 45950
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.568031837187044e-06,
+      "loss": 0.2418,
+      "step": 45951
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.5676960018806785e-06,
+      "loss": 0.3003,
+      "step": 45952
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.567360166574312e-06,
+      "loss": 0.3511,
+      "step": 45953
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.567024331267946e-06,
+      "loss": 0.3037,
+      "step": 45954
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 4.566688495961581e-06,
+      "loss": 0.2244,
+      "step": 45955
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.566352660655215e-06,
+      "loss": 0.223,
+      "step": 45956
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.566016825348849e-06,
+      "loss": 0.2155,
+      "step": 45957
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.565680990042484e-06,
+      "loss": 0.1858,
+      "step": 45958
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.565345154736118e-06,
+      "loss": 0.1831,
+      "step": 45959
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5650093194297514e-06,
+      "loss": 0.2057,
+      "step": 45960
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.564673484123386e-06,
+      "loss": 0.2893,
+      "step": 45961
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.564337648817021e-06,
+      "loss": 0.3343,
+      "step": 45962
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5640018135106545e-06,
+      "loss": 0.3708,
+      "step": 45963
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.563665978204289e-06,
+      "loss": 0.5116,
+      "step": 45964
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.563330142897923e-06,
+      "loss": 0.2564,
+      "step": 45965
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.562994307591558e-06,
+      "loss": 0.2075,
+      "step": 45966
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.562658472285192e-06,
+      "loss": 0.2302,
+      "step": 45967
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.562322636978826e-06,
+      "loss": 0.2831,
+      "step": 45968
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.56198680167246e-06,
+      "loss": 0.2433,
+      "step": 45969
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5616509663660945e-06,
+      "loss": 0.2689,
+      "step": 45970
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.561315131059728e-06,
+      "loss": 0.1771,
+      "step": 45971
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.560979295753363e-06,
+      "loss": 0.3699,
+      "step": 45972
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.560643460446998e-06,
+      "loss": 0.2057,
+      "step": 45973
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.560307625140631e-06,
+      "loss": 0.3938,
+      "step": 45974
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.559971789834265e-06,
+      "loss": 0.3462,
+      "step": 45975
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5596359545279e-06,
+      "loss": 0.3392,
+      "step": 45976
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5593001192215345e-06,
+      "loss": 0.2641,
+      "step": 45977
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.558964283915168e-06,
+      "loss": 0.3521,
+      "step": 45978
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.558628448608803e-06,
+      "loss": 0.2224,
+      "step": 45979
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.558292613302437e-06,
+      "loss": 0.188,
+      "step": 45980
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5579567779960706e-06,
+      "loss": 0.2525,
+      "step": 45981
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.557620942689705e-06,
+      "loss": 0.3408,
+      "step": 45982
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.55728510738334e-06,
+      "loss": 0.1663,
+      "step": 45983
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.556949272076974e-06,
+      "loss": 0.3559,
+      "step": 45984
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.556613436770608e-06,
+      "loss": 0.2013,
+      "step": 45985
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.556277601464242e-06,
+      "loss": 0.1665,
+      "step": 45986
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.555941766157877e-06,
+      "loss": 0.288,
+      "step": 45987
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.555605930851511e-06,
+      "loss": 0.3283,
+      "step": 45988
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.555270095545145e-06,
+      "loss": 0.2112,
+      "step": 45989
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.554934260238779e-06,
+      "loss": 0.2285,
+      "step": 45990
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.554598424932414e-06,
+      "loss": 0.228,
+      "step": 45991
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5542625896260474e-06,
+      "loss": 0.3792,
+      "step": 45992
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.553926754319682e-06,
+      "loss": 0.2236,
+      "step": 45993
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.553590919013317e-06,
+      "loss": 0.2298,
+      "step": 45994
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5532550837069505e-06,
+      "loss": 0.3507,
+      "step": 45995
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.552919248400584e-06,
+      "loss": 0.2038,
+      "step": 45996
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.552583413094219e-06,
+      "loss": 0.3716,
+      "step": 45997
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.552247577787854e-06,
+      "loss": 0.4442,
+      "step": 45998
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5519117424814874e-06,
+      "loss": 0.1949,
+      "step": 45999
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.551575907175121e-06,
+      "loss": 0.1599,
+      "step": 46000
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.551240071868756e-06,
+      "loss": 0.3987,
+      "step": 46001
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5509042365623905e-06,
+      "loss": 0.2862,
+      "step": 46002
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.550568401256024e-06,
+      "loss": 0.2394,
+      "step": 46003
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.550232565949659e-06,
+      "loss": 0.3008,
+      "step": 46004
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.549896730643293e-06,
+      "loss": 0.3277,
+      "step": 46005
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5495608953369266e-06,
+      "loss": 0.2659,
+      "step": 46006
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.549225060030561e-06,
+      "loss": 0.2209,
+      "step": 46007
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.548889224724196e-06,
+      "loss": 0.1824,
+      "step": 46008
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.54855338941783e-06,
+      "loss": 0.3245,
+      "step": 46009
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.548217554111464e-06,
+      "loss": 0.2498,
+      "step": 46010
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.547881718805098e-06,
+      "loss": 0.3924,
+      "step": 46011
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.547545883498733e-06,
+      "loss": 0.3679,
+      "step": 46012
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.547210048192367e-06,
+      "loss": 0.2373,
+      "step": 46013
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.546874212886001e-06,
+      "loss": 0.1944,
+      "step": 46014
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.546538377579635e-06,
+      "loss": 0.2843,
+      "step": 46015
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.54620254227327e-06,
+      "loss": 0.3847,
+      "step": 46016
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5458667069669035e-06,
+      "loss": 0.233,
+      "step": 46017
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.545530871660538e-06,
+      "loss": 0.1624,
+      "step": 46018
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.545195036354173e-06,
+      "loss": 0.2551,
+      "step": 46019
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5448592010478066e-06,
+      "loss": 0.2586,
+      "step": 46020
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.54452336574144e-06,
+      "loss": 0.2889,
+      "step": 46021
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.544187530435075e-06,
+      "loss": 0.43,
+      "step": 46022
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.54385169512871e-06,
+      "loss": 0.3742,
+      "step": 46023
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5435158598223434e-06,
+      "loss": 0.2045,
+      "step": 46024
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.543180024515978e-06,
+      "loss": 0.2881,
+      "step": 46025
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.542844189209612e-06,
+      "loss": 0.1692,
+      "step": 46026
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.542508353903246e-06,
+      "loss": 0.3222,
+      "step": 46027
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.54217251859688e-06,
+      "loss": 0.3002,
+      "step": 46028
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.541836683290515e-06,
+      "loss": 0.2413,
+      "step": 46029
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.541500847984149e-06,
+      "loss": 0.24,
+      "step": 46030
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5411650126777834e-06,
+      "loss": 0.1932,
+      "step": 46031
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.540829177371417e-06,
+      "loss": 0.2213,
+      "step": 46032
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.540493342065052e-06,
+      "loss": 0.2125,
+      "step": 46033
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5401575067586865e-06,
+      "loss": 0.2244,
+      "step": 46034
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.53982167145232e-06,
+      "loss": 0.1245,
+      "step": 46035
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.539485836145954e-06,
+      "loss": 0.3493,
+      "step": 46036
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.539150000839589e-06,
+      "loss": 0.3651,
+      "step": 46037
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.538814165533223e-06,
+      "loss": 0.2958,
+      "step": 46038
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.538478330226857e-06,
+      "loss": 0.3351,
+      "step": 46039
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.538142494920492e-06,
+      "loss": 0.4377,
+      "step": 46040
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.537806659614126e-06,
+      "loss": 0.314,
+      "step": 46041
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5374708243077595e-06,
+      "loss": 0.3956,
+      "step": 46042
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.537134989001394e-06,
+      "loss": 0.1512,
+      "step": 46043
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.536799153695029e-06,
+      "loss": 0.3086,
+      "step": 46044
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5364633183886626e-06,
+      "loss": 0.3179,
+      "step": 46045
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.536127483082296e-06,
+      "loss": 0.2286,
+      "step": 46046
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.535791647775931e-06,
+      "loss": 0.1935,
+      "step": 46047
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.535455812469565e-06,
+      "loss": 0.3312,
+      "step": 46048
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5351199771631995e-06,
+      "loss": 0.3282,
+      "step": 46049
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.534784141856834e-06,
+      "loss": 0.199,
+      "step": 46050
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.534448306550468e-06,
+      "loss": 0.1389,
+      "step": 46051
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.534112471244102e-06,
+      "loss": 0.2807,
+      "step": 46052
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.533776635937736e-06,
+      "loss": 0.1258,
+      "step": 46053
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.533440800631371e-06,
+      "loss": 0.3048,
+      "step": 46054
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.533104965325005e-06,
+      "loss": 0.247,
+      "step": 46055
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5327691300186395e-06,
+      "loss": 0.2805,
+      "step": 46056
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.532433294712273e-06,
+      "loss": 0.2133,
+      "step": 46057
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.532097459405908e-06,
+      "loss": 0.1849,
+      "step": 46058
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.531761624099542e-06,
+      "loss": 0.2136,
+      "step": 46059
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.531425788793176e-06,
+      "loss": 0.3826,
+      "step": 46060
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.53108995348681e-06,
+      "loss": 0.1888,
+      "step": 46061
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.530754118180445e-06,
+      "loss": 0.3245,
+      "step": 46062
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.530418282874079e-06,
+      "loss": 0.1944,
+      "step": 46063
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.530082447567713e-06,
+      "loss": 0.2568,
+      "step": 46064
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.529746612261348e-06,
+      "loss": 0.3674,
+      "step": 46065
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.529410776954982e-06,
+      "loss": 0.2632,
+      "step": 46066
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5290749416486155e-06,
+      "loss": 0.2738,
+      "step": 46067
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.52873910634225e-06,
+      "loss": 0.217,
+      "step": 46068
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.528403271035885e-06,
+      "loss": 0.2091,
+      "step": 46069
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.528067435729519e-06,
+      "loss": 0.282,
+      "step": 46070
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.527731600423153e-06,
+      "loss": 0.3451,
+      "step": 46071
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.527395765116787e-06,
+      "loss": 0.213,
+      "step": 46072
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.527059929810421e-06,
+      "loss": 0.3133,
+      "step": 46073
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5267240945040555e-06,
+      "loss": 0.2531,
+      "step": 46074
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.52638825919769e-06,
+      "loss": 0.3139,
+      "step": 46075
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.526052423891324e-06,
+      "loss": 0.1911,
+      "step": 46076
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.525716588584959e-06,
+      "loss": 0.1538,
+      "step": 46077
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.525380753278592e-06,
+      "loss": 0.2534,
+      "step": 46078
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.525044917972227e-06,
+      "loss": 0.2688,
+      "step": 46079
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.524709082665862e-06,
+      "loss": 0.2969,
+      "step": 46080
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5243732473594955e-06,
+      "loss": 0.167,
+      "step": 46081
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.524037412053129e-06,
+      "loss": 0.3103,
+      "step": 46082
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.523701576746764e-06,
+      "loss": 0.2225,
+      "step": 46083
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.523365741440398e-06,
+      "loss": 0.2753,
+      "step": 46084
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.523029906134032e-06,
+      "loss": 0.3386,
+      "step": 46085
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.522694070827666e-06,
+      "loss": 0.1951,
+      "step": 46086
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.522358235521301e-06,
+      "loss": 0.3045,
+      "step": 46087
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.522022400214935e-06,
+      "loss": 0.2083,
+      "step": 46088
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.521686564908569e-06,
+      "loss": 0.1979,
+      "step": 46089
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.521350729602204e-06,
+      "loss": 0.2366,
+      "step": 46090
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.521014894295838e-06,
+      "loss": 0.3488,
+      "step": 46091
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5206790589894715e-06,
+      "loss": 0.2578,
+      "step": 46092
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.520343223683106e-06,
+      "loss": 0.2689,
+      "step": 46093
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.52000738837674e-06,
+      "loss": 0.275,
+      "step": 46094
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.519671553070375e-06,
+      "loss": 0.321,
+      "step": 46095
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.519335717764009e-06,
+      "loss": 0.2844,
+      "step": 46096
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.518999882457643e-06,
+      "loss": 0.3843,
+      "step": 46097
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.518664047151277e-06,
+      "loss": 0.4482,
+      "step": 46098
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5183282118449115e-06,
+      "loss": 0.4335,
+      "step": 46099
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.517992376538546e-06,
+      "loss": 0.3234,
+      "step": 46100
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.51765654123218e-06,
+      "loss": 0.2522,
+      "step": 46101
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.517320705925815e-06,
+      "loss": 0.2623,
+      "step": 46102
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.516984870619448e-06,
+      "loss": 0.2727,
+      "step": 46103
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.516649035313083e-06,
+      "loss": 0.196,
+      "step": 46104
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.516313200006717e-06,
+      "loss": 0.1843,
+      "step": 46105
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5159773647003515e-06,
+      "loss": 0.2788,
+      "step": 46106
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.515641529393985e-06,
+      "loss": 0.314,
+      "step": 46107
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.51530569408762e-06,
+      "loss": 0.3134,
+      "step": 46108
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.514969858781254e-06,
+      "loss": 0.1994,
+      "step": 46109
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.514634023474888e-06,
+      "loss": 0.2189,
+      "step": 46110
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.514298188168523e-06,
+      "loss": 0.2693,
+      "step": 46111
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.513962352862157e-06,
+      "loss": 0.252,
+      "step": 46112
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.513626517555791e-06,
+      "loss": 0.3784,
+      "step": 46113
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.513290682249425e-06,
+      "loss": 0.3558,
+      "step": 46114
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.51295484694306e-06,
+      "loss": 0.242,
+      "step": 46115
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.512619011636694e-06,
+      "loss": 0.1823,
+      "step": 46116
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.512283176330328e-06,
+      "loss": 0.3057,
+      "step": 46117
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.511947341023962e-06,
+      "loss": 0.1788,
+      "step": 46118
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.511611505717596e-06,
+      "loss": 0.2182,
+      "step": 46119
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.511275670411231e-06,
+      "loss": 0.3921,
+      "step": 46120
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.510939835104865e-06,
+      "loss": 0.3729,
+      "step": 46121
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.510603999798499e-06,
+      "loss": 0.2413,
+      "step": 46122
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.510268164492134e-06,
+      "loss": 0.1882,
+      "step": 46123
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5099323291857675e-06,
+      "loss": 0.3007,
+      "step": 46124
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.509596493879402e-06,
+      "loss": 0.2133,
+      "step": 46125
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.509260658573037e-06,
+      "loss": 0.306,
+      "step": 46126
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.508924823266671e-06,
+      "loss": 0.2773,
+      "step": 46127
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5085889879603044e-06,
+      "loss": 0.2788,
+      "step": 46128
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.508253152653939e-06,
+      "loss": 0.181,
+      "step": 46129
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.507917317347573e-06,
+      "loss": 0.1549,
+      "step": 46130
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5075814820412075e-06,
+      "loss": 0.3994,
+      "step": 46131
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.507245646734841e-06,
+      "loss": 0.3365,
+      "step": 46132
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.506909811428476e-06,
+      "loss": 0.2815,
+      "step": 46133
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.50657397612211e-06,
+      "loss": 0.3156,
+      "step": 46134
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.506238140815744e-06,
+      "loss": 0.3406,
+      "step": 46135
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.505902305509379e-06,
+      "loss": 0.1343,
+      "step": 46136
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.505566470203013e-06,
+      "loss": 0.2237,
+      "step": 46137
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.505230634896647e-06,
+      "loss": 0.1929,
+      "step": 46138
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.504894799590281e-06,
+      "loss": 0.2317,
+      "step": 46139
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.504558964283915e-06,
+      "loss": 0.2576,
+      "step": 46140
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.50422312897755e-06,
+      "loss": 0.2058,
+      "step": 46141
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.503887293671184e-06,
+      "loss": 0.3774,
+      "step": 46142
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.503551458364818e-06,
+      "loss": 0.1661,
+      "step": 46143
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.503215623058452e-06,
+      "loss": 0.3846,
+      "step": 46144
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.502879787752087e-06,
+      "loss": 0.2426,
+      "step": 46145
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.502543952445721e-06,
+      "loss": 0.1591,
+      "step": 46146
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.502208117139355e-06,
+      "loss": 0.309,
+      "step": 46147
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.50187228183299e-06,
+      "loss": 0.1921,
+      "step": 46148
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5015364465266236e-06,
+      "loss": 0.3405,
+      "step": 46149
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.501200611220257e-06,
+      "loss": 0.4187,
+      "step": 46150
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.500864775913892e-06,
+      "loss": 0.2993,
+      "step": 46151
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.500528940607527e-06,
+      "loss": 0.2727,
+      "step": 46152
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 4.5001931053011604e-06,
+      "loss": 0.2237,
+      "step": 46153
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.499857269994795e-06,
+      "loss": 0.2632,
+      "step": 46154
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.499521434688429e-06,
+      "loss": 0.1848,
+      "step": 46155
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4991855993820635e-06,
+      "loss": 0.441,
+      "step": 46156
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.498849764075698e-06,
+      "loss": 0.2368,
+      "step": 46157
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.498513928769332e-06,
+      "loss": 0.1968,
+      "step": 46158
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.498178093462966e-06,
+      "loss": 0.2452,
+      "step": 46159
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4978422581566004e-06,
+      "loss": 0.3859,
+      "step": 46160
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.497506422850234e-06,
+      "loss": 0.1905,
+      "step": 46161
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.497170587543869e-06,
+      "loss": 0.2719,
+      "step": 46162
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4968347522375035e-06,
+      "loss": 0.3873,
+      "step": 46163
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.496498916931137e-06,
+      "loss": 0.3107,
+      "step": 46164
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.496163081624771e-06,
+      "loss": 0.1933,
+      "step": 46165
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.495827246318406e-06,
+      "loss": 0.1496,
+      "step": 46166
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4954914110120404e-06,
+      "loss": 0.2885,
+      "step": 46167
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.495155575705674e-06,
+      "loss": 0.3039,
+      "step": 46168
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.494819740399309e-06,
+      "loss": 0.15,
+      "step": 46169
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.494483905092943e-06,
+      "loss": 0.3595,
+      "step": 46170
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.494148069786577e-06,
+      "loss": 0.1795,
+      "step": 46171
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.493812234480211e-06,
+      "loss": 0.2376,
+      "step": 46172
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.493476399173846e-06,
+      "loss": 0.3267,
+      "step": 46173
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4931405638674796e-06,
+      "loss": 0.2961,
+      "step": 46174
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.492804728561114e-06,
+      "loss": 0.3603,
+      "step": 46175
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.492468893254748e-06,
+      "loss": 0.2768,
+      "step": 46176
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.492133057948383e-06,
+      "loss": 0.2783,
+      "step": 46177
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4917972226420165e-06,
+      "loss": 0.1748,
+      "step": 46178
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.491461387335651e-06,
+      "loss": 0.2583,
+      "step": 46179
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.491125552029285e-06,
+      "loss": 0.1516,
+      "step": 46180
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4907897167229196e-06,
+      "loss": 0.2837,
+      "step": 46181
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.490453881416554e-06,
+      "loss": 0.2557,
+      "step": 46182
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.490118046110188e-06,
+      "loss": 0.2668,
+      "step": 46183
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.489782210803822e-06,
+      "loss": 0.4117,
+      "step": 46184
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4894463754974565e-06,
+      "loss": 0.5278,
+      "step": 46185
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.48911054019109e-06,
+      "loss": 0.4437,
+      "step": 46186
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.488774704884725e-06,
+      "loss": 0.2529,
+      "step": 46187
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4884388695783596e-06,
+      "loss": 0.2134,
+      "step": 46188
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.488103034271993e-06,
+      "loss": 0.2272,
+      "step": 46189
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.487767198965627e-06,
+      "loss": 0.223,
+      "step": 46190
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.487431363659262e-06,
+      "loss": 0.1916,
+      "step": 46191
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4870955283528964e-06,
+      "loss": 0.1504,
+      "step": 46192
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.48675969304653e-06,
+      "loss": 0.3823,
+      "step": 46193
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.486423857740165e-06,
+      "loss": 0.231,
+      "step": 46194
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.486088022433799e-06,
+      "loss": 0.3049,
+      "step": 46195
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4857521871274325e-06,
+      "loss": 0.205,
+      "step": 46196
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.485416351821067e-06,
+      "loss": 0.2677,
+      "step": 46197
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.485080516514702e-06,
+      "loss": 0.3659,
+      "step": 46198
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.484744681208336e-06,
+      "loss": 0.2791,
+      "step": 46199
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.48440884590197e-06,
+      "loss": 0.2458,
+      "step": 46200
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.484073010595604e-06,
+      "loss": 0.2899,
+      "step": 46201
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.483737175289239e-06,
+      "loss": 0.244,
+      "step": 46202
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.483401339982873e-06,
+      "loss": 0.281,
+      "step": 46203
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.483065504676507e-06,
+      "loss": 0.2756,
+      "step": 46204
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.482729669370141e-06,
+      "loss": 0.1792,
+      "step": 46205
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.482393834063776e-06,
+      "loss": 0.1631,
+      "step": 46206
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.482057998757409e-06,
+      "loss": 0.2582,
+      "step": 46207
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.481722163451044e-06,
+      "loss": 0.183,
+      "step": 46208
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.481386328144679e-06,
+      "loss": 0.3809,
+      "step": 46209
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4810504928383125e-06,
+      "loss": 0.3051,
+      "step": 46210
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.480714657531946e-06,
+      "loss": 0.3602,
+      "step": 46211
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.480378822225581e-06,
+      "loss": 0.3034,
+      "step": 46212
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4800429869192156e-06,
+      "loss": 0.2348,
+      "step": 46213
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.479707151612849e-06,
+      "loss": 0.2426,
+      "step": 46214
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.479371316306484e-06,
+      "loss": 0.199,
+      "step": 46215
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.479035481000118e-06,
+      "loss": 0.1908,
+      "step": 46216
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.478699645693752e-06,
+      "loss": 0.1743,
+      "step": 46217
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.478363810387386e-06,
+      "loss": 0.4134,
+      "step": 46218
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.478027975081021e-06,
+      "loss": 0.3765,
+      "step": 46219
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.477692139774655e-06,
+      "loss": 0.2719,
+      "step": 46220
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.477356304468289e-06,
+      "loss": 0.4438,
+      "step": 46221
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.477020469161923e-06,
+      "loss": 0.2476,
+      "step": 46222
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.476684633855558e-06,
+      "loss": 0.1799,
+      "step": 46223
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.476348798549192e-06,
+      "loss": 0.3785,
+      "step": 46224
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.476012963242826e-06,
+      "loss": 0.1627,
+      "step": 46225
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.47567712793646e-06,
+      "loss": 0.4628,
+      "step": 46226
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.475341292630095e-06,
+      "loss": 0.3855,
+      "step": 46227
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4750054573237285e-06,
+      "loss": 0.2727,
+      "step": 46228
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.474669622017363e-06,
+      "loss": 0.2297,
+      "step": 46229
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.474333786710997e-06,
+      "loss": 0.3831,
+      "step": 46230
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.473997951404632e-06,
+      "loss": 0.2288,
+      "step": 46231
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.473662116098265e-06,
+      "loss": 0.2215,
+      "step": 46232
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4733262807919e-06,
+      "loss": 0.3684,
+      "step": 46233
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.472990445485535e-06,
+      "loss": 0.2088,
+      "step": 46234
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4726546101791685e-06,
+      "loss": 0.2407,
+      "step": 46235
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.472318774872802e-06,
+      "loss": 0.48,
+      "step": 46236
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.471982939566437e-06,
+      "loss": 0.3008,
+      "step": 46237
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.471647104260072e-06,
+      "loss": 0.3378,
+      "step": 46238
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.471311268953705e-06,
+      "loss": 0.3048,
+      "step": 46239
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.47097543364734e-06,
+      "loss": 0.2058,
+      "step": 46240
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.470639598340974e-06,
+      "loss": 0.4167,
+      "step": 46241
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.470303763034608e-06,
+      "loss": 0.1495,
+      "step": 46242
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.469967927728242e-06,
+      "loss": 0.436,
+      "step": 46243
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.469632092421877e-06,
+      "loss": 0.2719,
+      "step": 46244
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.469296257115511e-06,
+      "loss": 0.2987,
+      "step": 46245
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.468960421809145e-06,
+      "loss": 0.2359,
+      "step": 46246
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.468624586502779e-06,
+      "loss": 0.2404,
+      "step": 46247
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.468288751196414e-06,
+      "loss": 0.3168,
+      "step": 46248
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4679529158900485e-06,
+      "loss": 0.2135,
+      "step": 46249
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.467617080583682e-06,
+      "loss": 0.2696,
+      "step": 46250
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.467281245277316e-06,
+      "loss": 0.2848,
+      "step": 46251
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.466945409970951e-06,
+      "loss": 0.3842,
+      "step": 46252
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4666095746645845e-06,
+      "loss": 0.1918,
+      "step": 46253
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.466273739358219e-06,
+      "loss": 0.1891,
+      "step": 46254
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.465937904051854e-06,
+      "loss": 0.3213,
+      "step": 46255
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.465602068745488e-06,
+      "loss": 0.3685,
+      "step": 46256
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.465266233439121e-06,
+      "loss": 0.2151,
+      "step": 46257
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.464930398132756e-06,
+      "loss": 0.2775,
+      "step": 46258
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.464594562826391e-06,
+      "loss": 0.2716,
+      "step": 46259
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4642587275200245e-06,
+      "loss": 0.3072,
+      "step": 46260
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.463922892213659e-06,
+      "loss": 0.2737,
+      "step": 46261
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.463587056907293e-06,
+      "loss": 0.3952,
+      "step": 46262
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.463251221600927e-06,
+      "loss": 0.1052,
+      "step": 46263
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.462915386294561e-06,
+      "loss": 0.223,
+      "step": 46264
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.462579550988196e-06,
+      "loss": 0.3765,
+      "step": 46265
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.46224371568183e-06,
+      "loss": 0.3006,
+      "step": 46266
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.461907880375464e-06,
+      "loss": 0.2612,
+      "step": 46267
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.461572045069098e-06,
+      "loss": 0.177,
+      "step": 46268
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.461236209762733e-06,
+      "loss": 0.1594,
+      "step": 46269
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.460900374456367e-06,
+      "loss": 0.3921,
+      "step": 46270
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.460564539150001e-06,
+      "loss": 0.2762,
+      "step": 46271
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.460228703843635e-06,
+      "loss": 0.2257,
+      "step": 46272
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.45989286853727e-06,
+      "loss": 0.4853,
+      "step": 46273
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.459557033230904e-06,
+      "loss": 0.2264,
+      "step": 46274
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.459221197924538e-06,
+      "loss": 0.2493,
+      "step": 46275
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.458885362618172e-06,
+      "loss": 0.188,
+      "step": 46276
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.458549527311807e-06,
+      "loss": 0.2294,
+      "step": 46277
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4582136920054405e-06,
+      "loss": 0.3717,
+      "step": 46278
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.457877856699075e-06,
+      "loss": 0.2329,
+      "step": 46279
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.45754202139271e-06,
+      "loss": 0.2314,
+      "step": 46280
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.457206186086344e-06,
+      "loss": 0.2446,
+      "step": 46281
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4568703507799774e-06,
+      "loss": 0.2269,
+      "step": 46282
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.456534515473612e-06,
+      "loss": 0.2276,
+      "step": 46283
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.456198680167247e-06,
+      "loss": 0.4021,
+      "step": 46284
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4558628448608805e-06,
+      "loss": 0.2274,
+      "step": 46285
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.455527009554515e-06,
+      "loss": 0.161,
+      "step": 46286
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.455191174248149e-06,
+      "loss": 0.1939,
+      "step": 46287
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.454855338941783e-06,
+      "loss": 0.2433,
+      "step": 46288
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4545195036354174e-06,
+      "loss": 0.3133,
+      "step": 46289
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.454183668329052e-06,
+      "loss": 0.6274,
+      "step": 46290
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.453847833022686e-06,
+      "loss": 0.203,
+      "step": 46291
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4535119977163205e-06,
+      "loss": 0.21,
+      "step": 46292
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.453176162409954e-06,
+      "loss": 0.2821,
+      "step": 46293
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.452840327103589e-06,
+      "loss": 0.1432,
+      "step": 46294
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.452504491797224e-06,
+      "loss": 0.1879,
+      "step": 46295
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4521686564908574e-06,
+      "loss": 0.2032,
+      "step": 46296
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.451832821184491e-06,
+      "loss": 0.2779,
+      "step": 46297
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.451496985878126e-06,
+      "loss": 0.5171,
+      "step": 46298
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.45116115057176e-06,
+      "loss": 0.2346,
+      "step": 46299
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.450825315265394e-06,
+      "loss": 0.2106,
+      "step": 46300
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.450489479959029e-06,
+      "loss": 0.3011,
+      "step": 46301
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.450153644652663e-06,
+      "loss": 0.3057,
+      "step": 46302
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4498178093462966e-06,
+      "loss": 0.1881,
+      "step": 46303
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.449481974039931e-06,
+      "loss": 0.1972,
+      "step": 46304
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.449146138733566e-06,
+      "loss": 0.202,
+      "step": 46305
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4488103034272e-06,
+      "loss": 0.2347,
+      "step": 46306
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.448474468120834e-06,
+      "loss": 0.5159,
+      "step": 46307
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.448138632814468e-06,
+      "loss": 0.2739,
+      "step": 46308
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.447802797508102e-06,
+      "loss": 0.266,
+      "step": 46309
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4474669622017366e-06,
+      "loss": 0.4232,
+      "step": 46310
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.447131126895371e-06,
+      "loss": 0.2937,
+      "step": 46311
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.446795291589005e-06,
+      "loss": 0.3428,
+      "step": 46312
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.446459456282639e-06,
+      "loss": 0.2393,
+      "step": 46313
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4461236209762735e-06,
+      "loss": 0.3021,
+      "step": 46314
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.445787785669908e-06,
+      "loss": 0.3831,
+      "step": 46315
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.445451950363542e-06,
+      "loss": 0.2644,
+      "step": 46316
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4451161150571765e-06,
+      "loss": 0.1682,
+      "step": 46317
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.44478027975081e-06,
+      "loss": 0.3353,
+      "step": 46318
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.444444444444444e-06,
+      "loss": 0.542,
+      "step": 46319
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.444108609138079e-06,
+      "loss": 0.256,
+      "step": 46320
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4437727738317134e-06,
+      "loss": 0.185,
+      "step": 46321
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.443436938525347e-06,
+      "loss": 0.304,
+      "step": 46322
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.443101103218982e-06,
+      "loss": 0.1869,
+      "step": 46323
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.442765267912616e-06,
+      "loss": 0.198,
+      "step": 46324
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.44242943260625e-06,
+      "loss": 0.1536,
+      "step": 46325
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.442093597299885e-06,
+      "loss": 0.1739,
+      "step": 46326
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.441757761993519e-06,
+      "loss": 0.4797,
+      "step": 46327
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.441421926687153e-06,
+      "loss": 0.249,
+      "step": 46328
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.441086091380787e-06,
+      "loss": 0.2424,
+      "step": 46329
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.440750256074421e-06,
+      "loss": 0.2908,
+      "step": 46330
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.440414420768056e-06,
+      "loss": 0.2545,
+      "step": 46331
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.44007858546169e-06,
+      "loss": 0.1482,
+      "step": 46332
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.439742750155324e-06,
+      "loss": 0.3214,
+      "step": 46333
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.439406914848958e-06,
+      "loss": 0.3049,
+      "step": 46334
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.439071079542593e-06,
+      "loss": 0.1951,
+      "step": 46335
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.438735244236227e-06,
+      "loss": 0.2693,
+      "step": 46336
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.438399408929861e-06,
+      "loss": 0.1486,
+      "step": 46337
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.438063573623496e-06,
+      "loss": 0.2294,
+      "step": 46338
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4377277383171295e-06,
+      "loss": 0.3151,
+      "step": 46339
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.437391903010764e-06,
+      "loss": 0.2807,
+      "step": 46340
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.437056067704398e-06,
+      "loss": 0.3192,
+      "step": 46341
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4367202323980326e-06,
+      "loss": 0.3493,
+      "step": 46342
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.436384397091666e-06,
+      "loss": 0.3668,
+      "step": 46343
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.436048561785301e-06,
+      "loss": 0.4752,
+      "step": 46344
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.435712726478935e-06,
+      "loss": 0.3265,
+      "step": 46345
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.4353768911725695e-06,
+      "loss": 0.2137,
+      "step": 46346
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.435041055866204e-06,
+      "loss": 0.345,
+      "step": 46347
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.434705220559838e-06,
+      "loss": 0.2034,
+      "step": 46348
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.434369385253472e-06,
+      "loss": 0.2121,
+      "step": 46349
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.434033549947106e-06,
+      "loss": 0.2079,
+      "step": 46350
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.433697714640741e-06,
+      "loss": 0.1945,
+      "step": 46351
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 4.433361879334375e-06,
+      "loss": 0.2757,
+      "step": 46352
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4330260440280095e-06,
+      "loss": 0.1854,
+      "step": 46353
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.432690208721643e-06,
+      "loss": 0.5148,
+      "step": 46354
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.432354373415277e-06,
+      "loss": 0.2408,
+      "step": 46355
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.432018538108912e-06,
+      "loss": 0.3027,
+      "step": 46356
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.431682702802546e-06,
+      "loss": 0.338,
+      "step": 46357
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.43134686749618e-06,
+      "loss": 0.272,
+      "step": 46358
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.431011032189814e-06,
+      "loss": 0.2716,
+      "step": 46359
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.430675196883449e-06,
+      "loss": 0.3323,
+      "step": 46360
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.430339361577083e-06,
+      "loss": 0.267,
+      "step": 46361
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.430003526270717e-06,
+      "loss": 0.4008,
+      "step": 46362
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.429667690964352e-06,
+      "loss": 0.2235,
+      "step": 46363
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4293318556579855e-06,
+      "loss": 0.2923,
+      "step": 46364
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.428996020351619e-06,
+      "loss": 0.2296,
+      "step": 46365
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.428660185045254e-06,
+      "loss": 0.277,
+      "step": 46366
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.428324349738889e-06,
+      "loss": 0.0713,
+      "step": 46367
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.427988514432522e-06,
+      "loss": 0.4104,
+      "step": 46368
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.427652679126157e-06,
+      "loss": 0.1747,
+      "step": 46369
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.427316843819791e-06,
+      "loss": 0.2916,
+      "step": 46370
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4269810085134255e-06,
+      "loss": 0.2713,
+      "step": 46371
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.42664517320706e-06,
+      "loss": 0.2581,
+      "step": 46372
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.426309337900694e-06,
+      "loss": 0.3278,
+      "step": 46373
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.425973502594328e-06,
+      "loss": 0.1834,
+      "step": 46374
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.425637667287962e-06,
+      "loss": 0.2584,
+      "step": 46375
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.425301831981596e-06,
+      "loss": 0.274,
+      "step": 46376
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.424965996675231e-06,
+      "loss": 0.2977,
+      "step": 46377
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4246301613688655e-06,
+      "loss": 0.3296,
+      "step": 46378
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.424294326062499e-06,
+      "loss": 0.1898,
+      "step": 46379
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.423958490756133e-06,
+      "loss": 0.1151,
+      "step": 46380
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.423622655449768e-06,
+      "loss": 0.4369,
+      "step": 46381
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.423286820143402e-06,
+      "loss": 0.4573,
+      "step": 46382
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.422950984837036e-06,
+      "loss": 0.1975,
+      "step": 46383
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.422615149530671e-06,
+      "loss": 0.1879,
+      "step": 46384
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.422279314224305e-06,
+      "loss": 0.4311,
+      "step": 46385
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.421943478917939e-06,
+      "loss": 0.2741,
+      "step": 46386
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.421607643611573e-06,
+      "loss": 0.3103,
+      "step": 46387
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.421271808305208e-06,
+      "loss": 0.1939,
+      "step": 46388
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4209359729988415e-06,
+      "loss": 0.2688,
+      "step": 46389
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.420600137692476e-06,
+      "loss": 0.3986,
+      "step": 46390
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.42026430238611e-06,
+      "loss": 0.3858,
+      "step": 46391
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.419928467079745e-06,
+      "loss": 0.279,
+      "step": 46392
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.419592631773379e-06,
+      "loss": 0.3264,
+      "step": 46393
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.419256796467013e-06,
+      "loss": 0.2685,
+      "step": 46394
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.418920961160647e-06,
+      "loss": 0.3473,
+      "step": 46395
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4185851258542815e-06,
+      "loss": 0.425,
+      "step": 46396
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.418249290547916e-06,
+      "loss": 0.235,
+      "step": 46397
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.41791345524155e-06,
+      "loss": 0.2011,
+      "step": 46398
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.417577619935185e-06,
+      "loss": 0.3244,
+      "step": 46399
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.417241784628818e-06,
+      "loss": 0.2887,
+      "step": 46400
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.416905949322452e-06,
+      "loss": 0.194,
+      "step": 46401
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.416570114016087e-06,
+      "loss": 0.2836,
+      "step": 46402
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4162342787097215e-06,
+      "loss": 0.178,
+      "step": 46403
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.415898443403355e-06,
+      "loss": 0.2055,
+      "step": 46404
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.415562608096989e-06,
+      "loss": 0.3915,
+      "step": 46405
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.415226772790624e-06,
+      "loss": 0.2907,
+      "step": 46406
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.414890937484258e-06,
+      "loss": 0.2567,
+      "step": 46407
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.414555102177892e-06,
+      "loss": 0.2103,
+      "step": 46408
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.414219266871527e-06,
+      "loss": 0.2657,
+      "step": 46409
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.413883431565161e-06,
+      "loss": 0.2431,
+      "step": 46410
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4135475962587944e-06,
+      "loss": 0.3171,
+      "step": 46411
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.413211760952429e-06,
+      "loss": 0.1564,
+      "step": 46412
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.412875925646064e-06,
+      "loss": 0.2101,
+      "step": 46413
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4125400903396975e-06,
+      "loss": 0.2269,
+      "step": 46414
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.412204255033332e-06,
+      "loss": 0.4094,
+      "step": 46415
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.411868419726966e-06,
+      "loss": 0.2707,
+      "step": 46416
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.411532584420601e-06,
+      "loss": 0.2831,
+      "step": 46417
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.411196749114235e-06,
+      "loss": 0.2226,
+      "step": 46418
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.410860913807869e-06,
+      "loss": 0.2972,
+      "step": 46419
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.410525078501503e-06,
+      "loss": 0.3504,
+      "step": 46420
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4101892431951375e-06,
+      "loss": 0.2839,
+      "step": 46421
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.409853407888771e-06,
+      "loss": 0.2291,
+      "step": 46422
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.409517572582406e-06,
+      "loss": 0.2854,
+      "step": 46423
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.409181737276041e-06,
+      "loss": 0.2876,
+      "step": 46424
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.408845901969674e-06,
+      "loss": 0.196,
+      "step": 46425
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.408510066663308e-06,
+      "loss": 0.3338,
+      "step": 46426
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.408174231356943e-06,
+      "loss": 0.3952,
+      "step": 46427
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4078383960505775e-06,
+      "loss": 0.2039,
+      "step": 46428
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.407502560744211e-06,
+      "loss": 0.2131,
+      "step": 46429
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.407166725437846e-06,
+      "loss": 0.2229,
+      "step": 46430
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.40683089013148e-06,
+      "loss": 0.2511,
+      "step": 46431
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4064950548251136e-06,
+      "loss": 0.3004,
+      "step": 46432
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.406159219518748e-06,
+      "loss": 0.3049,
+      "step": 46433
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.405823384212383e-06,
+      "loss": 0.3811,
+      "step": 46434
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.405487548906017e-06,
+      "loss": 0.3282,
+      "step": 46435
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.405151713599651e-06,
+      "loss": 0.2566,
+      "step": 46436
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.404815878293285e-06,
+      "loss": 0.2463,
+      "step": 46437
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.40448004298692e-06,
+      "loss": 0.2349,
+      "step": 46438
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.404144207680554e-06,
+      "loss": 0.1888,
+      "step": 46439
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.403808372374188e-06,
+      "loss": 0.3421,
+      "step": 46440
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.403472537067822e-06,
+      "loss": 0.2,
+      "step": 46441
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.403136701761457e-06,
+      "loss": 0.2154,
+      "step": 46442
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4028008664550905e-06,
+      "loss": 0.1934,
+      "step": 46443
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.402465031148725e-06,
+      "loss": 0.0838,
+      "step": 46444
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.40212919584236e-06,
+      "loss": 0.1649,
+      "step": 46445
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4017933605359935e-06,
+      "loss": 0.2102,
+      "step": 46446
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.401457525229627e-06,
+      "loss": 0.1655,
+      "step": 46447
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.401121689923262e-06,
+      "loss": 0.3797,
+      "step": 46448
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.400785854616897e-06,
+      "loss": 0.2962,
+      "step": 46449
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.4004500193105304e-06,
+      "loss": 0.2727,
+      "step": 46450
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.400114184004164e-06,
+      "loss": 0.2895,
+      "step": 46451
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.399778348697799e-06,
+      "loss": 0.1726,
+      "step": 46452
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.3994425133914335e-06,
+      "loss": 0.1554,
+      "step": 46453
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.399106678085067e-06,
+      "loss": 0.4541,
+      "step": 46454
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.398770842778702e-06,
+      "loss": 0.2583,
+      "step": 46455
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.398435007472336e-06,
+      "loss": 0.2211,
+      "step": 46456
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.39809917216597e-06,
+      "loss": 0.3624,
+      "step": 46457
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.397763336859604e-06,
+      "loss": 0.4012,
+      "step": 46458
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.397427501553239e-06,
+      "loss": 0.2432,
+      "step": 46459
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.397091666246873e-06,
+      "loss": 0.3393,
+      "step": 46460
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.396755830940507e-06,
+      "loss": 0.2519,
+      "step": 46461
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.396419995634141e-06,
+      "loss": 0.3292,
+      "step": 46462
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.396084160327776e-06,
+      "loss": 0.2229,
+      "step": 46463
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.39574832502141e-06,
+      "loss": 0.3969,
+      "step": 46464
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.395412489715044e-06,
+      "loss": 0.335,
+      "step": 46465
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.395076654408678e-06,
+      "loss": 0.3448,
+      "step": 46466
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.394740819102313e-06,
+      "loss": 0.2873,
+      "step": 46467
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.3944049837959465e-06,
+      "loss": 0.2864,
+      "step": 46468
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.394069148489581e-06,
+      "loss": 0.2422,
+      "step": 46469
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.393733313183216e-06,
+      "loss": 0.1634,
+      "step": 46470
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.3933974778768496e-06,
+      "loss": 0.2049,
+      "step": 46471
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.393061642570483e-06,
+      "loss": 0.3312,
+      "step": 46472
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.392725807264118e-06,
+      "loss": 0.2616,
+      "step": 46473
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.392389971957753e-06,
+      "loss": 0.3101,
+      "step": 46474
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.3920541366513865e-06,
+      "loss": 0.3971,
+      "step": 46475
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.391718301345021e-06,
+      "loss": 0.214,
+      "step": 46476
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.391382466038655e-06,
+      "loss": 0.2738,
+      "step": 46477
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.391046630732289e-06,
+      "loss": 0.4354,
+      "step": 46478
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.390710795425923e-06,
+      "loss": 0.3347,
+      "step": 46479
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.390374960119558e-06,
+      "loss": 0.1724,
+      "step": 46480
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.390039124813192e-06,
+      "loss": 0.2506,
+      "step": 46481
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.3897032895068265e-06,
+      "loss": 0.2631,
+      "step": 46482
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.38936745420046e-06,
+      "loss": 0.1857,
+      "step": 46483
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.389031618894095e-06,
+      "loss": 0.3013,
+      "step": 46484
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.3886957835877295e-06,
+      "loss": 0.2164,
+      "step": 46485
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.388359948281363e-06,
+      "loss": 0.3115,
+      "step": 46486
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.388024112974997e-06,
+      "loss": 0.2379,
+      "step": 46487
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.387688277668632e-06,
+      "loss": 0.2996,
+      "step": 46488
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.387352442362266e-06,
+      "loss": 0.3059,
+      "step": 46489
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.3870166070559e-06,
+      "loss": 0.3749,
+      "step": 46490
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.386680771749535e-06,
+      "loss": 0.2998,
+      "step": 46491
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.386344936443169e-06,
+      "loss": 0.1666,
+      "step": 46492
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.3860091011368025e-06,
+      "loss": 0.3176,
+      "step": 46493
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.385673265830437e-06,
+      "loss": 0.2843,
+      "step": 46494
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.385337430524072e-06,
+      "loss": 0.2495,
+      "step": 46495
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.385001595217706e-06,
+      "loss": 0.2615,
+      "step": 46496
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.384665759911339e-06,
+      "loss": 0.3379,
+      "step": 46497
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.384329924604974e-06,
+      "loss": 0.2101,
+      "step": 46498
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.383994089298609e-06,
+      "loss": 0.2291,
+      "step": 46499
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.3836582539922425e-06,
+      "loss": 0.3776,
+      "step": 46500
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.383322418685877e-06,
+      "loss": 0.3017,
+      "step": 46501
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.382986583379511e-06,
+      "loss": 0.1455,
+      "step": 46502
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.382650748073145e-06,
+      "loss": 0.1888,
+      "step": 46503
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.382314912766779e-06,
+      "loss": 0.2306,
+      "step": 46504
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.381979077460414e-06,
+      "loss": 0.3373,
+      "step": 46505
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.381643242154048e-06,
+      "loss": 0.3283,
+      "step": 46506
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.3813074068476825e-06,
+      "loss": 0.2804,
+      "step": 46507
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.380971571541316e-06,
+      "loss": 0.2322,
+      "step": 46508
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.380635736234951e-06,
+      "loss": 0.1403,
+      "step": 46509
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.3802999009285856e-06,
+      "loss": 0.2488,
+      "step": 46510
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.379964065622219e-06,
+      "loss": 0.215,
+      "step": 46511
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.379628230315853e-06,
+      "loss": 0.3542,
+      "step": 46512
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.379292395009488e-06,
+      "loss": 0.2589,
+      "step": 46513
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.378956559703122e-06,
+      "loss": 0.0898,
+      "step": 46514
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.378620724396756e-06,
+      "loss": 0.3368,
+      "step": 46515
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.378284889090391e-06,
+      "loss": 0.273,
+      "step": 46516
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.377949053784025e-06,
+      "loss": 0.258,
+      "step": 46517
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.3776132184776585e-06,
+      "loss": 0.2222,
+      "step": 46518
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.377277383171293e-06,
+      "loss": 0.2556,
+      "step": 46519
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.376941547864928e-06,
+      "loss": 0.2543,
+      "step": 46520
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.376605712558562e-06,
+      "loss": 0.3124,
+      "step": 46521
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.376269877252196e-06,
+      "loss": 0.2176,
+      "step": 46522
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.37593404194583e-06,
+      "loss": 0.2597,
+      "step": 46523
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.375598206639464e-06,
+      "loss": 0.285,
+      "step": 46524
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.3752623713330985e-06,
+      "loss": 0.3234,
+      "step": 46525
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.374926536026733e-06,
+      "loss": 0.2451,
+      "step": 46526
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.374590700720367e-06,
+      "loss": 0.1779,
+      "step": 46527
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.374254865414002e-06,
+      "loss": 0.2307,
+      "step": 46528
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.373919030107635e-06,
+      "loss": 0.2316,
+      "step": 46529
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.37358319480127e-06,
+      "loss": 0.3632,
+      "step": 46530
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.373247359494905e-06,
+      "loss": 0.1788,
+      "step": 46531
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.3729115241885385e-06,
+      "loss": 0.2891,
+      "step": 46532
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.372575688882172e-06,
+      "loss": 0.2823,
+      "step": 46533
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.372239853575807e-06,
+      "loss": 0.4283,
+      "step": 46534
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.371904018269441e-06,
+      "loss": 0.1922,
+      "step": 46535
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.371568182963075e-06,
+      "loss": 0.1677,
+      "step": 46536
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.37123234765671e-06,
+      "loss": 0.3603,
+      "step": 46537
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.370896512350344e-06,
+      "loss": 0.2788,
+      "step": 46538
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.370560677043978e-06,
+      "loss": 0.281,
+      "step": 46539
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.370224841737612e-06,
+      "loss": 0.1849,
+      "step": 46540
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.369889006431247e-06,
+      "loss": 0.1707,
+      "step": 46541
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.369553171124881e-06,
+      "loss": 0.1807,
+      "step": 46542
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.3692173358185145e-06,
+      "loss": 0.1763,
+      "step": 46543
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.368881500512149e-06,
+      "loss": 0.199,
+      "step": 46544
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.368545665205783e-06,
+      "loss": 0.3115,
+      "step": 46545
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.368209829899418e-06,
+      "loss": 0.2895,
+      "step": 46546
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.367873994593052e-06,
+      "loss": 0.2383,
+      "step": 46547
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.367538159286686e-06,
+      "loss": 0.2438,
+      "step": 46548
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.36720232398032e-06,
+      "loss": 0.2411,
+      "step": 46549
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 4.3668664886739545e-06,
+      "loss": 0.2899,
+      "step": 46550
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.366530653367589e-06,
+      "loss": 0.3714,
+      "step": 46551
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.366194818061223e-06,
+      "loss": 0.1834,
+      "step": 46552
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.365858982754858e-06,
+      "loss": 0.2991,
+      "step": 46553
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.365523147448491e-06,
+      "loss": 0.2451,
+      "step": 46554
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.365187312142126e-06,
+      "loss": 0.3265,
+      "step": 46555
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.36485147683576e-06,
+      "loss": 0.208,
+      "step": 46556
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3645156415293945e-06,
+      "loss": 0.1607,
+      "step": 46557
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.364179806223028e-06,
+      "loss": 0.3032,
+      "step": 46558
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.363843970916663e-06,
+      "loss": 0.2438,
+      "step": 46559
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.363508135610297e-06,
+      "loss": 0.23,
+      "step": 46560
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.363172300303931e-06,
+      "loss": 0.2287,
+      "step": 46561
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.362836464997566e-06,
+      "loss": 0.2578,
+      "step": 46562
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3625006296912e-06,
+      "loss": 0.2579,
+      "step": 46563
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.362164794384834e-06,
+      "loss": 0.2089,
+      "step": 46564
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.361828959078468e-06,
+      "loss": 0.2674,
+      "step": 46565
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.361493123772103e-06,
+      "loss": 0.3082,
+      "step": 46566
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.361157288465737e-06,
+      "loss": 0.2496,
+      "step": 46567
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.360821453159371e-06,
+      "loss": 0.3102,
+      "step": 46568
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.360485617853005e-06,
+      "loss": 0.3013,
+      "step": 46569
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.360149782546639e-06,
+      "loss": 0.1243,
+      "step": 46570
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.359813947240274e-06,
+      "loss": 0.2515,
+      "step": 46571
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.359478111933908e-06,
+      "loss": 0.1659,
+      "step": 46572
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.359142276627542e-06,
+      "loss": 0.2256,
+      "step": 46573
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.358806441321177e-06,
+      "loss": 0.3177,
+      "step": 46574
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3584706060148105e-06,
+      "loss": 0.2309,
+      "step": 46575
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.358134770708445e-06,
+      "loss": 0.2982,
+      "step": 46576
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.35779893540208e-06,
+      "loss": 0.3535,
+      "step": 46577
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.357463100095714e-06,
+      "loss": 0.1236,
+      "step": 46578
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3571272647893474e-06,
+      "loss": 0.4468,
+      "step": 46579
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.356791429482982e-06,
+      "loss": 0.2575,
+      "step": 46580
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.356455594176616e-06,
+      "loss": 0.2836,
+      "step": 46581
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3561197588702505e-06,
+      "loss": 0.2448,
+      "step": 46582
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.355783923563884e-06,
+      "loss": 0.2506,
+      "step": 46583
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.355448088257519e-06,
+      "loss": 0.1751,
+      "step": 46584
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.355112252951153e-06,
+      "loss": 0.1807,
+      "step": 46585
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3547764176447874e-06,
+      "loss": 0.3974,
+      "step": 46586
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.354440582338422e-06,
+      "loss": 0.3377,
+      "step": 46587
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.354104747032056e-06,
+      "loss": 0.1548,
+      "step": 46588
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.35376891172569e-06,
+      "loss": 0.2377,
+      "step": 46589
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.353433076419324e-06,
+      "loss": 0.5021,
+      "step": 46590
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.353097241112958e-06,
+      "loss": 0.207,
+      "step": 46591
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.352761405806593e-06,
+      "loss": 0.2225,
+      "step": 46592
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.352425570500227e-06,
+      "loss": 0.2762,
+      "step": 46593
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.352089735193861e-06,
+      "loss": 0.4105,
+      "step": 46594
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.351753899887495e-06,
+      "loss": 0.2706,
+      "step": 46595
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.35141806458113e-06,
+      "loss": 0.1918,
+      "step": 46596
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.351082229274764e-06,
+      "loss": 0.3497,
+      "step": 46597
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.350746393968398e-06,
+      "loss": 0.3099,
+      "step": 46598
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.350410558662033e-06,
+      "loss": 0.3596,
+      "step": 46599
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3500747233556666e-06,
+      "loss": 0.3279,
+      "step": 46600
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3497388880493e-06,
+      "loss": 0.2127,
+      "step": 46601
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.349403052742935e-06,
+      "loss": 0.2383,
+      "step": 46602
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.34906721743657e-06,
+      "loss": 0.1425,
+      "step": 46603
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3487313821302035e-06,
+      "loss": 0.2484,
+      "step": 46604
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.348395546823838e-06,
+      "loss": 0.1808,
+      "step": 46605
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.348059711517472e-06,
+      "loss": 0.2701,
+      "step": 46606
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3477238762111066e-06,
+      "loss": 0.4596,
+      "step": 46607
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.347388040904741e-06,
+      "loss": 0.1695,
+      "step": 46608
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.347052205598375e-06,
+      "loss": 0.2304,
+      "step": 46609
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.346716370292009e-06,
+      "loss": 0.232,
+      "step": 46610
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3463805349856434e-06,
+      "loss": 0.1683,
+      "step": 46611
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.346044699679277e-06,
+      "loss": 0.2205,
+      "step": 46612
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.345708864372912e-06,
+      "loss": 0.3802,
+      "step": 46613
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3453730290665465e-06,
+      "loss": 0.3171,
+      "step": 46614
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.34503719376018e-06,
+      "loss": 0.2489,
+      "step": 46615
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.344701358453814e-06,
+      "loss": 0.3289,
+      "step": 46616
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.344365523147449e-06,
+      "loss": 0.2108,
+      "step": 46617
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3440296878410834e-06,
+      "loss": 0.2373,
+      "step": 46618
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.343693852534717e-06,
+      "loss": 0.3068,
+      "step": 46619
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.343358017228352e-06,
+      "loss": 0.3654,
+      "step": 46620
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.343022181921986e-06,
+      "loss": 0.1775,
+      "step": 46621
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.34268634661562e-06,
+      "loss": 0.1606,
+      "step": 46622
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.342350511309254e-06,
+      "loss": 0.192,
+      "step": 46623
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.342014676002889e-06,
+      "loss": 0.3016,
+      "step": 46624
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.341678840696523e-06,
+      "loss": 0.2214,
+      "step": 46625
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.341343005390157e-06,
+      "loss": 0.2066,
+      "step": 46626
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.341007170083791e-06,
+      "loss": 0.4099,
+      "step": 46627
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.340671334777426e-06,
+      "loss": 0.3431,
+      "step": 46628
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3403354994710595e-06,
+      "loss": 0.3156,
+      "step": 46629
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.339999664164694e-06,
+      "loss": 0.2402,
+      "step": 46630
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.339663828858328e-06,
+      "loss": 0.2414,
+      "step": 46631
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3393279935519626e-06,
+      "loss": 0.2534,
+      "step": 46632
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.338992158245597e-06,
+      "loss": 0.2581,
+      "step": 46633
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.338656322939231e-06,
+      "loss": 0.2772,
+      "step": 46634
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.338320487632865e-06,
+      "loss": 0.3733,
+      "step": 46635
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3379846523264995e-06,
+      "loss": 0.369,
+      "step": 46636
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.337648817020133e-06,
+      "loss": 0.3235,
+      "step": 46637
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.337312981713768e-06,
+      "loss": 0.1203,
+      "step": 46638
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3369771464074026e-06,
+      "loss": 0.504,
+      "step": 46639
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.336641311101036e-06,
+      "loss": 0.3576,
+      "step": 46640
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.33630547579467e-06,
+      "loss": 0.1858,
+      "step": 46641
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.335969640488305e-06,
+      "loss": 0.3828,
+      "step": 46642
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3356338051819395e-06,
+      "loss": 0.276,
+      "step": 46643
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.335297969875573e-06,
+      "loss": 0.3639,
+      "step": 46644
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.334962134569208e-06,
+      "loss": 0.268,
+      "step": 46645
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.334626299262842e-06,
+      "loss": 0.4672,
+      "step": 46646
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3342904639564755e-06,
+      "loss": 0.3886,
+      "step": 46647
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.33395462865011e-06,
+      "loss": 0.2617,
+      "step": 46648
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.333618793343745e-06,
+      "loss": 0.1256,
+      "step": 46649
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.333282958037379e-06,
+      "loss": 0.2673,
+      "step": 46650
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.332947122731013e-06,
+      "loss": 0.2926,
+      "step": 46651
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.332611287424647e-06,
+      "loss": 0.3382,
+      "step": 46652
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.332275452118282e-06,
+      "loss": 0.2869,
+      "step": 46653
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.331939616811916e-06,
+      "loss": 0.3704,
+      "step": 46654
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.33160378150555e-06,
+      "loss": 0.2055,
+      "step": 46655
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.331267946199184e-06,
+      "loss": 0.2122,
+      "step": 46656
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.330932110892819e-06,
+      "loss": 0.2714,
+      "step": 46657
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.330596275586452e-06,
+      "loss": 0.2303,
+      "step": 46658
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.330260440280087e-06,
+      "loss": 0.3132,
+      "step": 46659
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.329924604973722e-06,
+      "loss": 0.2117,
+      "step": 46660
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3295887696673555e-06,
+      "loss": 0.3836,
+      "step": 46661
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.329252934360989e-06,
+      "loss": 0.284,
+      "step": 46662
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.328917099054624e-06,
+      "loss": 0.3185,
+      "step": 46663
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.328581263748259e-06,
+      "loss": 0.2982,
+      "step": 46664
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.328245428441892e-06,
+      "loss": 0.2948,
+      "step": 46665
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.327909593135527e-06,
+      "loss": 0.3463,
+      "step": 46666
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.327573757829161e-06,
+      "loss": 0.3392,
+      "step": 46667
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3272379225227955e-06,
+      "loss": 0.1681,
+      "step": 46668
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.326902087216429e-06,
+      "loss": 0.1981,
+      "step": 46669
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.326566251910064e-06,
+      "loss": 0.2816,
+      "step": 46670
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.326230416603698e-06,
+      "loss": 0.2735,
+      "step": 46671
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.325894581297332e-06,
+      "loss": 0.3043,
+      "step": 46672
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.325558745990966e-06,
+      "loss": 0.2847,
+      "step": 46673
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.325222910684601e-06,
+      "loss": 0.4734,
+      "step": 46674
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.324887075378235e-06,
+      "loss": 0.3065,
+      "step": 46675
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.324551240071869e-06,
+      "loss": 0.2148,
+      "step": 46676
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.324215404765503e-06,
+      "loss": 0.2342,
+      "step": 46677
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.323879569459138e-06,
+      "loss": 0.3027,
+      "step": 46678
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.323543734152772e-06,
+      "loss": 0.2774,
+      "step": 46679
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.323207898846406e-06,
+      "loss": 0.1289,
+      "step": 46680
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.32287206354004e-06,
+      "loss": 0.3745,
+      "step": 46681
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.322536228233675e-06,
+      "loss": 0.3512,
+      "step": 46682
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.322200392927308e-06,
+      "loss": 0.1465,
+      "step": 46683
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.321864557620943e-06,
+      "loss": 0.224,
+      "step": 46684
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.321528722314578e-06,
+      "loss": 0.1265,
+      "step": 46685
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3211928870082115e-06,
+      "loss": 0.31,
+      "step": 46686
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.320857051701845e-06,
+      "loss": 0.1633,
+      "step": 46687
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.32052121639548e-06,
+      "loss": 0.2031,
+      "step": 46688
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.320185381089115e-06,
+      "loss": 0.3084,
+      "step": 46689
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.319849545782748e-06,
+      "loss": 0.3799,
+      "step": 46690
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.319513710476383e-06,
+      "loss": 0.2002,
+      "step": 46691
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.319177875170017e-06,
+      "loss": 0.2559,
+      "step": 46692
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.318842039863651e-06,
+      "loss": 0.1477,
+      "step": 46693
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.318506204557285e-06,
+      "loss": 0.2409,
+      "step": 46694
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.31817036925092e-06,
+      "loss": 0.2411,
+      "step": 46695
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.317834533944554e-06,
+      "loss": 0.1838,
+      "step": 46696
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.317498698638188e-06,
+      "loss": 0.314,
+      "step": 46697
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.317162863331822e-06,
+      "loss": 0.1925,
+      "step": 46698
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.316827028025457e-06,
+      "loss": 0.2754,
+      "step": 46699
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3164911927190915e-06,
+      "loss": 0.1926,
+      "step": 46700
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.316155357412725e-06,
+      "loss": 0.2165,
+      "step": 46701
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.315819522106359e-06,
+      "loss": 0.4281,
+      "step": 46702
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.315483686799994e-06,
+      "loss": 0.2159,
+      "step": 46703
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3151478514936275e-06,
+      "loss": 0.315,
+      "step": 46704
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.314812016187262e-06,
+      "loss": 0.209,
+      "step": 46705
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.314476180880897e-06,
+      "loss": 0.3182,
+      "step": 46706
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.314140345574531e-06,
+      "loss": 0.1944,
+      "step": 46707
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3138045102681644e-06,
+      "loss": 0.1561,
+      "step": 46708
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.313468674961799e-06,
+      "loss": 0.1728,
+      "step": 46709
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.313132839655434e-06,
+      "loss": 0.4904,
+      "step": 46710
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3127970043490675e-06,
+      "loss": 0.3164,
+      "step": 46711
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.312461169042702e-06,
+      "loss": 0.336,
+      "step": 46712
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.312125333736336e-06,
+      "loss": 0.3138,
+      "step": 46713
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.31178949842997e-06,
+      "loss": 0.372,
+      "step": 46714
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3114536631236044e-06,
+      "loss": 0.2541,
+      "step": 46715
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.311117827817239e-06,
+      "loss": 0.1958,
+      "step": 46716
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.310781992510873e-06,
+      "loss": 0.2295,
+      "step": 46717
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.310446157204507e-06,
+      "loss": 0.2508,
+      "step": 46718
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.310110321898141e-06,
+      "loss": 0.202,
+      "step": 46719
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.309774486591776e-06,
+      "loss": 0.2083,
+      "step": 46720
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.30943865128541e-06,
+      "loss": 0.3454,
+      "step": 46721
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.309102815979044e-06,
+      "loss": 0.3068,
+      "step": 46722
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.308766980672678e-06,
+      "loss": 0.141,
+      "step": 46723
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.308431145366313e-06,
+      "loss": 0.3145,
+      "step": 46724
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.308095310059947e-06,
+      "loss": 0.28,
+      "step": 46725
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.307759474753581e-06,
+      "loss": 0.3136,
+      "step": 46726
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.307423639447215e-06,
+      "loss": 0.2927,
+      "step": 46727
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.30708780414085e-06,
+      "loss": 0.3177,
+      "step": 46728
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3067519688344836e-06,
+      "loss": 0.1957,
+      "step": 46729
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.306416133528118e-06,
+      "loss": 0.2586,
+      "step": 46730
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.306080298221753e-06,
+      "loss": 0.2325,
+      "step": 46731
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.305744462915387e-06,
+      "loss": 0.3205,
+      "step": 46732
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3054086276090205e-06,
+      "loss": 0.2372,
+      "step": 46733
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.305072792302655e-06,
+      "loss": 0.4538,
+      "step": 46734
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.30473695699629e-06,
+      "loss": 0.2882,
+      "step": 46735
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3044011216899235e-06,
+      "loss": 0.2056,
+      "step": 46736
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.304065286383558e-06,
+      "loss": 0.2027,
+      "step": 46737
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.303729451077192e-06,
+      "loss": 0.1734,
+      "step": 46738
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.303393615770826e-06,
+      "loss": 0.2137,
+      "step": 46739
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3030577804644604e-06,
+      "loss": 0.1414,
+      "step": 46740
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.302721945158095e-06,
+      "loss": 0.1719,
+      "step": 46741
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.302386109851729e-06,
+      "loss": 0.2382,
+      "step": 46742
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3020502745453635e-06,
+      "loss": 0.3024,
+      "step": 46743
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.301714439238997e-06,
+      "loss": 0.2696,
+      "step": 46744
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.301378603932632e-06,
+      "loss": 0.2968,
+      "step": 46745
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.301042768626267e-06,
+      "loss": 0.1798,
+      "step": 46746
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.3007069333199004e-06,
+      "loss": 0.2573,
+      "step": 46747
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.300371098013534e-06,
+      "loss": 0.2777,
+      "step": 46748
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 4.300035262707169e-06,
+      "loss": 0.274,
+      "step": 46749
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.299699427400803e-06,
+      "loss": 0.157,
+      "step": 46750
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.299363592094437e-06,
+      "loss": 0.3078,
+      "step": 46751
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.299027756788072e-06,
+      "loss": 0.2446,
+      "step": 46752
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.298691921481706e-06,
+      "loss": 0.312,
+      "step": 46753
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.29835608617534e-06,
+      "loss": 0.2187,
+      "step": 46754
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.298020250868974e-06,
+      "loss": 0.1634,
+      "step": 46755
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.297684415562609e-06,
+      "loss": 0.4152,
+      "step": 46756
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.297348580256243e-06,
+      "loss": 0.3022,
+      "step": 46757
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.297012744949877e-06,
+      "loss": 0.3533,
+      "step": 46758
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.296676909643511e-06,
+      "loss": 0.3312,
+      "step": 46759
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.296341074337145e-06,
+      "loss": 0.1768,
+      "step": 46760
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2960052390307796e-06,
+      "loss": 0.3008,
+      "step": 46761
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.295669403724414e-06,
+      "loss": 0.2433,
+      "step": 46762
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.295333568418048e-06,
+      "loss": 0.2767,
+      "step": 46763
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.294997733111682e-06,
+      "loss": 0.1928,
+      "step": 46764
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2946618978053165e-06,
+      "loss": 0.2474,
+      "step": 46765
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.294326062498951e-06,
+      "loss": 0.3615,
+      "step": 46766
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.293990227192585e-06,
+      "loss": 0.2142,
+      "step": 46767
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2936543918862196e-06,
+      "loss": 0.1912,
+      "step": 46768
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.293318556579853e-06,
+      "loss": 0.4301,
+      "step": 46769
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.292982721273488e-06,
+      "loss": 0.2277,
+      "step": 46770
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.292646885967122e-06,
+      "loss": 0.3121,
+      "step": 46771
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2923110506607565e-06,
+      "loss": 0.393,
+      "step": 46772
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.29197521535439e-06,
+      "loss": 0.1629,
+      "step": 46773
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.291639380048025e-06,
+      "loss": 0.2967,
+      "step": 46774
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.291303544741659e-06,
+      "loss": 0.2332,
+      "step": 46775
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.290967709435293e-06,
+      "loss": 0.1841,
+      "step": 46776
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.290631874128928e-06,
+      "loss": 0.2269,
+      "step": 46777
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.290296038822562e-06,
+      "loss": 0.2122,
+      "step": 46778
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.289960203516196e-06,
+      "loss": 0.3228,
+      "step": 46779
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.28962436820983e-06,
+      "loss": 0.2936,
+      "step": 46780
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.289288532903465e-06,
+      "loss": 0.3278,
+      "step": 46781
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.288952697597099e-06,
+      "loss": 0.3596,
+      "step": 46782
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.288616862290733e-06,
+      "loss": 0.2566,
+      "step": 46783
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.288281026984367e-06,
+      "loss": 0.1981,
+      "step": 46784
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.287945191678001e-06,
+      "loss": 0.1646,
+      "step": 46785
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.287609356371636e-06,
+      "loss": 0.1003,
+      "step": 46786
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.28727352106527e-06,
+      "loss": 0.1672,
+      "step": 46787
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.286937685758904e-06,
+      "loss": 0.3396,
+      "step": 46788
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.286601850452539e-06,
+      "loss": 0.2914,
+      "step": 46789
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2862660151461725e-06,
+      "loss": 0.142,
+      "step": 46790
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.285930179839807e-06,
+      "loss": 0.2235,
+      "step": 46791
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.285594344533442e-06,
+      "loss": 0.2621,
+      "step": 46792
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.285258509227076e-06,
+      "loss": 0.3213,
+      "step": 46793
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.284922673920709e-06,
+      "loss": 0.2169,
+      "step": 46794
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.284586838614344e-06,
+      "loss": 0.2678,
+      "step": 46795
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.284251003307978e-06,
+      "loss": 0.4462,
+      "step": 46796
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2839151680016125e-06,
+      "loss": 0.1834,
+      "step": 46797
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.283579332695247e-06,
+      "loss": 0.1923,
+      "step": 46798
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.283243497388881e-06,
+      "loss": 0.2672,
+      "step": 46799
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.282907662082515e-06,
+      "loss": 0.5053,
+      "step": 46800
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.282571826776149e-06,
+      "loss": 0.1625,
+      "step": 46801
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.282235991469784e-06,
+      "loss": 0.3083,
+      "step": 46802
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.281900156163418e-06,
+      "loss": 0.4497,
+      "step": 46803
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2815643208570525e-06,
+      "loss": 0.2817,
+      "step": 46804
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.281228485550686e-06,
+      "loss": 0.25,
+      "step": 46805
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.28089265024432e-06,
+      "loss": 0.1589,
+      "step": 46806
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.280556814937955e-06,
+      "loss": 0.2928,
+      "step": 46807
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.280220979631589e-06,
+      "loss": 0.3194,
+      "step": 46808
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.279885144325223e-06,
+      "loss": 0.2421,
+      "step": 46809
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.279549309018857e-06,
+      "loss": 0.2585,
+      "step": 46810
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.279213473712492e-06,
+      "loss": 0.2474,
+      "step": 46811
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.278877638406126e-06,
+      "loss": 0.2528,
+      "step": 46812
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.27854180309976e-06,
+      "loss": 0.1562,
+      "step": 46813
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.278205967793395e-06,
+      "loss": 0.2987,
+      "step": 46814
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2778701324870285e-06,
+      "loss": 0.1875,
+      "step": 46815
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.277534297180662e-06,
+      "loss": 0.2478,
+      "step": 46816
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.277198461874297e-06,
+      "loss": 0.2697,
+      "step": 46817
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.276862626567932e-06,
+      "loss": 0.4368,
+      "step": 46818
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.276526791261565e-06,
+      "loss": 0.2512,
+      "step": 46819
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2761909559552e-06,
+      "loss": 0.2003,
+      "step": 46820
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.275855120648834e-06,
+      "loss": 0.2891,
+      "step": 46821
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2755192853424685e-06,
+      "loss": 0.2306,
+      "step": 46822
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.275183450036103e-06,
+      "loss": 0.1914,
+      "step": 46823
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.274847614729737e-06,
+      "loss": 0.289,
+      "step": 46824
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.274511779423371e-06,
+      "loss": 0.1761,
+      "step": 46825
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.274175944117005e-06,
+      "loss": 0.2542,
+      "step": 46826
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.273840108810639e-06,
+      "loss": 0.1667,
+      "step": 46827
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.273504273504274e-06,
+      "loss": 0.2599,
+      "step": 46828
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2731684381979085e-06,
+      "loss": 0.259,
+      "step": 46829
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.272832602891542e-06,
+      "loss": 0.1256,
+      "step": 46830
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.272496767585176e-06,
+      "loss": 0.3287,
+      "step": 46831
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.272160932278811e-06,
+      "loss": 0.2826,
+      "step": 46832
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.271825096972445e-06,
+      "loss": 0.2016,
+      "step": 46833
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.271489261666079e-06,
+      "loss": 0.3132,
+      "step": 46834
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.271153426359714e-06,
+      "loss": 0.1467,
+      "step": 46835
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.270817591053348e-06,
+      "loss": 0.3449,
+      "step": 46836
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.270481755746982e-06,
+      "loss": 0.3128,
+      "step": 46837
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.270145920440616e-06,
+      "loss": 0.3295,
+      "step": 46838
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.269810085134251e-06,
+      "loss": 0.4105,
+      "step": 46839
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2694742498278845e-06,
+      "loss": 0.348,
+      "step": 46840
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.269138414521519e-06,
+      "loss": 0.2287,
+      "step": 46841
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.268802579215153e-06,
+      "loss": 0.331,
+      "step": 46842
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.268466743908788e-06,
+      "loss": 0.2333,
+      "step": 46843
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.268130908602422e-06,
+      "loss": 0.2982,
+      "step": 46844
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.267795073296056e-06,
+      "loss": 0.2661,
+      "step": 46845
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.26745923798969e-06,
+      "loss": 0.1778,
+      "step": 46846
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2671234026833245e-06,
+      "loss": 0.2426,
+      "step": 46847
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.266787567376959e-06,
+      "loss": 0.3755,
+      "step": 46848
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.266451732070593e-06,
+      "loss": 0.3154,
+      "step": 46849
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.266115896764228e-06,
+      "loss": 0.305,
+      "step": 46850
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.265780061457861e-06,
+      "loss": 0.4639,
+      "step": 46851
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.265444226151495e-06,
+      "loss": 0.2957,
+      "step": 46852
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.26510839084513e-06,
+      "loss": 0.1399,
+      "step": 46853
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2647725555387645e-06,
+      "loss": 0.1563,
+      "step": 46854
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.264436720232398e-06,
+      "loss": 0.2632,
+      "step": 46855
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.264100884926032e-06,
+      "loss": 0.2581,
+      "step": 46856
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.263765049619667e-06,
+      "loss": 0.3461,
+      "step": 46857
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.263429214313301e-06,
+      "loss": 0.2621,
+      "step": 46858
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.263093379006935e-06,
+      "loss": 0.2289,
+      "step": 46859
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.26275754370057e-06,
+      "loss": 0.268,
+      "step": 46860
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.262421708394204e-06,
+      "loss": 0.2049,
+      "step": 46861
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2620858730878375e-06,
+      "loss": 0.2231,
+      "step": 46862
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.261750037781472e-06,
+      "loss": 0.2136,
+      "step": 46863
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.261414202475107e-06,
+      "loss": 0.263,
+      "step": 46864
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2610783671687405e-06,
+      "loss": 0.3028,
+      "step": 46865
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.260742531862375e-06,
+      "loss": 0.283,
+      "step": 46866
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.260406696556009e-06,
+      "loss": 0.2256,
+      "step": 46867
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.260070861249644e-06,
+      "loss": 0.1612,
+      "step": 46868
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.259735025943278e-06,
+      "loss": 0.4384,
+      "step": 46869
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.259399190636912e-06,
+      "loss": 0.2995,
+      "step": 46870
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.259063355330546e-06,
+      "loss": 0.1726,
+      "step": 46871
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2587275200241805e-06,
+      "loss": 0.3408,
+      "step": 46872
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.258391684717814e-06,
+      "loss": 0.2201,
+      "step": 46873
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.258055849411449e-06,
+      "loss": 0.3092,
+      "step": 46874
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.257720014105084e-06,
+      "loss": 0.2961,
+      "step": 46875
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2573841787987174e-06,
+      "loss": 0.3924,
+      "step": 46876
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.257048343492351e-06,
+      "loss": 0.2053,
+      "step": 46877
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.256712508185986e-06,
+      "loss": 0.2778,
+      "step": 46878
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2563766728796205e-06,
+      "loss": 0.3907,
+      "step": 46879
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.256040837573254e-06,
+      "loss": 0.1091,
+      "step": 46880
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.255705002266889e-06,
+      "loss": 0.147,
+      "step": 46881
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.255369166960523e-06,
+      "loss": 0.217,
+      "step": 46882
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2550333316541566e-06,
+      "loss": 0.2278,
+      "step": 46883
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.254697496347791e-06,
+      "loss": 0.2521,
+      "step": 46884
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.254361661041426e-06,
+      "loss": 0.3853,
+      "step": 46885
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.25402582573506e-06,
+      "loss": 0.2406,
+      "step": 46886
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.253689990428694e-06,
+      "loss": 0.1807,
+      "step": 46887
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.253354155122328e-06,
+      "loss": 0.188,
+      "step": 46888
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.253018319815963e-06,
+      "loss": 0.2661,
+      "step": 46889
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.252682484509597e-06,
+      "loss": 0.165,
+      "step": 46890
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.252346649203231e-06,
+      "loss": 0.2469,
+      "step": 46891
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.252010813896865e-06,
+      "loss": 0.3322,
+      "step": 46892
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2516749785905e-06,
+      "loss": 0.2756,
+      "step": 46893
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2513391432841335e-06,
+      "loss": 0.2746,
+      "step": 46894
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.251003307977768e-06,
+      "loss": 0.3158,
+      "step": 46895
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.250667472671403e-06,
+      "loss": 0.2919,
+      "step": 46896
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2503316373650366e-06,
+      "loss": 0.2299,
+      "step": 46897
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.24999580205867e-06,
+      "loss": 0.2571,
+      "step": 46898
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.249659966752305e-06,
+      "loss": 0.2284,
+      "step": 46899
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.24932413144594e-06,
+      "loss": 0.3453,
+      "step": 46900
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2489882961395735e-06,
+      "loss": 0.2115,
+      "step": 46901
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.248652460833207e-06,
+      "loss": 0.2912,
+      "step": 46902
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.248316625526842e-06,
+      "loss": 0.1478,
+      "step": 46903
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2479807902204765e-06,
+      "loss": 0.3304,
+      "step": 46904
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.24764495491411e-06,
+      "loss": 0.2441,
+      "step": 46905
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.247309119607745e-06,
+      "loss": 0.1359,
+      "step": 46906
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.246973284301379e-06,
+      "loss": 0.2539,
+      "step": 46907
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.246637448995013e-06,
+      "loss": 0.2023,
+      "step": 46908
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.246301613688647e-06,
+      "loss": 0.2367,
+      "step": 46909
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.245965778382282e-06,
+      "loss": 0.3461,
+      "step": 46910
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.245629943075916e-06,
+      "loss": 0.4323,
+      "step": 46911
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.24529410776955e-06,
+      "loss": 0.3172,
+      "step": 46912
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.244958272463184e-06,
+      "loss": 0.2587,
+      "step": 46913
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.244622437156819e-06,
+      "loss": 0.2771,
+      "step": 46914
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2442866018504534e-06,
+      "loss": 0.2156,
+      "step": 46915
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.243950766544087e-06,
+      "loss": 0.326,
+      "step": 46916
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.243614931237721e-06,
+      "loss": 0.4029,
+      "step": 46917
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.243279095931356e-06,
+      "loss": 0.3455,
+      "step": 46918
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2429432606249895e-06,
+      "loss": 0.2572,
+      "step": 46919
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.242607425318624e-06,
+      "loss": 0.2744,
+      "step": 46920
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.242271590012259e-06,
+      "loss": 0.1769,
+      "step": 46921
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.241935754705893e-06,
+      "loss": 0.2628,
+      "step": 46922
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.241599919399526e-06,
+      "loss": 0.1878,
+      "step": 46923
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.241264084093161e-06,
+      "loss": 0.1985,
+      "step": 46924
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.240928248786796e-06,
+      "loss": 0.3485,
+      "step": 46925
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2405924134804295e-06,
+      "loss": 0.2987,
+      "step": 46926
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.240256578174064e-06,
+      "loss": 0.1749,
+      "step": 46927
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.239920742867698e-06,
+      "loss": 0.2866,
+      "step": 46928
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.239584907561332e-06,
+      "loss": 0.2666,
+      "step": 46929
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.239249072254966e-06,
+      "loss": 0.2407,
+      "step": 46930
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.238913236948601e-06,
+      "loss": 0.3537,
+      "step": 46931
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.238577401642235e-06,
+      "loss": 0.1412,
+      "step": 46932
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2382415663358695e-06,
+      "loss": 0.1729,
+      "step": 46933
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.237905731029503e-06,
+      "loss": 0.2043,
+      "step": 46934
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.237569895723138e-06,
+      "loss": 0.233,
+      "step": 46935
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2372340604167726e-06,
+      "loss": 0.1764,
+      "step": 46936
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.236898225110406e-06,
+      "loss": 0.359,
+      "step": 46937
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.23656238980404e-06,
+      "loss": 0.2517,
+      "step": 46938
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.236226554497675e-06,
+      "loss": 0.2425,
+      "step": 46939
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.235890719191309e-06,
+      "loss": 0.2273,
+      "step": 46940
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.235554883884943e-06,
+      "loss": 0.3939,
+      "step": 46941
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.235219048578578e-06,
+      "loss": 0.2569,
+      "step": 46942
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.234883213272212e-06,
+      "loss": 0.3201,
+      "step": 46943
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.2345473779658455e-06,
+      "loss": 0.303,
+      "step": 46944
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.23421154265948e-06,
+      "loss": 0.3239,
+      "step": 46945
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.233875707353115e-06,
+      "loss": 0.2076,
+      "step": 46946
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 4.233539872046749e-06,
+      "loss": 0.3227,
+      "step": 46947
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.233204036740382e-06,
+      "loss": 0.1751,
+      "step": 46948
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.232868201434017e-06,
+      "loss": 0.1249,
+      "step": 46949
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.232532366127652e-06,
+      "loss": 0.226,
+      "step": 46950
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.2321965308212855e-06,
+      "loss": 0.1843,
+      "step": 46951
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.23186069551492e-06,
+      "loss": 0.2799,
+      "step": 46952
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.231524860208554e-06,
+      "loss": 0.2342,
+      "step": 46953
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.231189024902188e-06,
+      "loss": 0.2579,
+      "step": 46954
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.230853189595822e-06,
+      "loss": 0.1375,
+      "step": 46955
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.230517354289457e-06,
+      "loss": 0.2703,
+      "step": 46956
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.230181518983091e-06,
+      "loss": 0.2396,
+      "step": 46957
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.2298456836767255e-06,
+      "loss": 0.088,
+      "step": 46958
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.229509848370359e-06,
+      "loss": 0.2942,
+      "step": 46959
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.229174013063994e-06,
+      "loss": 0.2009,
+      "step": 46960
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.228838177757629e-06,
+      "loss": 0.1258,
+      "step": 46961
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.228502342451262e-06,
+      "loss": 0.463,
+      "step": 46962
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.228166507144896e-06,
+      "loss": 0.3055,
+      "step": 46963
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.227830671838531e-06,
+      "loss": 0.199,
+      "step": 46964
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.227494836532165e-06,
+      "loss": 0.2642,
+      "step": 46965
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.227159001225799e-06,
+      "loss": 0.3783,
+      "step": 46966
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.226823165919434e-06,
+      "loss": 0.3466,
+      "step": 46967
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.226487330613068e-06,
+      "loss": 0.3305,
+      "step": 46968
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.2261514953067015e-06,
+      "loss": 0.4533,
+      "step": 46969
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.225815660000336e-06,
+      "loss": 0.3184,
+      "step": 46970
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.225479824693971e-06,
+      "loss": 0.2513,
+      "step": 46971
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.225143989387605e-06,
+      "loss": 0.1274,
+      "step": 46972
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.224808154081239e-06,
+      "loss": 0.2424,
+      "step": 46973
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.224472318774873e-06,
+      "loss": 0.3289,
+      "step": 46974
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.224136483468507e-06,
+      "loss": 0.2329,
+      "step": 46975
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.2238006481621415e-06,
+      "loss": 0.1851,
+      "step": 46976
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.223464812855776e-06,
+      "loss": 0.134,
+      "step": 46977
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.22312897754941e-06,
+      "loss": 0.1792,
+      "step": 46978
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.222793142243045e-06,
+      "loss": 0.1981,
+      "step": 46979
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.222457306936678e-06,
+      "loss": 0.3378,
+      "step": 46980
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.222121471630313e-06,
+      "loss": 0.4022,
+      "step": 46981
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.221785636323948e-06,
+      "loss": 0.3114,
+      "step": 46982
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.2214498010175815e-06,
+      "loss": 0.2418,
+      "step": 46983
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.221113965711215e-06,
+      "loss": 0.2192,
+      "step": 46984
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.22077813040485e-06,
+      "loss": 0.0796,
+      "step": 46985
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.220442295098484e-06,
+      "loss": 0.3178,
+      "step": 46986
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.220106459792118e-06,
+      "loss": 0.1681,
+      "step": 46987
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.219770624485753e-06,
+      "loss": 0.2772,
+      "step": 46988
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.219434789179387e-06,
+      "loss": 0.2217,
+      "step": 46989
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.219098953873021e-06,
+      "loss": 0.1608,
+      "step": 46990
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.218763118566655e-06,
+      "loss": 0.1796,
+      "step": 46991
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.21842728326029e-06,
+      "loss": 0.2587,
+      "step": 46992
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.218091447953924e-06,
+      "loss": 0.3795,
+      "step": 46993
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.2177556126475575e-06,
+      "loss": 0.1384,
+      "step": 46994
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.217419777341192e-06,
+      "loss": 0.2913,
+      "step": 46995
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.217083942034826e-06,
+      "loss": 0.2409,
+      "step": 46996
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.216748106728461e-06,
+      "loss": 0.3956,
+      "step": 46997
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.216412271422095e-06,
+      "loss": 0.5572,
+      "step": 46998
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.216076436115729e-06,
+      "loss": 0.362,
+      "step": 46999
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.215740600809363e-06,
+      "loss": 0.3805,
+      "step": 47000
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.2154047655029975e-06,
+      "loss": 0.4206,
+      "step": 47001
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.215068930196632e-06,
+      "loss": 0.1475,
+      "step": 47002
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.214733094890266e-06,
+      "loss": 0.2961,
+      "step": 47003
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.214397259583901e-06,
+      "loss": 0.1494,
+      "step": 47004
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.2140614242775344e-06,
+      "loss": 0.3117,
+      "step": 47005
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.213725588971169e-06,
+      "loss": 0.3837,
+      "step": 47006
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.213389753664803e-06,
+      "loss": 0.2343,
+      "step": 47007
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.2130539183584375e-06,
+      "loss": 0.1918,
+      "step": 47008
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.212718083052071e-06,
+      "loss": 0.2656,
+      "step": 47009
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.212382247745706e-06,
+      "loss": 0.2177,
+      "step": 47010
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.21204641243934e-06,
+      "loss": 0.241,
+      "step": 47011
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.211710577132974e-06,
+      "loss": 0.1868,
+      "step": 47012
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.211374741826609e-06,
+      "loss": 0.2103,
+      "step": 47013
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.211038906520243e-06,
+      "loss": 0.1677,
+      "step": 47014
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.210703071213877e-06,
+      "loss": 0.351,
+      "step": 47015
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.210367235907511e-06,
+      "loss": 0.239,
+      "step": 47016
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.210031400601146e-06,
+      "loss": 0.207,
+      "step": 47017
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.20969556529478e-06,
+      "loss": 0.2721,
+      "step": 47018
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.209359729988414e-06,
+      "loss": 0.2372,
+      "step": 47019
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.209023894682048e-06,
+      "loss": 0.2209,
+      "step": 47020
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.208688059375682e-06,
+      "loss": 0.1481,
+      "step": 47021
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.208352224069317e-06,
+      "loss": 0.2454,
+      "step": 47022
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.208016388762951e-06,
+      "loss": 0.205,
+      "step": 47023
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.207680553456585e-06,
+      "loss": 0.3146,
+      "step": 47024
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.20734471815022e-06,
+      "loss": 0.2979,
+      "step": 47025
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.2070088828438536e-06,
+      "loss": 0.1881,
+      "step": 47026
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.206673047537488e-06,
+      "loss": 0.1605,
+      "step": 47027
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.206337212231123e-06,
+      "loss": 0.349,
+      "step": 47028
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.206001376924757e-06,
+      "loss": 0.2045,
+      "step": 47029
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.2056655416183904e-06,
+      "loss": 0.257,
+      "step": 47030
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.205329706312025e-06,
+      "loss": 0.448,
+      "step": 47031
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.204993871005659e-06,
+      "loss": 0.2087,
+      "step": 47032
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.2046580356992935e-06,
+      "loss": 0.2572,
+      "step": 47033
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.204322200392927e-06,
+      "loss": 0.293,
+      "step": 47034
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.203986365086562e-06,
+      "loss": 0.3185,
+      "step": 47035
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.203650529780196e-06,
+      "loss": 0.1685,
+      "step": 47036
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.2033146944738304e-06,
+      "loss": 0.2836,
+      "step": 47037
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.202978859167465e-06,
+      "loss": 0.2941,
+      "step": 47038
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.202643023861099e-06,
+      "loss": 0.2148,
+      "step": 47039
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.202307188554733e-06,
+      "loss": 0.2759,
+      "step": 47040
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.201971353248367e-06,
+      "loss": 0.4298,
+      "step": 47041
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.201635517942001e-06,
+      "loss": 0.2226,
+      "step": 47042
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.201299682635636e-06,
+      "loss": 0.2328,
+      "step": 47043
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.2009638473292704e-06,
+      "loss": 0.3737,
+      "step": 47044
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.200628012022904e-06,
+      "loss": 0.2976,
+      "step": 47045
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.200292176716538e-06,
+      "loss": 0.1389,
+      "step": 47046
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.199956341410173e-06,
+      "loss": 0.1873,
+      "step": 47047
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.199620506103807e-06,
+      "loss": 0.3294,
+      "step": 47048
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.199284670797441e-06,
+      "loss": 0.1542,
+      "step": 47049
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.198948835491076e-06,
+      "loss": 0.1337,
+      "step": 47050
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.1986130001847096e-06,
+      "loss": 0.2966,
+      "step": 47051
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.198277164878344e-06,
+      "loss": 0.1922,
+      "step": 47052
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.197941329571978e-06,
+      "loss": 0.4379,
+      "step": 47053
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.197605494265613e-06,
+      "loss": 0.3689,
+      "step": 47054
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.1972696589592465e-06,
+      "loss": 0.4387,
+      "step": 47055
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.196933823652881e-06,
+      "loss": 0.2015,
+      "step": 47056
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.196597988346515e-06,
+      "loss": 0.3966,
+      "step": 47057
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.1962621530401496e-06,
+      "loss": 0.3366,
+      "step": 47058
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.195926317733784e-06,
+      "loss": 0.3333,
+      "step": 47059
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.195590482427418e-06,
+      "loss": 0.3975,
+      "step": 47060
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.195254647121052e-06,
+      "loss": 0.326,
+      "step": 47061
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.1949188118146865e-06,
+      "loss": 0.3624,
+      "step": 47062
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.194582976508321e-06,
+      "loss": 0.202,
+      "step": 47063
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.194247141201955e-06,
+      "loss": 0.1941,
+      "step": 47064
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.1939113058955896e-06,
+      "loss": 0.3241,
+      "step": 47065
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.193575470589223e-06,
+      "loss": 0.2405,
+      "step": 47066
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.193239635282857e-06,
+      "loss": 0.2275,
+      "step": 47067
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.192903799976492e-06,
+      "loss": 0.1772,
+      "step": 47068
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.1925679646701265e-06,
+      "loss": 0.1681,
+      "step": 47069
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.19223212936376e-06,
+      "loss": 0.4291,
+      "step": 47070
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.191896294057395e-06,
+      "loss": 0.1689,
+      "step": 47071
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.191560458751029e-06,
+      "loss": 0.1399,
+      "step": 47072
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.191224623444663e-06,
+      "loss": 0.2677,
+      "step": 47073
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.190888788138298e-06,
+      "loss": 0.2789,
+      "step": 47074
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.190552952831932e-06,
+      "loss": 0.373,
+      "step": 47075
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.190217117525566e-06,
+      "loss": 0.1904,
+      "step": 47076
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.1898812822192e-06,
+      "loss": 0.2392,
+      "step": 47077
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.189545446912834e-06,
+      "loss": 0.1826,
+      "step": 47078
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.189209611606469e-06,
+      "loss": 0.2002,
+      "step": 47079
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.1888737763001025e-06,
+      "loss": 0.2615,
+      "step": 47080
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.188537940993737e-06,
+      "loss": 0.3271,
+      "step": 47081
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.188202105687371e-06,
+      "loss": 0.3042,
+      "step": 47082
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.187866270381006e-06,
+      "loss": 0.2215,
+      "step": 47083
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.18753043507464e-06,
+      "loss": 0.2815,
+      "step": 47084
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.187194599768274e-06,
+      "loss": 0.2212,
+      "step": 47085
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.186858764461908e-06,
+      "loss": 0.3137,
+      "step": 47086
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.1865229291555425e-06,
+      "loss": 0.273,
+      "step": 47087
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.186187093849176e-06,
+      "loss": 0.2545,
+      "step": 47088
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.185851258542811e-06,
+      "loss": 0.237,
+      "step": 47089
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.1855154232364456e-06,
+      "loss": 0.1069,
+      "step": 47090
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.185179587930079e-06,
+      "loss": 0.3019,
+      "step": 47091
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.184843752623713e-06,
+      "loss": 0.3414,
+      "step": 47092
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.184507917317348e-06,
+      "loss": 0.1294,
+      "step": 47093
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.1841720820109825e-06,
+      "loss": 0.1675,
+      "step": 47094
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.183836246704616e-06,
+      "loss": 0.1866,
+      "step": 47095
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.183500411398251e-06,
+      "loss": 0.2285,
+      "step": 47096
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.183164576091885e-06,
+      "loss": 0.3569,
+      "step": 47097
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.1828287407855185e-06,
+      "loss": 0.2497,
+      "step": 47098
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.182492905479153e-06,
+      "loss": 0.3924,
+      "step": 47099
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.182157070172788e-06,
+      "loss": 0.4376,
+      "step": 47100
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.181821234866422e-06,
+      "loss": 0.4049,
+      "step": 47101
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.181485399560056e-06,
+      "loss": 0.1743,
+      "step": 47102
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.18114956425369e-06,
+      "loss": 0.1723,
+      "step": 47103
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.180813728947325e-06,
+      "loss": 0.1778,
+      "step": 47104
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.180477893640959e-06,
+      "loss": 0.2741,
+      "step": 47105
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.180142058334593e-06,
+      "loss": 0.2853,
+      "step": 47106
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.179806223028227e-06,
+      "loss": 0.1934,
+      "step": 47107
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.179470387721862e-06,
+      "loss": 0.3114,
+      "step": 47108
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.179134552415495e-06,
+      "loss": 0.2954,
+      "step": 47109
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.17879871710913e-06,
+      "loss": 0.3025,
+      "step": 47110
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.178462881802765e-06,
+      "loss": 0.1126,
+      "step": 47111
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.1781270464963985e-06,
+      "loss": 0.1295,
+      "step": 47112
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.177791211190032e-06,
+      "loss": 0.187,
+      "step": 47113
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.177455375883667e-06,
+      "loss": 0.1686,
+      "step": 47114
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.177119540577302e-06,
+      "loss": 0.2473,
+      "step": 47115
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.176783705270935e-06,
+      "loss": 0.1955,
+      "step": 47116
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.17644786996457e-06,
+      "loss": 0.3786,
+      "step": 47117
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.176112034658204e-06,
+      "loss": 0.2045,
+      "step": 47118
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.1757761993518385e-06,
+      "loss": 0.3074,
+      "step": 47119
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.175440364045472e-06,
+      "loss": 0.3596,
+      "step": 47120
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.175104528739107e-06,
+      "loss": 0.2118,
+      "step": 47121
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.174768693432741e-06,
+      "loss": 0.2356,
+      "step": 47122
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.174432858126375e-06,
+      "loss": 0.1985,
+      "step": 47123
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.174097022820009e-06,
+      "loss": 0.388,
+      "step": 47124
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.173761187513644e-06,
+      "loss": 0.1707,
+      "step": 47125
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.173425352207278e-06,
+      "loss": 0.3417,
+      "step": 47126
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.173089516900912e-06,
+      "loss": 0.2801,
+      "step": 47127
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.172753681594546e-06,
+      "loss": 0.199,
+      "step": 47128
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.172417846288181e-06,
+      "loss": 0.2305,
+      "step": 47129
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.172082010981815e-06,
+      "loss": 0.4885,
+      "step": 47130
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.171746175675449e-06,
+      "loss": 0.1874,
+      "step": 47131
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.171410340369083e-06,
+      "loss": 0.3422,
+      "step": 47132
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.171074505062718e-06,
+      "loss": 0.3265,
+      "step": 47133
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.1707386697563514e-06,
+      "loss": 0.4161,
+      "step": 47134
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.170402834449986e-06,
+      "loss": 0.2535,
+      "step": 47135
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.170066999143621e-06,
+      "loss": 0.2587,
+      "step": 47136
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.1697311638372545e-06,
+      "loss": 0.1725,
+      "step": 47137
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.169395328530888e-06,
+      "loss": 0.3661,
+      "step": 47138
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.169059493224523e-06,
+      "loss": 0.2656,
+      "step": 47139
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.168723657918158e-06,
+      "loss": 0.2371,
+      "step": 47140
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.168387822611791e-06,
+      "loss": 0.4142,
+      "step": 47141
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.168051987305426e-06,
+      "loss": 0.1747,
+      "step": 47142
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.16771615199906e-06,
+      "loss": 0.1943,
+      "step": 47143
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.167380316692694e-06,
+      "loss": 0.1977,
+      "step": 47144
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.167044481386328e-06,
+      "loss": 0.3741,
+      "step": 47145
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.166708646079963e-06,
+      "loss": 0.3724,
+      "step": 47146
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.166372810773597e-06,
+      "loss": 0.3329,
+      "step": 47147
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.166036975467231e-06,
+      "loss": 0.2755,
+      "step": 47148
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.165701140160865e-06,
+      "loss": 0.2324,
+      "step": 47149
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1653653048545e-06,
+      "loss": 0.2198,
+      "step": 47150
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1650294695481345e-06,
+      "loss": 0.233,
+      "step": 47151
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.164693634241768e-06,
+      "loss": 0.2325,
+      "step": 47152
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.164357798935402e-06,
+      "loss": 0.306,
+      "step": 47153
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.164021963629037e-06,
+      "loss": 0.3171,
+      "step": 47154
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1636861283226706e-06,
+      "loss": 0.2424,
+      "step": 47155
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.163350293016305e-06,
+      "loss": 0.1938,
+      "step": 47156
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.16301445770994e-06,
+      "loss": 0.3026,
+      "step": 47157
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.162678622403574e-06,
+      "loss": 0.1896,
+      "step": 47158
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1623427870972074e-06,
+      "loss": 0.2139,
+      "step": 47159
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.162006951790842e-06,
+      "loss": 0.2821,
+      "step": 47160
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.161671116484477e-06,
+      "loss": 0.1553,
+      "step": 47161
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1613352811781105e-06,
+      "loss": 0.2238,
+      "step": 47162
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.160999445871745e-06,
+      "loss": 0.1603,
+      "step": 47163
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.160663610565379e-06,
+      "loss": 0.193,
+      "step": 47164
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.160327775259014e-06,
+      "loss": 0.2841,
+      "step": 47165
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1599919399526474e-06,
+      "loss": 0.2006,
+      "step": 47166
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.159656104646282e-06,
+      "loss": 0.2997,
+      "step": 47167
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.159320269339916e-06,
+      "loss": 0.1586,
+      "step": 47168
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.15898443403355e-06,
+      "loss": 0.1945,
+      "step": 47169
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.158648598727184e-06,
+      "loss": 0.3687,
+      "step": 47170
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.158312763420819e-06,
+      "loss": 0.1806,
+      "step": 47171
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.157976928114453e-06,
+      "loss": 0.4014,
+      "step": 47172
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1576410928080874e-06,
+      "loss": 0.291,
+      "step": 47173
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.157305257501721e-06,
+      "loss": 0.2607,
+      "step": 47174
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.156969422195356e-06,
+      "loss": 0.2198,
+      "step": 47175
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1566335868889905e-06,
+      "loss": 0.2066,
+      "step": 47176
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.156297751582624e-06,
+      "loss": 0.4131,
+      "step": 47177
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.155961916276258e-06,
+      "loss": 0.2441,
+      "step": 47178
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.155626080969893e-06,
+      "loss": 0.2704,
+      "step": 47179
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1552902456635266e-06,
+      "loss": 0.3278,
+      "step": 47180
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.154954410357161e-06,
+      "loss": 0.2305,
+      "step": 47181
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.154618575050796e-06,
+      "loss": 0.3396,
+      "step": 47182
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.15428273974443e-06,
+      "loss": 0.3132,
+      "step": 47183
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1539469044380635e-06,
+      "loss": 0.3666,
+      "step": 47184
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.153611069131698e-06,
+      "loss": 0.3352,
+      "step": 47185
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.153275233825333e-06,
+      "loss": 0.7321,
+      "step": 47186
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1529393985189666e-06,
+      "loss": 0.2174,
+      "step": 47187
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.152603563212601e-06,
+      "loss": 0.3336,
+      "step": 47188
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.152267727906235e-06,
+      "loss": 0.2116,
+      "step": 47189
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.151931892599869e-06,
+      "loss": 0.2374,
+      "step": 47190
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1515960572935035e-06,
+      "loss": 0.3277,
+      "step": 47191
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.151260221987138e-06,
+      "loss": 0.305,
+      "step": 47192
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.150924386680772e-06,
+      "loss": 0.1733,
+      "step": 47193
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1505885513744066e-06,
+      "loss": 0.1174,
+      "step": 47194
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.15025271606804e-06,
+      "loss": 0.3628,
+      "step": 47195
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.149916880761675e-06,
+      "loss": 0.3423,
+      "step": 47196
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.14958104545531e-06,
+      "loss": 0.4336,
+      "step": 47197
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1492452101489434e-06,
+      "loss": 0.2106,
+      "step": 47198
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.148909374842577e-06,
+      "loss": 0.3756,
+      "step": 47199
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.148573539536212e-06,
+      "loss": 0.3628,
+      "step": 47200
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.148237704229846e-06,
+      "loss": 0.2919,
+      "step": 47201
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.14790186892348e-06,
+      "loss": 0.1801,
+      "step": 47202
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.147566033617115e-06,
+      "loss": 0.2004,
+      "step": 47203
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.147230198310749e-06,
+      "loss": 0.2245,
+      "step": 47204
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.146894363004383e-06,
+      "loss": 0.4041,
+      "step": 47205
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.146558527698017e-06,
+      "loss": 0.2581,
+      "step": 47206
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.146222692391652e-06,
+      "loss": 0.2366,
+      "step": 47207
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.145886857085286e-06,
+      "loss": 0.3668,
+      "step": 47208
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.14555102177892e-06,
+      "loss": 0.2852,
+      "step": 47209
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.145215186472554e-06,
+      "loss": 0.2361,
+      "step": 47210
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.144879351166188e-06,
+      "loss": 0.353,
+      "step": 47211
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.144543515859823e-06,
+      "loss": 0.2516,
+      "step": 47212
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.144207680553457e-06,
+      "loss": 0.1793,
+      "step": 47213
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.143871845247091e-06,
+      "loss": 0.3065,
+      "step": 47214
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.143536009940725e-06,
+      "loss": 0.3435,
+      "step": 47215
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1432001746343595e-06,
+      "loss": 0.3146,
+      "step": 47216
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.142864339327994e-06,
+      "loss": 0.2054,
+      "step": 47217
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.142528504021628e-06,
+      "loss": 0.3168,
+      "step": 47218
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1421926687152626e-06,
+      "loss": 0.1862,
+      "step": 47219
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.141856833408896e-06,
+      "loss": 0.295,
+      "step": 47220
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.141520998102531e-06,
+      "loss": 0.1912,
+      "step": 47221
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.141185162796165e-06,
+      "loss": 0.2903,
+      "step": 47222
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1408493274897995e-06,
+      "loss": 0.0965,
+      "step": 47223
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.140513492183433e-06,
+      "loss": 0.2418,
+      "step": 47224
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.140177656877068e-06,
+      "loss": 0.4462,
+      "step": 47225
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.139841821570702e-06,
+      "loss": 0.1312,
+      "step": 47226
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.139505986264336e-06,
+      "loss": 0.2695,
+      "step": 47227
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.139170150957971e-06,
+      "loss": 0.1499,
+      "step": 47228
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.138834315651605e-06,
+      "loss": 0.1757,
+      "step": 47229
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.138498480345239e-06,
+      "loss": 0.2596,
+      "step": 47230
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.138162645038873e-06,
+      "loss": 0.2695,
+      "step": 47231
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.137826809732508e-06,
+      "loss": 0.2354,
+      "step": 47232
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.137490974426142e-06,
+      "loss": 0.2427,
+      "step": 47233
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.137155139119776e-06,
+      "loss": 0.193,
+      "step": 47234
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.13681930381341e-06,
+      "loss": 0.237,
+      "step": 47235
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.136483468507044e-06,
+      "loss": 0.1382,
+      "step": 47236
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.136147633200679e-06,
+      "loss": 0.2181,
+      "step": 47237
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.135811797894313e-06,
+      "loss": 0.3907,
+      "step": 47238
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.135475962587947e-06,
+      "loss": 0.1623,
+      "step": 47239
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.135140127281582e-06,
+      "loss": 0.1931,
+      "step": 47240
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1348042919752155e-06,
+      "loss": 0.2357,
+      "step": 47241
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.13446845666885e-06,
+      "loss": 0.1899,
+      "step": 47242
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.134132621362485e-06,
+      "loss": 0.2057,
+      "step": 47243
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.133796786056119e-06,
+      "loss": 0.2997,
+      "step": 47244
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.133460950749752e-06,
+      "loss": 0.2852,
+      "step": 47245
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.133125115443387e-06,
+      "loss": 0.216,
+      "step": 47246
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.132789280137021e-06,
+      "loss": 0.3632,
+      "step": 47247
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1324534448306555e-06,
+      "loss": 0.1613,
+      "step": 47248
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.13211760952429e-06,
+      "loss": 0.3807,
+      "step": 47249
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.131781774217924e-06,
+      "loss": 0.1362,
+      "step": 47250
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.131445938911558e-06,
+      "loss": 0.1946,
+      "step": 47251
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.131110103605192e-06,
+      "loss": 0.0961,
+      "step": 47252
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.130774268298827e-06,
+      "loss": 0.2348,
+      "step": 47253
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.130438432992461e-06,
+      "loss": 0.177,
+      "step": 47254
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1301025976860955e-06,
+      "loss": 0.1475,
+      "step": 47255
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.129766762379729e-06,
+      "loss": 0.24,
+      "step": 47256
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.129430927073363e-06,
+      "loss": 0.2511,
+      "step": 47257
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.129095091766998e-06,
+      "loss": 0.2634,
+      "step": 47258
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.128759256460632e-06,
+      "loss": 0.3299,
+      "step": 47259
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.128423421154266e-06,
+      "loss": 0.3168,
+      "step": 47260
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1280875858479e-06,
+      "loss": 0.2417,
+      "step": 47261
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.127751750541535e-06,
+      "loss": 0.2034,
+      "step": 47262
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.127415915235169e-06,
+      "loss": 0.2305,
+      "step": 47263
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.127080079928803e-06,
+      "loss": 0.2889,
+      "step": 47264
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.126744244622438e-06,
+      "loss": 0.2791,
+      "step": 47265
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1264084093160715e-06,
+      "loss": 0.1636,
+      "step": 47266
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.126072574009705e-06,
+      "loss": 0.2648,
+      "step": 47267
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.12573673870334e-06,
+      "loss": 0.1032,
+      "step": 47268
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.125400903396975e-06,
+      "loss": 0.286,
+      "step": 47269
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.125065068090608e-06,
+      "loss": 0.1973,
+      "step": 47270
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.124729232784243e-06,
+      "loss": 0.2154,
+      "step": 47271
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.124393397477877e-06,
+      "loss": 0.3388,
+      "step": 47272
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1240575621715115e-06,
+      "loss": 0.3309,
+      "step": 47273
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.123721726865146e-06,
+      "loss": 0.3033,
+      "step": 47274
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.12338589155878e-06,
+      "loss": 0.1776,
+      "step": 47275
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.123050056252414e-06,
+      "loss": 0.1117,
+      "step": 47276
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.122714220946048e-06,
+      "loss": 0.3704,
+      "step": 47277
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.122378385639682e-06,
+      "loss": 0.2161,
+      "step": 47278
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.122042550333317e-06,
+      "loss": 0.4036,
+      "step": 47279
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1217067150269515e-06,
+      "loss": 0.2283,
+      "step": 47280
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.121370879720585e-06,
+      "loss": 0.2447,
+      "step": 47281
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.121035044414219e-06,
+      "loss": 0.2329,
+      "step": 47282
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.120699209107854e-06,
+      "loss": 0.1637,
+      "step": 47283
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.120363373801488e-06,
+      "loss": 0.2618,
+      "step": 47284
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.120027538495122e-06,
+      "loss": 0.1521,
+      "step": 47285
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.119691703188757e-06,
+      "loss": 0.3171,
+      "step": 47286
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.119355867882391e-06,
+      "loss": 0.346,
+      "step": 47287
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.119020032576025e-06,
+      "loss": 0.1049,
+      "step": 47288
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.118684197269659e-06,
+      "loss": 0.2504,
+      "step": 47289
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.118348361963294e-06,
+      "loss": 0.307,
+      "step": 47290
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1180125266569275e-06,
+      "loss": 0.1482,
+      "step": 47291
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.117676691350562e-06,
+      "loss": 0.3453,
+      "step": 47292
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.117340856044196e-06,
+      "loss": 0.2054,
+      "step": 47293
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.117005020737831e-06,
+      "loss": 0.3274,
+      "step": 47294
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.116669185431465e-06,
+      "loss": 0.2121,
+      "step": 47295
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.116333350125099e-06,
+      "loss": 0.3284,
+      "step": 47296
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.115997514818733e-06,
+      "loss": 0.2546,
+      "step": 47297
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1156616795123675e-06,
+      "loss": 0.2604,
+      "step": 47298
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.115325844206002e-06,
+      "loss": 0.3018,
+      "step": 47299
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.114990008899636e-06,
+      "loss": 0.3956,
+      "step": 47300
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.114654173593271e-06,
+      "loss": 0.2295,
+      "step": 47301
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1143183382869044e-06,
+      "loss": 0.1595,
+      "step": 47302
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.113982502980538e-06,
+      "loss": 0.2432,
+      "step": 47303
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.113646667674173e-06,
+      "loss": 0.3321,
+      "step": 47304
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1133108323678075e-06,
+      "loss": 0.2971,
+      "step": 47305
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.112974997061441e-06,
+      "loss": 0.3185,
+      "step": 47306
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.112639161755075e-06,
+      "loss": 0.1828,
+      "step": 47307
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.11230332644871e-06,
+      "loss": 0.2225,
+      "step": 47308
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.111967491142344e-06,
+      "loss": 0.1784,
+      "step": 47309
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.111631655835978e-06,
+      "loss": 0.3645,
+      "step": 47310
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.111295820529613e-06,
+      "loss": 0.3029,
+      "step": 47311
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.110959985223247e-06,
+      "loss": 0.3483,
+      "step": 47312
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1106241499168805e-06,
+      "loss": 0.2103,
+      "step": 47313
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.110288314610515e-06,
+      "loss": 0.2106,
+      "step": 47314
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.10995247930415e-06,
+      "loss": 0.2065,
+      "step": 47315
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1096166439977836e-06,
+      "loss": 0.2717,
+      "step": 47316
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.109280808691418e-06,
+      "loss": 0.2243,
+      "step": 47317
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.108944973385052e-06,
+      "loss": 0.1436,
+      "step": 47318
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.108609138078687e-06,
+      "loss": 0.3802,
+      "step": 47319
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.108273302772321e-06,
+      "loss": 0.2534,
+      "step": 47320
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.107937467465955e-06,
+      "loss": 0.3596,
+      "step": 47321
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.107601632159589e-06,
+      "loss": 0.4874,
+      "step": 47322
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1072657968532235e-06,
+      "loss": 0.2981,
+      "step": 47323
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.106929961546857e-06,
+      "loss": 0.2602,
+      "step": 47324
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.106594126240492e-06,
+      "loss": 0.1685,
+      "step": 47325
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.106258290934127e-06,
+      "loss": 0.2784,
+      "step": 47326
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1059224556277604e-06,
+      "loss": 0.2373,
+      "step": 47327
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.105586620321394e-06,
+      "loss": 0.3294,
+      "step": 47328
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.105250785015029e-06,
+      "loss": 0.2344,
+      "step": 47329
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1049149497086635e-06,
+      "loss": 0.2142,
+      "step": 47330
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.104579114402297e-06,
+      "loss": 0.2545,
+      "step": 47331
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.104243279095932e-06,
+      "loss": 0.2757,
+      "step": 47332
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.103907443789566e-06,
+      "loss": 0.2156,
+      "step": 47333
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1035716084832004e-06,
+      "loss": 0.2732,
+      "step": 47334
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.103235773176834e-06,
+      "loss": 0.2114,
+      "step": 47335
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.102899937870469e-06,
+      "loss": 0.2447,
+      "step": 47336
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.102564102564103e-06,
+      "loss": 0.2959,
+      "step": 47337
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.102228267257737e-06,
+      "loss": 0.2689,
+      "step": 47338
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.101892431951371e-06,
+      "loss": 0.2307,
+      "step": 47339
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.101556596645006e-06,
+      "loss": 0.1798,
+      "step": 47340
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.1012207613386404e-06,
+      "loss": 0.27,
+      "step": 47341
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.100884926032274e-06,
+      "loss": 0.2384,
+      "step": 47342
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.100549090725908e-06,
+      "loss": 0.2258,
+      "step": 47343
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 4.100213255419543e-06,
+      "loss": 0.2859,
+      "step": 47344
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.099877420113177e-06,
+      "loss": 0.3983,
+      "step": 47345
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.099541584806811e-06,
+      "loss": 0.1512,
+      "step": 47346
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.099205749500446e-06,
+      "loss": 0.3873,
+      "step": 47347
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0988699141940796e-06,
+      "loss": 0.4566,
+      "step": 47348
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.098534078887713e-06,
+      "loss": 0.3844,
+      "step": 47349
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.098198243581348e-06,
+      "loss": 0.317,
+      "step": 47350
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.097862408274983e-06,
+      "loss": 0.1965,
+      "step": 47351
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0975265729686165e-06,
+      "loss": 0.2465,
+      "step": 47352
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.09719073766225e-06,
+      "loss": 0.2826,
+      "step": 47353
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.096854902355885e-06,
+      "loss": 0.2402,
+      "step": 47354
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0965190670495196e-06,
+      "loss": 0.2942,
+      "step": 47355
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.096183231743153e-06,
+      "loss": 0.3042,
+      "step": 47356
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.095847396436788e-06,
+      "loss": 0.3217,
+      "step": 47357
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.095511561130422e-06,
+      "loss": 0.1064,
+      "step": 47358
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.095175725824056e-06,
+      "loss": 0.3854,
+      "step": 47359
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.09483989051769e-06,
+      "loss": 0.2438,
+      "step": 47360
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.094504055211325e-06,
+      "loss": 0.1782,
+      "step": 47361
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.094168219904959e-06,
+      "loss": 0.2282,
+      "step": 47362
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.093832384598593e-06,
+      "loss": 0.117,
+      "step": 47363
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.093496549292227e-06,
+      "loss": 0.2229,
+      "step": 47364
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.093160713985862e-06,
+      "loss": 0.2086,
+      "step": 47365
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0928248786794964e-06,
+      "loss": 0.2358,
+      "step": 47366
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.09248904337313e-06,
+      "loss": 0.2167,
+      "step": 47367
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.092153208066764e-06,
+      "loss": 0.266,
+      "step": 47368
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.091817372760399e-06,
+      "loss": 0.3034,
+      "step": 47369
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0914815374540325e-06,
+      "loss": 0.423,
+      "step": 47370
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.091145702147667e-06,
+      "loss": 0.3231,
+      "step": 47371
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.090809866841302e-06,
+      "loss": 0.2945,
+      "step": 47372
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.090474031534936e-06,
+      "loss": 0.2181,
+      "step": 47373
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.090138196228569e-06,
+      "loss": 0.2814,
+      "step": 47374
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.089802360922204e-06,
+      "loss": 0.2454,
+      "step": 47375
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.089466525615839e-06,
+      "loss": 0.3761,
+      "step": 47376
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0891306903094725e-06,
+      "loss": 0.2781,
+      "step": 47377
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.088794855003107e-06,
+      "loss": 0.4686,
+      "step": 47378
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.088459019696741e-06,
+      "loss": 0.2245,
+      "step": 47379
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.088123184390375e-06,
+      "loss": 0.3166,
+      "step": 47380
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.087787349084009e-06,
+      "loss": 0.2922,
+      "step": 47381
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.087451513777644e-06,
+      "loss": 0.2669,
+      "step": 47382
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.087115678471278e-06,
+      "loss": 0.1204,
+      "step": 47383
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0867798431649125e-06,
+      "loss": 0.175,
+      "step": 47384
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.086444007858546e-06,
+      "loss": 0.3524,
+      "step": 47385
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.086108172552181e-06,
+      "loss": 0.2826,
+      "step": 47386
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0857723372458156e-06,
+      "loss": 0.1753,
+      "step": 47387
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.085436501939449e-06,
+      "loss": 0.2685,
+      "step": 47388
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.085100666633083e-06,
+      "loss": 0.4066,
+      "step": 47389
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.084764831326718e-06,
+      "loss": 0.2913,
+      "step": 47390
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.084428996020352e-06,
+      "loss": 0.2817,
+      "step": 47391
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.084093160713986e-06,
+      "loss": 0.1827,
+      "step": 47392
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.083757325407621e-06,
+      "loss": 0.4174,
+      "step": 47393
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.083421490101255e-06,
+      "loss": 0.2095,
+      "step": 47394
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0830856547948885e-06,
+      "loss": 0.2551,
+      "step": 47395
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.082749819488523e-06,
+      "loss": 0.2118,
+      "step": 47396
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.082413984182158e-06,
+      "loss": 0.3077,
+      "step": 47397
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.082078148875792e-06,
+      "loss": 0.234,
+      "step": 47398
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.081742313569425e-06,
+      "loss": 0.2085,
+      "step": 47399
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.08140647826306e-06,
+      "loss": 0.22,
+      "step": 47400
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.081070642956695e-06,
+      "loss": 0.2511,
+      "step": 47401
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0807348076503285e-06,
+      "loss": 0.4216,
+      "step": 47402
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.080398972343963e-06,
+      "loss": 0.2079,
+      "step": 47403
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.080063137037597e-06,
+      "loss": 0.3489,
+      "step": 47404
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.079727301731231e-06,
+      "loss": 0.3209,
+      "step": 47405
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.079391466424865e-06,
+      "loss": 0.2987,
+      "step": 47406
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0790556311185e-06,
+      "loss": 0.2203,
+      "step": 47407
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.078719795812134e-06,
+      "loss": 0.3387,
+      "step": 47408
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0783839605057685e-06,
+      "loss": 0.3256,
+      "step": 47409
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.078048125199402e-06,
+      "loss": 0.228,
+      "step": 47410
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.077712289893037e-06,
+      "loss": 0.3195,
+      "step": 47411
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.077376454586672e-06,
+      "loss": 0.1496,
+      "step": 47412
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.077040619280305e-06,
+      "loss": 0.2882,
+      "step": 47413
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.076704783973939e-06,
+      "loss": 0.1708,
+      "step": 47414
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.076368948667574e-06,
+      "loss": 0.3194,
+      "step": 47415
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.076033113361208e-06,
+      "loss": 0.2798,
+      "step": 47416
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.075697278054842e-06,
+      "loss": 0.3305,
+      "step": 47417
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.075361442748477e-06,
+      "loss": 0.4892,
+      "step": 47418
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.075025607442111e-06,
+      "loss": 0.1713,
+      "step": 47419
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0746897721357445e-06,
+      "loss": 0.2808,
+      "step": 47420
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.074353936829379e-06,
+      "loss": 0.1444,
+      "step": 47421
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.074018101523014e-06,
+      "loss": 0.158,
+      "step": 47422
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.073682266216648e-06,
+      "loss": 0.2858,
+      "step": 47423
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.073346430910282e-06,
+      "loss": 0.2406,
+      "step": 47424
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.073010595603916e-06,
+      "loss": 0.3714,
+      "step": 47425
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.07267476029755e-06,
+      "loss": 0.374,
+      "step": 47426
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0723389249911845e-06,
+      "loss": 0.1982,
+      "step": 47427
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.072003089684819e-06,
+      "loss": 0.3244,
+      "step": 47428
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.071667254378453e-06,
+      "loss": 0.4861,
+      "step": 47429
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.071331419072088e-06,
+      "loss": 0.2433,
+      "step": 47430
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.070995583765721e-06,
+      "loss": 0.2965,
+      "step": 47431
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.070659748459356e-06,
+      "loss": 0.3667,
+      "step": 47432
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.070323913152991e-06,
+      "loss": 0.1984,
+      "step": 47433
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0699880778466245e-06,
+      "loss": 0.2006,
+      "step": 47434
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.069652242540258e-06,
+      "loss": 0.3242,
+      "step": 47435
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.069316407233893e-06,
+      "loss": 0.2901,
+      "step": 47436
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.068980571927527e-06,
+      "loss": 0.1626,
+      "step": 47437
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.068644736621161e-06,
+      "loss": 0.3598,
+      "step": 47438
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.068308901314796e-06,
+      "loss": 0.2744,
+      "step": 47439
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.06797306600843e-06,
+      "loss": 0.3153,
+      "step": 47440
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.067637230702064e-06,
+      "loss": 0.3954,
+      "step": 47441
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.067301395395698e-06,
+      "loss": 0.4225,
+      "step": 47442
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.066965560089333e-06,
+      "loss": 0.1438,
+      "step": 47443
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.066629724782967e-06,
+      "loss": 0.3752,
+      "step": 47444
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0662938894766006e-06,
+      "loss": 0.3153,
+      "step": 47445
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.065958054170235e-06,
+      "loss": 0.2443,
+      "step": 47446
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.06562221886387e-06,
+      "loss": 0.2003,
+      "step": 47447
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.065286383557504e-06,
+      "loss": 0.1627,
+      "step": 47448
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.064950548251138e-06,
+      "loss": 0.2504,
+      "step": 47449
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.064614712944772e-06,
+      "loss": 0.3674,
+      "step": 47450
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.064278877638406e-06,
+      "loss": 0.5652,
+      "step": 47451
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0639430423320405e-06,
+      "loss": 0.3938,
+      "step": 47452
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.063607207025675e-06,
+      "loss": 0.304,
+      "step": 47453
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.063271371719309e-06,
+      "loss": 0.3253,
+      "step": 47454
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.062935536412944e-06,
+      "loss": 0.2519,
+      "step": 47455
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0625997011065774e-06,
+      "loss": 0.3859,
+      "step": 47456
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.062263865800212e-06,
+      "loss": 0.3289,
+      "step": 47457
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.061928030493847e-06,
+      "loss": 0.2819,
+      "step": 47458
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0615921951874805e-06,
+      "loss": 0.2171,
+      "step": 47459
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.061256359881114e-06,
+      "loss": 0.372,
+      "step": 47460
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.060920524574749e-06,
+      "loss": 0.237,
+      "step": 47461
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.060584689268383e-06,
+      "loss": 0.1915,
+      "step": 47462
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0602488539620174e-06,
+      "loss": 0.1915,
+      "step": 47463
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.059913018655652e-06,
+      "loss": 0.2057,
+      "step": 47464
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.059577183349286e-06,
+      "loss": 0.2838,
+      "step": 47465
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.05924134804292e-06,
+      "loss": 0.3,
+      "step": 47466
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.058905512736554e-06,
+      "loss": 0.3427,
+      "step": 47467
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.058569677430189e-06,
+      "loss": 0.2382,
+      "step": 47468
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.058233842123823e-06,
+      "loss": 0.3088,
+      "step": 47469
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.057898006817457e-06,
+      "loss": 0.1819,
+      "step": 47470
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.057562171511091e-06,
+      "loss": 0.2677,
+      "step": 47471
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.057226336204725e-06,
+      "loss": 0.2689,
+      "step": 47472
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.05689050089836e-06,
+      "loss": 0.2757,
+      "step": 47473
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.056554665591994e-06,
+      "loss": 0.2849,
+      "step": 47474
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.056218830285628e-06,
+      "loss": 0.3609,
+      "step": 47475
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.055882994979263e-06,
+      "loss": 0.218,
+      "step": 47476
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0555471596728966e-06,
+      "loss": 0.1961,
+      "step": 47477
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.055211324366531e-06,
+      "loss": 0.3091,
+      "step": 47478
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.054875489060166e-06,
+      "loss": 0.2471,
+      "step": 47479
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0545396537538e-06,
+      "loss": 0.2248,
+      "step": 47480
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0542038184474335e-06,
+      "loss": 0.2085,
+      "step": 47481
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.053867983141068e-06,
+      "loss": 0.2389,
+      "step": 47482
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.053532147834702e-06,
+      "loss": 0.2223,
+      "step": 47483
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0531963125283366e-06,
+      "loss": 0.306,
+      "step": 47484
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.05286047722197e-06,
+      "loss": 0.278,
+      "step": 47485
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.052524641915605e-06,
+      "loss": 0.3269,
+      "step": 47486
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.052188806609239e-06,
+      "loss": 0.3661,
+      "step": 47487
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0518529713028735e-06,
+      "loss": 0.2545,
+      "step": 47488
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.051517135996508e-06,
+      "loss": 0.1818,
+      "step": 47489
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.051181300690142e-06,
+      "loss": 0.2687,
+      "step": 47490
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.050845465383776e-06,
+      "loss": 0.172,
+      "step": 47491
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.05050963007741e-06,
+      "loss": 0.254,
+      "step": 47492
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.050173794771044e-06,
+      "loss": 0.1932,
+      "step": 47493
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.049837959464679e-06,
+      "loss": 0.2935,
+      "step": 47494
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0495021241583134e-06,
+      "loss": 0.2569,
+      "step": 47495
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.049166288851947e-06,
+      "loss": 0.2401,
+      "step": 47496
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.048830453545581e-06,
+      "loss": 0.1792,
+      "step": 47497
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.048494618239216e-06,
+      "loss": 0.1549,
+      "step": 47498
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.04815878293285e-06,
+      "loss": 0.4407,
+      "step": 47499
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.047822947626484e-06,
+      "loss": 0.2875,
+      "step": 47500
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.047487112320119e-06,
+      "loss": 0.2495,
+      "step": 47501
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.047151277013753e-06,
+      "loss": 0.2138,
+      "step": 47502
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.046815441707387e-06,
+      "loss": 0.27,
+      "step": 47503
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.046479606401021e-06,
+      "loss": 0.4196,
+      "step": 47504
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.046143771094656e-06,
+      "loss": 0.1994,
+      "step": 47505
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0458079357882895e-06,
+      "loss": 0.1963,
+      "step": 47506
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.045472100481924e-06,
+      "loss": 0.154,
+      "step": 47507
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.045136265175558e-06,
+      "loss": 0.1719,
+      "step": 47508
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0448004298691926e-06,
+      "loss": 0.368,
+      "step": 47509
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.044464594562827e-06,
+      "loss": 0.2472,
+      "step": 47510
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.044128759256461e-06,
+      "loss": 0.3939,
+      "step": 47511
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.043792923950095e-06,
+      "loss": 0.2777,
+      "step": 47512
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0434570886437295e-06,
+      "loss": 0.2467,
+      "step": 47513
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.043121253337364e-06,
+      "loss": 0.3055,
+      "step": 47514
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.042785418030998e-06,
+      "loss": 0.2955,
+      "step": 47515
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0424495827246326e-06,
+      "loss": 0.1646,
+      "step": 47516
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.042113747418266e-06,
+      "loss": 0.1945,
+      "step": 47517
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0417779121119e-06,
+      "loss": 0.2591,
+      "step": 47518
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.041442076805535e-06,
+      "loss": 0.2401,
+      "step": 47519
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0411062414991695e-06,
+      "loss": 0.3113,
+      "step": 47520
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.040770406192803e-06,
+      "loss": 0.3778,
+      "step": 47521
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.040434570886438e-06,
+      "loss": 0.3132,
+      "step": 47522
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.040098735580072e-06,
+      "loss": 0.3007,
+      "step": 47523
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.039762900273706e-06,
+      "loss": 0.3171,
+      "step": 47524
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.039427064967341e-06,
+      "loss": 0.1599,
+      "step": 47525
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.039091229660975e-06,
+      "loss": 0.3688,
+      "step": 47526
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.038755394354609e-06,
+      "loss": 0.2634,
+      "step": 47527
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.038419559048243e-06,
+      "loss": 0.2291,
+      "step": 47528
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.038083723741877e-06,
+      "loss": 0.2251,
+      "step": 47529
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.037747888435512e-06,
+      "loss": 0.18,
+      "step": 47530
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0374120531291455e-06,
+      "loss": 0.2131,
+      "step": 47531
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.03707621782278e-06,
+      "loss": 0.188,
+      "step": 47532
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.036740382516414e-06,
+      "loss": 0.2608,
+      "step": 47533
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.036404547210049e-06,
+      "loss": 0.247,
+      "step": 47534
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.036068711903683e-06,
+      "loss": 0.1408,
+      "step": 47535
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.035732876597317e-06,
+      "loss": 0.402,
+      "step": 47536
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.035397041290951e-06,
+      "loss": 0.3198,
+      "step": 47537
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.0350612059845855e-06,
+      "loss": 0.2696,
+      "step": 47538
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.034725370678219e-06,
+      "loss": 0.2268,
+      "step": 47539
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.034389535371854e-06,
+      "loss": 0.308,
+      "step": 47540
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.034053700065489e-06,
+      "loss": 0.3044,
+      "step": 47541
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.033717864759122e-06,
+      "loss": 0.3111,
+      "step": 47542
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.033382029452756e-06,
+      "loss": 0.2619,
+      "step": 47543
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.033046194146391e-06,
+      "loss": 0.2728,
+      "step": 47544
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.0327103588400255e-06,
+      "loss": 0.3826,
+      "step": 47545
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.032374523533659e-06,
+      "loss": 0.2927,
+      "step": 47546
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.032038688227294e-06,
+      "loss": 0.1804,
+      "step": 47547
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.031702852920928e-06,
+      "loss": 0.283,
+      "step": 47548
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.0313670176145615e-06,
+      "loss": 0.2264,
+      "step": 47549
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.031031182308196e-06,
+      "loss": 0.5982,
+      "step": 47550
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.030695347001831e-06,
+      "loss": 0.23,
+      "step": 47551
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.030359511695465e-06,
+      "loss": 0.3223,
+      "step": 47552
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.030023676389099e-06,
+      "loss": 0.3216,
+      "step": 47553
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.029687841082733e-06,
+      "loss": 0.4078,
+      "step": 47554
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.029352005776368e-06,
+      "loss": 0.1308,
+      "step": 47555
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.029016170470002e-06,
+      "loss": 0.3282,
+      "step": 47556
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.028680335163636e-06,
+      "loss": 0.3097,
+      "step": 47557
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.02834449985727e-06,
+      "loss": 0.1891,
+      "step": 47558
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.028008664550905e-06,
+      "loss": 0.2236,
+      "step": 47559
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.027672829244538e-06,
+      "loss": 0.2958,
+      "step": 47560
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.027336993938173e-06,
+      "loss": 0.4037,
+      "step": 47561
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.027001158631808e-06,
+      "loss": 0.2082,
+      "step": 47562
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.0266653233254415e-06,
+      "loss": 0.3106,
+      "step": 47563
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.026329488019075e-06,
+      "loss": 0.2637,
+      "step": 47564
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.02599365271271e-06,
+      "loss": 0.1981,
+      "step": 47565
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.025657817406345e-06,
+      "loss": 0.2551,
+      "step": 47566
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.025321982099978e-06,
+      "loss": 0.4497,
+      "step": 47567
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.024986146793613e-06,
+      "loss": 0.227,
+      "step": 47568
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.024650311487247e-06,
+      "loss": 0.2797,
+      "step": 47569
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.0243144761808815e-06,
+      "loss": 0.2874,
+      "step": 47570
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.023978640874516e-06,
+      "loss": 0.3187,
+      "step": 47571
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.02364280556815e-06,
+      "loss": 0.2874,
+      "step": 47572
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.023306970261784e-06,
+      "loss": 0.1993,
+      "step": 47573
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.022971134955418e-06,
+      "loss": 0.2228,
+      "step": 47574
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.022635299649052e-06,
+      "loss": 0.2205,
+      "step": 47575
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.022299464342687e-06,
+      "loss": 0.1942,
+      "step": 47576
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.021963629036321e-06,
+      "loss": 0.2011,
+      "step": 47577
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.021627793729955e-06,
+      "loss": 0.4222,
+      "step": 47578
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.021291958423589e-06,
+      "loss": 0.3917,
+      "step": 47579
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.020956123117224e-06,
+      "loss": 0.2607,
+      "step": 47580
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.020620287810858e-06,
+      "loss": 0.2065,
+      "step": 47581
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.020284452504492e-06,
+      "loss": 0.3797,
+      "step": 47582
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.019948617198126e-06,
+      "loss": 0.2223,
+      "step": 47583
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.019612781891761e-06,
+      "loss": 0.3614,
+      "step": 47584
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.0192769465853944e-06,
+      "loss": 0.3017,
+      "step": 47585
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.018941111279029e-06,
+      "loss": 0.1605,
+      "step": 47586
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.018605275972664e-06,
+      "loss": 0.2911,
+      "step": 47587
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.0182694406662975e-06,
+      "loss": 0.2219,
+      "step": 47588
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.017933605359931e-06,
+      "loss": 0.1922,
+      "step": 47589
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.017597770053566e-06,
+      "loss": 0.1585,
+      "step": 47590
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.017261934747201e-06,
+      "loss": 0.2908,
+      "step": 47591
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.0169260994408344e-06,
+      "loss": 0.1364,
+      "step": 47592
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.016590264134469e-06,
+      "loss": 0.3905,
+      "step": 47593
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.016254428828103e-06,
+      "loss": 0.2409,
+      "step": 47594
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.015918593521737e-06,
+      "loss": 0.1797,
+      "step": 47595
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.015582758215371e-06,
+      "loss": 0.3038,
+      "step": 47596
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.015246922909006e-06,
+      "loss": 0.2436,
+      "step": 47597
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.01491108760264e-06,
+      "loss": 0.218,
+      "step": 47598
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.014575252296274e-06,
+      "loss": 0.3121,
+      "step": 47599
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.014239416989908e-06,
+      "loss": 0.3112,
+      "step": 47600
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.013903581683543e-06,
+      "loss": 0.2604,
+      "step": 47601
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.0135677463771775e-06,
+      "loss": 0.1167,
+      "step": 47602
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.013231911070811e-06,
+      "loss": 0.2726,
+      "step": 47603
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.012896075764445e-06,
+      "loss": 0.1631,
+      "step": 47604
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.01256024045808e-06,
+      "loss": 0.2931,
+      "step": 47605
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.0122244051517136e-06,
+      "loss": 0.2371,
+      "step": 47606
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.011888569845348e-06,
+      "loss": 0.1532,
+      "step": 47607
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.011552734538983e-06,
+      "loss": 0.1352,
+      "step": 47608
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.011216899232617e-06,
+      "loss": 0.2062,
+      "step": 47609
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.0108810639262505e-06,
+      "loss": 0.3634,
+      "step": 47610
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.010545228619885e-06,
+      "loss": 0.2216,
+      "step": 47611
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.01020939331352e-06,
+      "loss": 0.2449,
+      "step": 47612
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.0098735580071536e-06,
+      "loss": 0.2057,
+      "step": 47613
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.009537722700788e-06,
+      "loss": 0.1565,
+      "step": 47614
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.009201887394422e-06,
+      "loss": 0.1422,
+      "step": 47615
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.008866052088057e-06,
+      "loss": 0.2837,
+      "step": 47616
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.0085302167816904e-06,
+      "loss": 0.1864,
+      "step": 47617
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.008194381475325e-06,
+      "loss": 0.1847,
+      "step": 47618
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.007858546168959e-06,
+      "loss": 0.2466,
+      "step": 47619
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.007522710862593e-06,
+      "loss": 0.2086,
+      "step": 47620
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.007186875556227e-06,
+      "loss": 0.2302,
+      "step": 47621
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.006851040249862e-06,
+      "loss": 0.2023,
+      "step": 47622
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.006515204943496e-06,
+      "loss": 0.49,
+      "step": 47623
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.0061793696371304e-06,
+      "loss": 0.1848,
+      "step": 47624
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.005843534330764e-06,
+      "loss": 0.1311,
+      "step": 47625
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.005507699024399e-06,
+      "loss": 0.3716,
+      "step": 47626
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.0051718637180335e-06,
+      "loss": 0.1318,
+      "step": 47627
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.004836028411667e-06,
+      "loss": 0.1354,
+      "step": 47628
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.004500193105301e-06,
+      "loss": 0.2388,
+      "step": 47629
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.004164357798936e-06,
+      "loss": 0.1439,
+      "step": 47630
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.00382852249257e-06,
+      "loss": 0.4288,
+      "step": 47631
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.003492687186204e-06,
+      "loss": 0.1835,
+      "step": 47632
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.003156851879839e-06,
+      "loss": 0.2691,
+      "step": 47633
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.002821016573473e-06,
+      "loss": 0.2675,
+      "step": 47634
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.0024851812671065e-06,
+      "loss": 0.4425,
+      "step": 47635
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.002149345960741e-06,
+      "loss": 0.1145,
+      "step": 47636
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.001813510654376e-06,
+      "loss": 0.241,
+      "step": 47637
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.0014776753480096e-06,
+      "loss": 0.3086,
+      "step": 47638
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.001141840041644e-06,
+      "loss": 0.1617,
+      "step": 47639
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.000806004735278e-06,
+      "loss": 0.2479,
+      "step": 47640
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.000470169428912e-06,
+      "loss": 0.1853,
+      "step": 47641
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 4.0001343341225465e-06,
+      "loss": 0.2666,
+      "step": 47642
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.999798498816181e-06,
+      "loss": 0.3115,
+      "step": 47643
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.999462663509815e-06,
+      "loss": 0.2284,
+      "step": 47644
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.9991268282034496e-06,
+      "loss": 0.2971,
+      "step": 47645
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.998790992897083e-06,
+      "loss": 0.396,
+      "step": 47646
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.998455157590718e-06,
+      "loss": 0.1561,
+      "step": 47647
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.998119322284353e-06,
+      "loss": 0.2025,
+      "step": 47648
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.9977834869779865e-06,
+      "loss": 0.2513,
+      "step": 47649
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.99744765167162e-06,
+      "loss": 0.1507,
+      "step": 47650
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.997111816365255e-06,
+      "loss": 0.326,
+      "step": 47651
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.996775981058889e-06,
+      "loss": 0.168,
+      "step": 47652
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.996440145752523e-06,
+      "loss": 0.3128,
+      "step": 47653
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.996104310446158e-06,
+      "loss": 0.2669,
+      "step": 47654
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.995768475139792e-06,
+      "loss": 0.4646,
+      "step": 47655
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.995432639833426e-06,
+      "loss": 0.3104,
+      "step": 47656
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.99509680452706e-06,
+      "loss": 0.1992,
+      "step": 47657
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.994760969220695e-06,
+      "loss": 0.3202,
+      "step": 47658
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.994425133914329e-06,
+      "loss": 0.2657,
+      "step": 47659
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.994089298607963e-06,
+      "loss": 0.3248,
+      "step": 47660
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.993753463301597e-06,
+      "loss": 0.3235,
+      "step": 47661
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.993417627995231e-06,
+      "loss": 0.1931,
+      "step": 47662
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.993081792688866e-06,
+      "loss": 0.531,
+      "step": 47663
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.9927459573825e-06,
+      "loss": 0.308,
+      "step": 47664
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.992410122076134e-06,
+      "loss": 0.4117,
+      "step": 47665
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.992074286769768e-06,
+      "loss": 0.2164,
+      "step": 47666
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.9917384514634025e-06,
+      "loss": 0.2324,
+      "step": 47667
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.991402616157037e-06,
+      "loss": 0.2925,
+      "step": 47668
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.991066780850671e-06,
+      "loss": 0.1919,
+      "step": 47669
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.990730945544306e-06,
+      "loss": 0.2149,
+      "step": 47670
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.990395110237939e-06,
+      "loss": 0.1699,
+      "step": 47671
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.990059274931574e-06,
+      "loss": 0.3738,
+      "step": 47672
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.989723439625208e-06,
+      "loss": 0.1689,
+      "step": 47673
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.9893876043188425e-06,
+      "loss": 0.2258,
+      "step": 47674
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.989051769012476e-06,
+      "loss": 0.16,
+      "step": 47675
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.988715933706111e-06,
+      "loss": 0.1292,
+      "step": 47676
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.988380098399745e-06,
+      "loss": 0.1559,
+      "step": 47677
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.988044263093379e-06,
+      "loss": 0.1535,
+      "step": 47678
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.987708427787014e-06,
+      "loss": 0.175,
+      "step": 47679
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.987372592480648e-06,
+      "loss": 0.2392,
+      "step": 47680
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.987036757174282e-06,
+      "loss": 0.3895,
+      "step": 47681
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.986700921867916e-06,
+      "loss": 0.2452,
+      "step": 47682
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.986365086561551e-06,
+      "loss": 0.2378,
+      "step": 47683
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.986029251255185e-06,
+      "loss": 0.2701,
+      "step": 47684
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.985693415948819e-06,
+      "loss": 0.5249,
+      "step": 47685
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.985357580642453e-06,
+      "loss": 0.202,
+      "step": 47686
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.985021745336087e-06,
+      "loss": 0.2365,
+      "step": 47687
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.984685910029722e-06,
+      "loss": 0.3082,
+      "step": 47688
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.984350074723356e-06,
+      "loss": 0.3182,
+      "step": 47689
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.98401423941699e-06,
+      "loss": 0.1771,
+      "step": 47690
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.983678404110625e-06,
+      "loss": 0.2365,
+      "step": 47691
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.9833425688042585e-06,
+      "loss": 0.1974,
+      "step": 47692
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.983006733497893e-06,
+      "loss": 0.2752,
+      "step": 47693
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.982670898191528e-06,
+      "loss": 0.3517,
+      "step": 47694
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.982335062885162e-06,
+      "loss": 0.3358,
+      "step": 47695
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.981999227578795e-06,
+      "loss": 0.3327,
+      "step": 47696
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.98166339227243e-06,
+      "loss": 0.3016,
+      "step": 47697
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.981327556966064e-06,
+      "loss": 0.1808,
+      "step": 47698
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.9809917216596985e-06,
+      "loss": 0.2694,
+      "step": 47699
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.980655886353333e-06,
+      "loss": 0.1943,
+      "step": 47700
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.980320051046967e-06,
+      "loss": 0.2022,
+      "step": 47701
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.979984215740601e-06,
+      "loss": 0.2479,
+      "step": 47702
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.979648380434235e-06,
+      "loss": 0.2085,
+      "step": 47703
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.97931254512787e-06,
+      "loss": 0.2348,
+      "step": 47704
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.978976709821504e-06,
+      "loss": 0.263,
+      "step": 47705
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.9786408745151385e-06,
+      "loss": 0.1763,
+      "step": 47706
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.978305039208772e-06,
+      "loss": 0.2844,
+      "step": 47707
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.977969203902406e-06,
+      "loss": 0.2007,
+      "step": 47708
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.977633368596041e-06,
+      "loss": 0.2995,
+      "step": 47709
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.977297533289675e-06,
+      "loss": 0.1367,
+      "step": 47710
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.976961697983309e-06,
+      "loss": 0.3607,
+      "step": 47711
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.976625862676943e-06,
+      "loss": 0.1746,
+      "step": 47712
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.976290027370578e-06,
+      "loss": 0.253,
+      "step": 47713
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.975954192064212e-06,
+      "loss": 0.3312,
+      "step": 47714
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.975618356757846e-06,
+      "loss": 0.2736,
+      "step": 47715
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.975282521451481e-06,
+      "loss": 0.2932,
+      "step": 47716
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.9749466861451145e-06,
+      "loss": 0.1459,
+      "step": 47717
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.974610850838749e-06,
+      "loss": 0.3521,
+      "step": 47718
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.974275015532383e-06,
+      "loss": 0.4261,
+      "step": 47719
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.973939180226018e-06,
+      "loss": 0.2843,
+      "step": 47720
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.9736033449196514e-06,
+      "loss": 0.2491,
+      "step": 47721
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.973267509613286e-06,
+      "loss": 0.3395,
+      "step": 47722
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.97293167430692e-06,
+      "loss": 0.2476,
+      "step": 47723
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.9725958390005545e-06,
+      "loss": 0.3549,
+      "step": 47724
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.972260003694189e-06,
+      "loss": 0.3307,
+      "step": 47725
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.971924168387823e-06,
+      "loss": 0.4043,
+      "step": 47726
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.971588333081457e-06,
+      "loss": 0.1697,
+      "step": 47727
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.971252497775091e-06,
+      "loss": 0.2127,
+      "step": 47728
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.970916662468726e-06,
+      "loss": 0.2524,
+      "step": 47729
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.97058082716236e-06,
+      "loss": 0.3117,
+      "step": 47730
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.9702449918559945e-06,
+      "loss": 0.1992,
+      "step": 47731
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.969909156549628e-06,
+      "loss": 0.1603,
+      "step": 47732
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.969573321243262e-06,
+      "loss": 0.3971,
+      "step": 47733
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.969237485936897e-06,
+      "loss": 0.2144,
+      "step": 47734
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.968901650630531e-06,
+      "loss": 0.1712,
+      "step": 47735
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.968565815324165e-06,
+      "loss": 0.2907,
+      "step": 47736
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.9682299800178e-06,
+      "loss": 0.1455,
+      "step": 47737
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.967894144711434e-06,
+      "loss": 0.3989,
+      "step": 47738
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.967558309405068e-06,
+      "loss": 0.1883,
+      "step": 47739
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.967222474098703e-06,
+      "loss": 0.3084,
+      "step": 47740
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 3.966886638792337e-06,
+      "loss": 0.3132,
+      "step": 47741
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9665508034859705e-06,
+      "loss": 0.2783,
+      "step": 47742
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.966214968179605e-06,
+      "loss": 0.3366,
+      "step": 47743
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.965879132873239e-06,
+      "loss": 0.1913,
+      "step": 47744
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.965543297566874e-06,
+      "loss": 0.3854,
+      "step": 47745
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.965207462260508e-06,
+      "loss": 0.4129,
+      "step": 47746
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.964871626954142e-06,
+      "loss": 0.3415,
+      "step": 47747
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.964535791647776e-06,
+      "loss": 0.3823,
+      "step": 47748
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9641999563414105e-06,
+      "loss": 0.2204,
+      "step": 47749
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.963864121035045e-06,
+      "loss": 0.1989,
+      "step": 47750
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.963528285728679e-06,
+      "loss": 0.2057,
+      "step": 47751
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.963192450422314e-06,
+      "loss": 0.1799,
+      "step": 47752
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9628566151159474e-06,
+      "loss": 0.2526,
+      "step": 47753
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.962520779809581e-06,
+      "loss": 0.2754,
+      "step": 47754
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.962184944503216e-06,
+      "loss": 0.163,
+      "step": 47755
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9618491091968505e-06,
+      "loss": 0.325,
+      "step": 47756
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.961513273890484e-06,
+      "loss": 0.3713,
+      "step": 47757
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.961177438584118e-06,
+      "loss": 0.2024,
+      "step": 47758
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.960841603277753e-06,
+      "loss": 0.2087,
+      "step": 47759
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9605057679713874e-06,
+      "loss": 0.1974,
+      "step": 47760
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.960169932665021e-06,
+      "loss": 0.1885,
+      "step": 47761
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.959834097358656e-06,
+      "loss": 0.2874,
+      "step": 47762
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.95949826205229e-06,
+      "loss": 0.1868,
+      "step": 47763
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9591624267459235e-06,
+      "loss": 0.4277,
+      "step": 47764
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.958826591439558e-06,
+      "loss": 0.1692,
+      "step": 47765
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.958490756133193e-06,
+      "loss": 0.3394,
+      "step": 47766
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9581549208268266e-06,
+      "loss": 0.4881,
+      "step": 47767
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.957819085520461e-06,
+      "loss": 0.4303,
+      "step": 47768
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.957483250214095e-06,
+      "loss": 0.1946,
+      "step": 47769
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.95714741490773e-06,
+      "loss": 0.3476,
+      "step": 47770
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.956811579601364e-06,
+      "loss": 0.2035,
+      "step": 47771
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.956475744294998e-06,
+      "loss": 0.3434,
+      "step": 47772
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.956139908988632e-06,
+      "loss": 0.1921,
+      "step": 47773
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9558040736822666e-06,
+      "loss": 0.3064,
+      "step": 47774
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9554682383759e-06,
+      "loss": 0.2835,
+      "step": 47775
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.955132403069535e-06,
+      "loss": 0.2389,
+      "step": 47776
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.95479656776317e-06,
+      "loss": 0.2601,
+      "step": 47777
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9544607324568035e-06,
+      "loss": 0.2264,
+      "step": 47778
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.954124897150437e-06,
+      "loss": 0.3625,
+      "step": 47779
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.953789061844072e-06,
+      "loss": 0.198,
+      "step": 47780
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9534532265377066e-06,
+      "loss": 0.2834,
+      "step": 47781
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.95311739123134e-06,
+      "loss": 0.2641,
+      "step": 47782
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.952781555924975e-06,
+      "loss": 0.2178,
+      "step": 47783
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.952445720618609e-06,
+      "loss": 0.2755,
+      "step": 47784
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9521098853122434e-06,
+      "loss": 0.137,
+      "step": 47785
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.951774050005877e-06,
+      "loss": 0.4808,
+      "step": 47786
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.951438214699512e-06,
+      "loss": 0.3194,
+      "step": 47787
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.951102379393146e-06,
+      "loss": 0.1613,
+      "step": 47788
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.95076654408678e-06,
+      "loss": 0.2242,
+      "step": 47789
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.950430708780414e-06,
+      "loss": 0.3388,
+      "step": 47790
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.950094873474049e-06,
+      "loss": 0.345,
+      "step": 47791
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9497590381676834e-06,
+      "loss": 0.152,
+      "step": 47792
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.949423202861317e-06,
+      "loss": 0.3378,
+      "step": 47793
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.949087367554951e-06,
+      "loss": 0.2073,
+      "step": 47794
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.948751532248586e-06,
+      "loss": 0.2988,
+      "step": 47795
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.94841569694222e-06,
+      "loss": 0.2889,
+      "step": 47796
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.948079861635854e-06,
+      "loss": 0.2234,
+      "step": 47797
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.947744026329489e-06,
+      "loss": 0.226,
+      "step": 47798
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.947408191023123e-06,
+      "loss": 0.2804,
+      "step": 47799
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.947072355716756e-06,
+      "loss": 0.2925,
+      "step": 47800
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.946736520410391e-06,
+      "loss": 0.219,
+      "step": 47801
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.946400685104026e-06,
+      "loss": 0.272,
+      "step": 47802
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9460648497976595e-06,
+      "loss": 0.2423,
+      "step": 47803
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.945729014491293e-06,
+      "loss": 0.4464,
+      "step": 47804
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.945393179184928e-06,
+      "loss": 0.412,
+      "step": 47805
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9450573438785626e-06,
+      "loss": 0.2602,
+      "step": 47806
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.944721508572196e-06,
+      "loss": 0.2682,
+      "step": 47807
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.944385673265831e-06,
+      "loss": 0.2668,
+      "step": 47808
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.944049837959465e-06,
+      "loss": 0.2919,
+      "step": 47809
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.943714002653099e-06,
+      "loss": 0.2902,
+      "step": 47810
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.943378167346733e-06,
+      "loss": 0.3276,
+      "step": 47811
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.943042332040368e-06,
+      "loss": 0.2424,
+      "step": 47812
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.942706496734002e-06,
+      "loss": 0.31,
+      "step": 47813
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.942370661427636e-06,
+      "loss": 0.3221,
+      "step": 47814
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.94203482612127e-06,
+      "loss": 0.1843,
+      "step": 47815
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.941698990814905e-06,
+      "loss": 0.395,
+      "step": 47816
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9413631555085395e-06,
+      "loss": 0.1913,
+      "step": 47817
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.941027320202173e-06,
+      "loss": 0.3106,
+      "step": 47818
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.940691484895807e-06,
+      "loss": 0.156,
+      "step": 47819
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.940355649589442e-06,
+      "loss": 0.264,
+      "step": 47820
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9400198142830755e-06,
+      "loss": 0.3806,
+      "step": 47821
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.93968397897671e-06,
+      "loss": 0.2479,
+      "step": 47822
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.939348143670345e-06,
+      "loss": 0.4306,
+      "step": 47823
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.939012308363979e-06,
+      "loss": 0.3012,
+      "step": 47824
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.938676473057612e-06,
+      "loss": 0.4089,
+      "step": 47825
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.938340637751247e-06,
+      "loss": 0.2183,
+      "step": 47826
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.938004802444882e-06,
+      "loss": 0.209,
+      "step": 47827
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9376689671385155e-06,
+      "loss": 0.146,
+      "step": 47828
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.93733313183215e-06,
+      "loss": 0.2253,
+      "step": 47829
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.936997296525784e-06,
+      "loss": 0.3021,
+      "step": 47830
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.936661461219419e-06,
+      "loss": 0.25,
+      "step": 47831
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.936325625913052e-06,
+      "loss": 0.3915,
+      "step": 47832
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.935989790606687e-06,
+      "loss": 0.3305,
+      "step": 47833
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.935653955300321e-06,
+      "loss": 0.2424,
+      "step": 47834
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9353181199939555e-06,
+      "loss": 0.093,
+      "step": 47835
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.934982284687589e-06,
+      "loss": 0.2381,
+      "step": 47836
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.934646449381224e-06,
+      "loss": 0.4823,
+      "step": 47837
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.934310614074859e-06,
+      "loss": 0.4433,
+      "step": 47838
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.933974778768492e-06,
+      "loss": 0.2533,
+      "step": 47839
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.933638943462126e-06,
+      "loss": 0.4204,
+      "step": 47840
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.933303108155761e-06,
+      "loss": 0.33,
+      "step": 47841
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9329672728493955e-06,
+      "loss": 0.2006,
+      "step": 47842
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.932631437543029e-06,
+      "loss": 0.1692,
+      "step": 47843
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.932295602236664e-06,
+      "loss": 0.218,
+      "step": 47844
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.931959766930298e-06,
+      "loss": 0.2961,
+      "step": 47845
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9316239316239315e-06,
+      "loss": 0.3758,
+      "step": 47846
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.931288096317566e-06,
+      "loss": 0.2751,
+      "step": 47847
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.930952261011201e-06,
+      "loss": 0.2783,
+      "step": 47848
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.930616425704835e-06,
+      "loss": 0.2657,
+      "step": 47849
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.930280590398468e-06,
+      "loss": 0.333,
+      "step": 47850
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.929944755092103e-06,
+      "loss": 0.0863,
+      "step": 47851
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.929608919785738e-06,
+      "loss": 0.1358,
+      "step": 47852
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9292730844793715e-06,
+      "loss": 0.2401,
+      "step": 47853
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.928937249173006e-06,
+      "loss": 0.1268,
+      "step": 47854
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.92860141386664e-06,
+      "loss": 0.1817,
+      "step": 47855
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.928265578560274e-06,
+      "loss": 0.3483,
+      "step": 47856
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.927929743253908e-06,
+      "loss": 0.2543,
+      "step": 47857
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.927593907947543e-06,
+      "loss": 0.3373,
+      "step": 47858
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.927258072641177e-06,
+      "loss": 0.2752,
+      "step": 47859
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9269222373348115e-06,
+      "loss": 0.1696,
+      "step": 47860
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.926586402028445e-06,
+      "loss": 0.2885,
+      "step": 47861
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.92625056672208e-06,
+      "loss": 0.4674,
+      "step": 47862
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.925914731415715e-06,
+      "loss": 0.1675,
+      "step": 47863
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.925578896109348e-06,
+      "loss": 0.2225,
+      "step": 47864
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.925243060802982e-06,
+      "loss": 0.2762,
+      "step": 47865
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.924907225496617e-06,
+      "loss": 0.3194,
+      "step": 47866
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.924571390190251e-06,
+      "loss": 0.2304,
+      "step": 47867
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.924235554883885e-06,
+      "loss": 0.3358,
+      "step": 47868
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.92389971957752e-06,
+      "loss": 0.2204,
+      "step": 47869
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.923563884271154e-06,
+      "loss": 0.188,
+      "step": 47870
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9232280489647875e-06,
+      "loss": 0.2508,
+      "step": 47871
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.922892213658422e-06,
+      "loss": 0.2061,
+      "step": 47872
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.922556378352057e-06,
+      "loss": 0.3303,
+      "step": 47873
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.922220543045691e-06,
+      "loss": 0.4337,
+      "step": 47874
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.921884707739325e-06,
+      "loss": 0.2111,
+      "step": 47875
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.921548872432959e-06,
+      "loss": 0.2257,
+      "step": 47876
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.921213037126593e-06,
+      "loss": 0.2797,
+      "step": 47877
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9208772018202275e-06,
+      "loss": 0.3496,
+      "step": 47878
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.920541366513862e-06,
+      "loss": 0.1645,
+      "step": 47879
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.920205531207496e-06,
+      "loss": 0.1518,
+      "step": 47880
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.919869695901131e-06,
+      "loss": 0.2422,
+      "step": 47881
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9195338605947644e-06,
+      "loss": 0.1498,
+      "step": 47882
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.919198025288399e-06,
+      "loss": 0.4364,
+      "step": 47883
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.918862189982034e-06,
+      "loss": 0.1176,
+      "step": 47884
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9185263546756675e-06,
+      "loss": 0.3603,
+      "step": 47885
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.918190519369301e-06,
+      "loss": 0.4343,
+      "step": 47886
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.917854684062936e-06,
+      "loss": 0.2141,
+      "step": 47887
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.91751884875657e-06,
+      "loss": 0.3515,
+      "step": 47888
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.917183013450204e-06,
+      "loss": 0.291,
+      "step": 47889
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.916847178143839e-06,
+      "loss": 0.105,
+      "step": 47890
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.916511342837473e-06,
+      "loss": 0.1295,
+      "step": 47891
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.916175507531107e-06,
+      "loss": 0.3274,
+      "step": 47892
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.915839672224741e-06,
+      "loss": 0.2916,
+      "step": 47893
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.915503836918376e-06,
+      "loss": 0.2891,
+      "step": 47894
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.91516800161201e-06,
+      "loss": 0.1977,
+      "step": 47895
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9148321663056436e-06,
+      "loss": 0.4817,
+      "step": 47896
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.914496330999278e-06,
+      "loss": 0.2697,
+      "step": 47897
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.914160495692913e-06,
+      "loss": 0.1842,
+      "step": 47898
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.913824660386547e-06,
+      "loss": 0.4021,
+      "step": 47899
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.913488825080181e-06,
+      "loss": 0.1555,
+      "step": 47900
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.913152989773815e-06,
+      "loss": 0.2159,
+      "step": 47901
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.912817154467449e-06,
+      "loss": 0.3456,
+      "step": 47902
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9124813191610836e-06,
+      "loss": 0.338,
+      "step": 47903
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.912145483854718e-06,
+      "loss": 0.1651,
+      "step": 47904
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.911809648548352e-06,
+      "loss": 0.1354,
+      "step": 47905
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.911473813241987e-06,
+      "loss": 0.263,
+      "step": 47906
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9111379779356205e-06,
+      "loss": 0.2593,
+      "step": 47907
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.910802142629255e-06,
+      "loss": 0.2871,
+      "step": 47908
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.91046630732289e-06,
+      "loss": 0.1016,
+      "step": 47909
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9101304720165235e-06,
+      "loss": 0.218,
+      "step": 47910
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.909794636710157e-06,
+      "loss": 0.2353,
+      "step": 47911
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.909458801403792e-06,
+      "loss": 0.3375,
+      "step": 47912
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.909122966097426e-06,
+      "loss": 0.2584,
+      "step": 47913
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9087871307910604e-06,
+      "loss": 0.1854,
+      "step": 47914
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.908451295484695e-06,
+      "loss": 0.3299,
+      "step": 47915
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.908115460178329e-06,
+      "loss": 0.2467,
+      "step": 47916
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.907779624871963e-06,
+      "loss": 0.2426,
+      "step": 47917
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.907443789565597e-06,
+      "loss": 0.3967,
+      "step": 47918
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.907107954259232e-06,
+      "loss": 0.144,
+      "step": 47919
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.906772118952866e-06,
+      "loss": 0.2413,
+      "step": 47920
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9064362836465004e-06,
+      "loss": 0.4658,
+      "step": 47921
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.906100448340134e-06,
+      "loss": 0.2572,
+      "step": 47922
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.905764613033768e-06,
+      "loss": 0.3699,
+      "step": 47923
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.905428777727403e-06,
+      "loss": 0.341,
+      "step": 47924
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.905092942421037e-06,
+      "loss": 0.1597,
+      "step": 47925
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.904757107114671e-06,
+      "loss": 0.1733,
+      "step": 47926
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.904421271808306e-06,
+      "loss": 0.3418,
+      "step": 47927
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.90408543650194e-06,
+      "loss": 0.2937,
+      "step": 47928
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.903749601195574e-06,
+      "loss": 0.3834,
+      "step": 47929
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.903413765889209e-06,
+      "loss": 0.2324,
+      "step": 47930
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.903077930582843e-06,
+      "loss": 0.5015,
+      "step": 47931
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9027420952764765e-06,
+      "loss": 0.2417,
+      "step": 47932
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.902406259970111e-06,
+      "loss": 0.1721,
+      "step": 47933
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.902070424663745e-06,
+      "loss": 0.232,
+      "step": 47934
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9017345893573796e-06,
+      "loss": 0.3866,
+      "step": 47935
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.901398754051013e-06,
+      "loss": 0.2147,
+      "step": 47936
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.901062918744648e-06,
+      "loss": 0.2648,
+      "step": 47937
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.900727083438282e-06,
+      "loss": 0.1839,
+      "step": 47938
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.9003912481319165e-06,
+      "loss": 0.3709,
+      "step": 47939
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 3.900055412825551e-06,
+      "loss": 0.1991,
+      "step": 47940
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.899719577519185e-06,
+      "loss": 0.3196,
+      "step": 47941
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.899383742212819e-06,
+      "loss": 0.2436,
+      "step": 47942
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.899047906906453e-06,
+      "loss": 0.2078,
+      "step": 47943
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.898712071600087e-06,
+      "loss": 0.2188,
+      "step": 47944
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.898376236293722e-06,
+      "loss": 0.3265,
+      "step": 47945
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8980404009873565e-06,
+      "loss": 0.3274,
+      "step": 47946
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.89770456568099e-06,
+      "loss": 0.1505,
+      "step": 47947
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.897368730374624e-06,
+      "loss": 0.1266,
+      "step": 47948
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.897032895068259e-06,
+      "loss": 0.2454,
+      "step": 47949
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.896697059761893e-06,
+      "loss": 0.3006,
+      "step": 47950
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.896361224455527e-06,
+      "loss": 0.1013,
+      "step": 47951
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.896025389149162e-06,
+      "loss": 0.3366,
+      "step": 47952
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.895689553842796e-06,
+      "loss": 0.2948,
+      "step": 47953
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.89535371853643e-06,
+      "loss": 0.1473,
+      "step": 47954
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.895017883230064e-06,
+      "loss": 0.2681,
+      "step": 47955
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.894682047923699e-06,
+      "loss": 0.3305,
+      "step": 47956
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8943462126173325e-06,
+      "loss": 0.1505,
+      "step": 47957
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.894010377310967e-06,
+      "loss": 0.3068,
+      "step": 47958
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.893674542004601e-06,
+      "loss": 0.1231,
+      "step": 47959
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.893338706698236e-06,
+      "loss": 0.3637,
+      "step": 47960
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.89300287139187e-06,
+      "loss": 0.2297,
+      "step": 47961
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.892667036085504e-06,
+      "loss": 0.3314,
+      "step": 47962
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.892331200779138e-06,
+      "loss": 0.2794,
+      "step": 47963
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8919953654727725e-06,
+      "loss": 0.238,
+      "step": 47964
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.891659530166407e-06,
+      "loss": 0.2526,
+      "step": 47965
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.891323694860041e-06,
+      "loss": 0.2718,
+      "step": 47966
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.890987859553676e-06,
+      "loss": 0.1936,
+      "step": 47967
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.890652024247309e-06,
+      "loss": 0.3071,
+      "step": 47968
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.890316188940943e-06,
+      "loss": 0.3079,
+      "step": 47969
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.889980353634578e-06,
+      "loss": 0.2747,
+      "step": 47970
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8896445183282125e-06,
+      "loss": 0.3985,
+      "step": 47971
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.889308683021846e-06,
+      "loss": 0.1856,
+      "step": 47972
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.888972847715481e-06,
+      "loss": 0.2025,
+      "step": 47973
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.888637012409115e-06,
+      "loss": 0.3033,
+      "step": 47974
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.888301177102749e-06,
+      "loss": 0.3753,
+      "step": 47975
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.887965341796384e-06,
+      "loss": 0.1623,
+      "step": 47976
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.887629506490018e-06,
+      "loss": 0.2814,
+      "step": 47977
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.887293671183652e-06,
+      "loss": 0.2872,
+      "step": 47978
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.886957835877286e-06,
+      "loss": 0.2048,
+      "step": 47979
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.88662200057092e-06,
+      "loss": 0.2399,
+      "step": 47980
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.886286165264555e-06,
+      "loss": 0.2462,
+      "step": 47981
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8859503299581885e-06,
+      "loss": 0.4181,
+      "step": 47982
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.885614494651823e-06,
+      "loss": 0.189,
+      "step": 47983
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.885278659345457e-06,
+      "loss": 0.3743,
+      "step": 47984
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.884942824039092e-06,
+      "loss": 0.2587,
+      "step": 47985
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.884606988732726e-06,
+      "loss": 0.2724,
+      "step": 47986
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.88427115342636e-06,
+      "loss": 0.2216,
+      "step": 47987
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.883935318119994e-06,
+      "loss": 0.2884,
+      "step": 47988
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8835994828136285e-06,
+      "loss": 0.2571,
+      "step": 47989
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.883263647507262e-06,
+      "loss": 0.2751,
+      "step": 47990
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.882927812200897e-06,
+      "loss": 0.1735,
+      "step": 47991
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.882591976894532e-06,
+      "loss": 0.3423,
+      "step": 47992
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.882256141588165e-06,
+      "loss": 0.2157,
+      "step": 47993
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.881920306281799e-06,
+      "loss": 0.1953,
+      "step": 47994
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.881584470975434e-06,
+      "loss": 0.235,
+      "step": 47995
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8812486356690685e-06,
+      "loss": 0.2447,
+      "step": 47996
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.880912800362702e-06,
+      "loss": 0.1546,
+      "step": 47997
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.880576965056337e-06,
+      "loss": 0.3435,
+      "step": 47998
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.880241129749971e-06,
+      "loss": 0.2443,
+      "step": 47999
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.879905294443605e-06,
+      "loss": 0.3289,
+      "step": 48000
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.879569459137239e-06,
+      "loss": 0.2929,
+      "step": 48001
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.879233623830874e-06,
+      "loss": 0.1949,
+      "step": 48002
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.878897788524508e-06,
+      "loss": 0.2624,
+      "step": 48003
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.878561953218142e-06,
+      "loss": 0.2101,
+      "step": 48004
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.878226117911776e-06,
+      "loss": 0.2743,
+      "step": 48005
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.877890282605411e-06,
+      "loss": 0.234,
+      "step": 48006
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.877554447299045e-06,
+      "loss": 0.3105,
+      "step": 48007
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.877218611992679e-06,
+      "loss": 0.2906,
+      "step": 48008
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.876882776686313e-06,
+      "loss": 0.1859,
+      "step": 48009
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.876546941379948e-06,
+      "loss": 0.2933,
+      "step": 48010
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.876211106073582e-06,
+      "loss": 0.2345,
+      "step": 48011
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.875875270767216e-06,
+      "loss": 0.4965,
+      "step": 48012
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.875539435460851e-06,
+      "loss": 0.4539,
+      "step": 48013
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8752036001544845e-06,
+      "loss": 0.4574,
+      "step": 48014
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.874867764848118e-06,
+      "loss": 0.1402,
+      "step": 48015
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.874531929541753e-06,
+      "loss": 0.2315,
+      "step": 48016
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.874196094235388e-06,
+      "loss": 0.393,
+      "step": 48017
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.873860258929021e-06,
+      "loss": 0.2265,
+      "step": 48018
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.873524423622656e-06,
+      "loss": 0.2929,
+      "step": 48019
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.87318858831629e-06,
+      "loss": 0.2513,
+      "step": 48020
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8728527530099245e-06,
+      "loss": 0.2655,
+      "step": 48021
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.872516917703559e-06,
+      "loss": 0.3005,
+      "step": 48022
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.872181082397193e-06,
+      "loss": 0.2564,
+      "step": 48023
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.871845247090827e-06,
+      "loss": 0.2076,
+      "step": 48024
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.871509411784461e-06,
+      "loss": 0.1878,
+      "step": 48025
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.871173576478095e-06,
+      "loss": 0.2549,
+      "step": 48026
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.87083774117173e-06,
+      "loss": 0.2468,
+      "step": 48027
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.870501905865364e-06,
+      "loss": 0.4072,
+      "step": 48028
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.870166070558998e-06,
+      "loss": 0.4257,
+      "step": 48029
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.869830235252632e-06,
+      "loss": 0.1601,
+      "step": 48030
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.869494399946267e-06,
+      "loss": 0.5043,
+      "step": 48031
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.869158564639901e-06,
+      "loss": 0.1167,
+      "step": 48032
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.868822729333535e-06,
+      "loss": 0.2201,
+      "step": 48033
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.868486894027169e-06,
+      "loss": 0.2182,
+      "step": 48034
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.868151058720804e-06,
+      "loss": 0.2694,
+      "step": 48035
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8678152234144374e-06,
+      "loss": 0.1893,
+      "step": 48036
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.867479388108072e-06,
+      "loss": 0.417,
+      "step": 48037
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.867143552801707e-06,
+      "loss": 0.2882,
+      "step": 48038
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8668077174953405e-06,
+      "loss": 0.3158,
+      "step": 48039
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.866471882188974e-06,
+      "loss": 0.3192,
+      "step": 48040
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.866136046882609e-06,
+      "loss": 0.41,
+      "step": 48041
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.865800211576244e-06,
+      "loss": 0.114,
+      "step": 48042
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8654643762698774e-06,
+      "loss": 0.2187,
+      "step": 48043
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.865128540963512e-06,
+      "loss": 0.3711,
+      "step": 48044
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.864792705657146e-06,
+      "loss": 0.1839,
+      "step": 48045
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.86445687035078e-06,
+      "loss": 0.2007,
+      "step": 48046
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.864121035044414e-06,
+      "loss": 0.3542,
+      "step": 48047
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.863785199738049e-06,
+      "loss": 0.1537,
+      "step": 48048
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.863449364431683e-06,
+      "loss": 0.2653,
+      "step": 48049
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8631135291253174e-06,
+      "loss": 0.2231,
+      "step": 48050
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.862777693818951e-06,
+      "loss": 0.1604,
+      "step": 48051
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.862441858512586e-06,
+      "loss": 0.3114,
+      "step": 48052
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8621060232062205e-06,
+      "loss": 0.3196,
+      "step": 48053
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.861770187899854e-06,
+      "loss": 0.1795,
+      "step": 48054
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.861434352593488e-06,
+      "loss": 0.1332,
+      "step": 48055
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.861098517287123e-06,
+      "loss": 0.2298,
+      "step": 48056
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8607626819807566e-06,
+      "loss": 0.2703,
+      "step": 48057
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.860426846674391e-06,
+      "loss": 0.2293,
+      "step": 48058
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.860091011368026e-06,
+      "loss": 0.2076,
+      "step": 48059
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.85975517606166e-06,
+      "loss": 0.424,
+      "step": 48060
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8594193407552935e-06,
+      "loss": 0.2081,
+      "step": 48061
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.859083505448928e-06,
+      "loss": 0.3942,
+      "step": 48062
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.858747670142563e-06,
+      "loss": 0.2647,
+      "step": 48063
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8584118348361966e-06,
+      "loss": 0.2318,
+      "step": 48064
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.858075999529831e-06,
+      "loss": 0.2859,
+      "step": 48065
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.857740164223465e-06,
+      "loss": 0.3518,
+      "step": 48066
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8574043289171e-06,
+      "loss": 0.1909,
+      "step": 48067
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8570684936107335e-06,
+      "loss": 0.1844,
+      "step": 48068
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.856732658304368e-06,
+      "loss": 0.2417,
+      "step": 48069
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.856396822998002e-06,
+      "loss": 0.3535,
+      "step": 48070
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8560609876916366e-06,
+      "loss": 0.3259,
+      "step": 48071
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.85572515238527e-06,
+      "loss": 0.2969,
+      "step": 48072
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.855389317078905e-06,
+      "loss": 0.2115,
+      "step": 48073
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.855053481772539e-06,
+      "loss": 0.2953,
+      "step": 48074
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8547176464661735e-06,
+      "loss": 0.2703,
+      "step": 48075
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.854381811159807e-06,
+      "loss": 0.3131,
+      "step": 48076
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.854045975853442e-06,
+      "loss": 0.4227,
+      "step": 48077
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8537101405470765e-06,
+      "loss": 0.3797,
+      "step": 48078
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.85337430524071e-06,
+      "loss": 0.3592,
+      "step": 48079
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.853038469934344e-06,
+      "loss": 0.2085,
+      "step": 48080
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.852702634627979e-06,
+      "loss": 0.1741,
+      "step": 48081
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.852366799321613e-06,
+      "loss": 0.4345,
+      "step": 48082
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.852030964015247e-06,
+      "loss": 0.4142,
+      "step": 48083
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.851695128708882e-06,
+      "loss": 0.3451,
+      "step": 48084
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.851359293402516e-06,
+      "loss": 0.2584,
+      "step": 48085
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8510234580961495e-06,
+      "loss": 0.1632,
+      "step": 48086
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.850687622789784e-06,
+      "loss": 0.3464,
+      "step": 48087
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.850351787483419e-06,
+      "loss": 0.1511,
+      "step": 48088
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.850015952177053e-06,
+      "loss": 0.3047,
+      "step": 48089
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.849680116870687e-06,
+      "loss": 0.1891,
+      "step": 48090
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.849344281564321e-06,
+      "loss": 0.3838,
+      "step": 48091
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.849008446257955e-06,
+      "loss": 0.2627,
+      "step": 48092
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8486726109515895e-06,
+      "loss": 0.2994,
+      "step": 48093
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.848336775645224e-06,
+      "loss": 0.1523,
+      "step": 48094
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.848000940338858e-06,
+      "loss": 0.4869,
+      "step": 48095
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8476651050324926e-06,
+      "loss": 0.1668,
+      "step": 48096
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.847329269726126e-06,
+      "loss": 0.4364,
+      "step": 48097
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.846993434419761e-06,
+      "loss": 0.4041,
+      "step": 48098
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.846657599113396e-06,
+      "loss": 0.2079,
+      "step": 48099
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8463217638070295e-06,
+      "loss": 0.1804,
+      "step": 48100
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.845985928500663e-06,
+      "loss": 0.1993,
+      "step": 48101
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.845650093194298e-06,
+      "loss": 0.2426,
+      "step": 48102
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.845314257887932e-06,
+      "loss": 0.395,
+      "step": 48103
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.844978422581566e-06,
+      "loss": 0.2771,
+      "step": 48104
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.844642587275201e-06,
+      "loss": 0.3661,
+      "step": 48105
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.844306751968835e-06,
+      "loss": 0.3151,
+      "step": 48106
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.843970916662469e-06,
+      "loss": 0.2995,
+      "step": 48107
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.843635081356103e-06,
+      "loss": 0.3149,
+      "step": 48108
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.843299246049738e-06,
+      "loss": 0.3737,
+      "step": 48109
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.842963410743372e-06,
+      "loss": 0.205,
+      "step": 48110
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.842627575437006e-06,
+      "loss": 0.1388,
+      "step": 48111
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.84229174013064e-06,
+      "loss": 0.1736,
+      "step": 48112
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.841955904824275e-06,
+      "loss": 0.3565,
+      "step": 48113
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.841620069517909e-06,
+      "loss": 0.2956,
+      "step": 48114
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.841284234211543e-06,
+      "loss": 0.5245,
+      "step": 48115
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.840948398905177e-06,
+      "loss": 0.3682,
+      "step": 48116
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.840612563598811e-06,
+      "loss": 0.2752,
+      "step": 48117
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8402767282924455e-06,
+      "loss": 0.2407,
+      "step": 48118
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.83994089298608e-06,
+      "loss": 0.2692,
+      "step": 48119
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.839605057679714e-06,
+      "loss": 0.3636,
+      "step": 48120
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.839269222373349e-06,
+      "loss": 0.2637,
+      "step": 48121
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.838933387066982e-06,
+      "loss": 0.2507,
+      "step": 48122
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.838597551760617e-06,
+      "loss": 0.423,
+      "step": 48123
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.838261716454252e-06,
+      "loss": 0.2202,
+      "step": 48124
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.8379258811478855e-06,
+      "loss": 0.2933,
+      "step": 48125
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.837590045841519e-06,
+      "loss": 0.2554,
+      "step": 48126
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.837254210535154e-06,
+      "loss": 0.1882,
+      "step": 48127
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.836918375228788e-06,
+      "loss": 0.2934,
+      "step": 48128
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.836582539922422e-06,
+      "loss": 0.2525,
+      "step": 48129
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.836246704616057e-06,
+      "loss": 0.2442,
+      "step": 48130
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.835910869309691e-06,
+      "loss": 0.2867,
+      "step": 48131
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.835575034003325e-06,
+      "loss": 0.1985,
+      "step": 48132
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.835239198696959e-06,
+      "loss": 0.3498,
+      "step": 48133
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.834903363390594e-06,
+      "loss": 0.3407,
+      "step": 48134
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.834567528084228e-06,
+      "loss": 0.405,
+      "step": 48135
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.834231692777862e-06,
+      "loss": 0.2558,
+      "step": 48136
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.833895857471496e-06,
+      "loss": 0.2624,
+      "step": 48137
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 3.83356002216513e-06,
+      "loss": 0.2238,
+      "step": 48138
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.833224186858765e-06,
+      "loss": 0.1956,
+      "step": 48139
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.832888351552399e-06,
+      "loss": 0.3651,
+      "step": 48140
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.832552516246033e-06,
+      "loss": 0.2192,
+      "step": 48141
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.832216680939668e-06,
+      "loss": 0.2508,
+      "step": 48142
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8318808456333015e-06,
+      "loss": 0.2792,
+      "step": 48143
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.831545010326936e-06,
+      "loss": 0.1472,
+      "step": 48144
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.831209175020571e-06,
+      "loss": 0.2384,
+      "step": 48145
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.830873339714205e-06,
+      "loss": 0.3082,
+      "step": 48146
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.830537504407838e-06,
+      "loss": 0.1257,
+      "step": 48147
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.830201669101473e-06,
+      "loss": 0.3119,
+      "step": 48148
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.829865833795107e-06,
+      "loss": 0.1549,
+      "step": 48149
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8295299984887415e-06,
+      "loss": 0.4211,
+      "step": 48150
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.829194163182376e-06,
+      "loss": 0.3629,
+      "step": 48151
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.82885832787601e-06,
+      "loss": 0.3557,
+      "step": 48152
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.828522492569644e-06,
+      "loss": 0.2894,
+      "step": 48153
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.828186657263278e-06,
+      "loss": 0.2386,
+      "step": 48154
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.827850821956913e-06,
+      "loss": 0.3526,
+      "step": 48155
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.827514986650547e-06,
+      "loss": 0.2328,
+      "step": 48156
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8271791513441815e-06,
+      "loss": 0.3212,
+      "step": 48157
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.826843316037815e-06,
+      "loss": 0.3905,
+      "step": 48158
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.826507480731449e-06,
+      "loss": 0.2837,
+      "step": 48159
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.826171645425084e-06,
+      "loss": 0.1589,
+      "step": 48160
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.825835810118718e-06,
+      "loss": 0.244,
+      "step": 48161
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.825499974812352e-06,
+      "loss": 0.1173,
+      "step": 48162
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.825164139505986e-06,
+      "loss": 0.1855,
+      "step": 48163
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.824828304199621e-06,
+      "loss": 0.208,
+      "step": 48164
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.824492468893255e-06,
+      "loss": 0.2076,
+      "step": 48165
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.824156633586889e-06,
+      "loss": 0.0965,
+      "step": 48166
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.823820798280524e-06,
+      "loss": 0.3926,
+      "step": 48167
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8234849629741575e-06,
+      "loss": 0.3073,
+      "step": 48168
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.823149127667792e-06,
+      "loss": 0.2492,
+      "step": 48169
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.822813292361426e-06,
+      "loss": 0.2979,
+      "step": 48170
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.822477457055061e-06,
+      "loss": 0.1488,
+      "step": 48171
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8221416217486944e-06,
+      "loss": 0.355,
+      "step": 48172
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.821805786442329e-06,
+      "loss": 0.2695,
+      "step": 48173
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.821469951135963e-06,
+      "loss": 0.3142,
+      "step": 48174
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8211341158295975e-06,
+      "loss": 0.1772,
+      "step": 48175
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.820798280523232e-06,
+      "loss": 0.5015,
+      "step": 48176
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.820462445216866e-06,
+      "loss": 0.3177,
+      "step": 48177
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8201266099105e-06,
+      "loss": 0.302,
+      "step": 48178
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8197907746041344e-06,
+      "loss": 0.2898,
+      "step": 48179
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.819454939297769e-06,
+      "loss": 0.2115,
+      "step": 48180
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.819119103991403e-06,
+      "loss": 0.1695,
+      "step": 48181
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8187832686850375e-06,
+      "loss": 0.26,
+      "step": 48182
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.818447433378671e-06,
+      "loss": 0.1954,
+      "step": 48183
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.818111598072305e-06,
+      "loss": 0.2477,
+      "step": 48184
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.81777576276594e-06,
+      "loss": 0.283,
+      "step": 48185
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.817439927459574e-06,
+      "loss": 0.3412,
+      "step": 48186
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.817104092153208e-06,
+      "loss": 0.2035,
+      "step": 48187
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.816768256846843e-06,
+      "loss": 0.2893,
+      "step": 48188
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.816432421540477e-06,
+      "loss": 0.3852,
+      "step": 48189
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.816096586234111e-06,
+      "loss": 0.5131,
+      "step": 48190
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.815760750927746e-06,
+      "loss": 0.279,
+      "step": 48191
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.81542491562138e-06,
+      "loss": 0.4277,
+      "step": 48192
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8150890803150136e-06,
+      "loss": 0.2672,
+      "step": 48193
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.814753245008648e-06,
+      "loss": 0.3105,
+      "step": 48194
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8144174097022824e-06,
+      "loss": 0.3144,
+      "step": 48195
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8140815743959162e-06,
+      "loss": 0.1667,
+      "step": 48196
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.813745739089551e-06,
+      "loss": 0.2827,
+      "step": 48197
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.813409903783185e-06,
+      "loss": 0.2008,
+      "step": 48198
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8130740684768193e-06,
+      "loss": 0.3221,
+      "step": 48199
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.812738233170454e-06,
+      "loss": 0.3253,
+      "step": 48200
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8124023978640878e-06,
+      "loss": 0.2442,
+      "step": 48201
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.812066562557722e-06,
+      "loss": 0.3478,
+      "step": 48202
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8117307272513566e-06,
+      "loss": 0.1871,
+      "step": 48203
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8113948919449904e-06,
+      "loss": 0.5092,
+      "step": 48204
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8110590566386247e-06,
+      "loss": 0.3272,
+      "step": 48205
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8107232213322593e-06,
+      "loss": 0.2765,
+      "step": 48206
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.810387386025893e-06,
+      "loss": 0.2782,
+      "step": 48207
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8100515507195273e-06,
+      "loss": 0.2097,
+      "step": 48208
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8097157154131616e-06,
+      "loss": 0.222,
+      "step": 48209
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8093798801067962e-06,
+      "loss": 0.3583,
+      "step": 48210
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.80904404480043e-06,
+      "loss": 0.2804,
+      "step": 48211
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8087082094940642e-06,
+      "loss": 0.326,
+      "step": 48212
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.808372374187699e-06,
+      "loss": 0.1978,
+      "step": 48213
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8080365388813327e-06,
+      "loss": 0.245,
+      "step": 48214
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.807700703574967e-06,
+      "loss": 0.2196,
+      "step": 48215
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8073648682686016e-06,
+      "loss": 0.1065,
+      "step": 48216
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8070290329622358e-06,
+      "loss": 0.2994,
+      "step": 48217
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8066931976558696e-06,
+      "loss": 0.1757,
+      "step": 48218
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8063573623495042e-06,
+      "loss": 0.4066,
+      "step": 48219
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8060215270431385e-06,
+      "loss": 0.117,
+      "step": 48220
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8056856917367723e-06,
+      "loss": 0.2851,
+      "step": 48221
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.805349856430407e-06,
+      "loss": 0.3982,
+      "step": 48222
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.805014021124041e-06,
+      "loss": 0.1364,
+      "step": 48223
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.804678185817675e-06,
+      "loss": 0.2605,
+      "step": 48224
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8043423505113096e-06,
+      "loss": 0.4093,
+      "step": 48225
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.804006515204944e-06,
+      "loss": 0.1724,
+      "step": 48226
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.803670679898578e-06,
+      "loss": 0.3455,
+      "step": 48227
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8033348445922127e-06,
+      "loss": 0.2442,
+      "step": 48228
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8029990092858465e-06,
+      "loss": 0.2198,
+      "step": 48229
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8026631739794807e-06,
+      "loss": 0.2297,
+      "step": 48230
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8023273386731153e-06,
+      "loss": 0.3098,
+      "step": 48231
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.801991503366749e-06,
+      "loss": 0.176,
+      "step": 48232
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8016556680603834e-06,
+      "loss": 0.2522,
+      "step": 48233
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.801319832754018e-06,
+      "loss": 0.1178,
+      "step": 48234
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.800983997447652e-06,
+      "loss": 0.1899,
+      "step": 48235
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.800648162141286e-06,
+      "loss": 0.2429,
+      "step": 48236
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.8003123268349207e-06,
+      "loss": 0.2701,
+      "step": 48237
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.799976491528555e-06,
+      "loss": 0.3156,
+      "step": 48238
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7996406562221887e-06,
+      "loss": 0.2254,
+      "step": 48239
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7993048209158234e-06,
+      "loss": 0.3248,
+      "step": 48240
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7989689856094576e-06,
+      "loss": 0.1749,
+      "step": 48241
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7986331503030914e-06,
+      "loss": 0.3414,
+      "step": 48242
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.798297314996726e-06,
+      "loss": 0.3218,
+      "step": 48243
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7979614796903602e-06,
+      "loss": 0.2959,
+      "step": 48244
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7976256443839945e-06,
+      "loss": 0.4086,
+      "step": 48245
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7972898090776287e-06,
+      "loss": 0.3178,
+      "step": 48246
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.796953973771263e-06,
+      "loss": 0.3035,
+      "step": 48247
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.796618138464897e-06,
+      "loss": 0.3033,
+      "step": 48248
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.796282303158532e-06,
+      "loss": 0.3585,
+      "step": 48249
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7959464678521656e-06,
+      "loss": 0.1407,
+      "step": 48250
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7956106325458e-06,
+      "loss": 0.2037,
+      "step": 48251
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7952747972394336e-06,
+      "loss": 0.2651,
+      "step": 48252
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7949389619330683e-06,
+      "loss": 0.2807,
+      "step": 48253
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7946031266267025e-06,
+      "loss": 0.4164,
+      "step": 48254
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7942672913203367e-06,
+      "loss": 0.2143,
+      "step": 48255
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7939314560139714e-06,
+      "loss": 0.1759,
+      "step": 48256
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.793595620707605e-06,
+      "loss": 0.3137,
+      "step": 48257
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7932597854012394e-06,
+      "loss": 0.3712,
+      "step": 48258
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.792923950094874e-06,
+      "loss": 0.186,
+      "step": 48259
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.792588114788508e-06,
+      "loss": 0.377,
+      "step": 48260
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.792252279482142e-06,
+      "loss": 0.0905,
+      "step": 48261
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7919164441757767e-06,
+      "loss": 0.2291,
+      "step": 48262
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7915806088694105e-06,
+      "loss": 0.272,
+      "step": 48263
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7912447735630447e-06,
+      "loss": 0.3007,
+      "step": 48264
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7909089382566794e-06,
+      "loss": 0.3154,
+      "step": 48265
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7905731029503136e-06,
+      "loss": 0.234,
+      "step": 48266
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7902372676439474e-06,
+      "loss": 0.2949,
+      "step": 48267
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.789901432337582e-06,
+      "loss": 0.2335,
+      "step": 48268
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7895655970312163e-06,
+      "loss": 0.2838,
+      "step": 48269
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.78922976172485e-06,
+      "loss": 0.4326,
+      "step": 48270
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7888939264184847e-06,
+      "loss": 0.3733,
+      "step": 48271
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.788558091112119e-06,
+      "loss": 0.2653,
+      "step": 48272
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.788222255805753e-06,
+      "loss": 0.1562,
+      "step": 48273
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7878864204993874e-06,
+      "loss": 0.132,
+      "step": 48274
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7875505851930216e-06,
+      "loss": 0.2411,
+      "step": 48275
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.787214749886656e-06,
+      "loss": 0.1261,
+      "step": 48276
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7868789145802905e-06,
+      "loss": 0.2276,
+      "step": 48277
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7865430792739243e-06,
+      "loss": 0.2136,
+      "step": 48278
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7862072439675585e-06,
+      "loss": 0.255,
+      "step": 48279
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.785871408661193e-06,
+      "loss": 0.373,
+      "step": 48280
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.785535573354827e-06,
+      "loss": 0.1841,
+      "step": 48281
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.785199738048461e-06,
+      "loss": 0.3083,
+      "step": 48282
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.784863902742096e-06,
+      "loss": 0.2172,
+      "step": 48283
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.78452806743573e-06,
+      "loss": 0.2434,
+      "step": 48284
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.784192232129364e-06,
+      "loss": 0.3061,
+      "step": 48285
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7838563968229985e-06,
+      "loss": 0.279,
+      "step": 48286
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7835205615166327e-06,
+      "loss": 0.1958,
+      "step": 48287
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7831847262102665e-06,
+      "loss": 0.2358,
+      "step": 48288
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.782848890903901e-06,
+      "loss": 0.2277,
+      "step": 48289
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7825130555975354e-06,
+      "loss": 0.3185,
+      "step": 48290
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.782177220291169e-06,
+      "loss": 0.2808,
+      "step": 48291
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.781841384984804e-06,
+      "loss": 0.176,
+      "step": 48292
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.781505549678438e-06,
+      "loss": 0.3118,
+      "step": 48293
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7811697143720723e-06,
+      "loss": 0.2512,
+      "step": 48294
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.780833879065707e-06,
+      "loss": 0.2034,
+      "step": 48295
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7804980437593407e-06,
+      "loss": 0.3306,
+      "step": 48296
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.780162208452975e-06,
+      "loss": 0.2224,
+      "step": 48297
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7798263731466088e-06,
+      "loss": 0.3467,
+      "step": 48298
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7794905378402434e-06,
+      "loss": 0.215,
+      "step": 48299
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7791547025338776e-06,
+      "loss": 0.2176,
+      "step": 48300
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.778818867227512e-06,
+      "loss": 0.265,
+      "step": 48301
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7784830319211465e-06,
+      "loss": 0.3273,
+      "step": 48302
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7781471966147803e-06,
+      "loss": 0.2402,
+      "step": 48303
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7778113613084145e-06,
+      "loss": 0.2119,
+      "step": 48304
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.777475526002049e-06,
+      "loss": 0.2512,
+      "step": 48305
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.777139690695683e-06,
+      "loss": 0.2005,
+      "step": 48306
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.776803855389317e-06,
+      "loss": 0.3106,
+      "step": 48307
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.776468020082952e-06,
+      "loss": 0.1548,
+      "step": 48308
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7761321847765856e-06,
+      "loss": 0.2052,
+      "step": 48309
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.77579634947022e-06,
+      "loss": 0.1786,
+      "step": 48310
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7754605141638545e-06,
+      "loss": 0.2636,
+      "step": 48311
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7751246788574887e-06,
+      "loss": 0.3492,
+      "step": 48312
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7747888435511225e-06,
+      "loss": 0.2644,
+      "step": 48313
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.774453008244757e-06,
+      "loss": 0.2044,
+      "step": 48314
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7741171729383914e-06,
+      "loss": 0.3921,
+      "step": 48315
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7737813376320252e-06,
+      "loss": 0.2477,
+      "step": 48316
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.77344550232566e-06,
+      "loss": 0.3083,
+      "step": 48317
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.773109667019294e-06,
+      "loss": 0.224,
+      "step": 48318
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7727738317129283e-06,
+      "loss": 0.2366,
+      "step": 48319
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7724379964065625e-06,
+      "loss": 0.2778,
+      "step": 48320
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7721021611001968e-06,
+      "loss": 0.2993,
+      "step": 48321
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.771766325793831e-06,
+      "loss": 0.4138,
+      "step": 48322
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7714304904874656e-06,
+      "loss": 0.3608,
+      "step": 48323
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7710946551810994e-06,
+      "loss": 0.3627,
+      "step": 48324
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7707588198747337e-06,
+      "loss": 0.2347,
+      "step": 48325
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7704229845683683e-06,
+      "loss": 0.2678,
+      "step": 48326
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.770087149262002e-06,
+      "loss": 0.1393,
+      "step": 48327
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7697513139556363e-06,
+      "loss": 0.1681,
+      "step": 48328
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.769415478649271e-06,
+      "loss": 0.1247,
+      "step": 48329
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.769079643342905e-06,
+      "loss": 0.251,
+      "step": 48330
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.768743808036539e-06,
+      "loss": 0.228,
+      "step": 48331
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7684079727301736e-06,
+      "loss": 0.3362,
+      "step": 48332
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.768072137423808e-06,
+      "loss": 0.4693,
+      "step": 48333
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7677363021174417e-06,
+      "loss": 0.3064,
+      "step": 48334
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7674004668110763e-06,
+      "loss": 0.154,
+      "step": 48335
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7670646315047105e-06,
+      "loss": 0.1874,
+      "step": 48336
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 3.7667287961983443e-06,
+      "loss": 0.2849,
+      "step": 48337
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.766392960891979e-06,
+      "loss": 0.2367,
+      "step": 48338
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.766057125585613e-06,
+      "loss": 0.1061,
+      "step": 48339
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7657212902792474e-06,
+      "loss": 0.2338,
+      "step": 48340
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.765385454972882e-06,
+      "loss": 0.1618,
+      "step": 48341
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.765049619666516e-06,
+      "loss": 0.3768,
+      "step": 48342
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.76471378436015e-06,
+      "loss": 0.3592,
+      "step": 48343
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.764377949053784e-06,
+      "loss": 0.2779,
+      "step": 48344
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7640421137474186e-06,
+      "loss": 0.3986,
+      "step": 48345
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7637062784410528e-06,
+      "loss": 0.301,
+      "step": 48346
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.763370443134687e-06,
+      "loss": 0.1713,
+      "step": 48347
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7630346078283212e-06,
+      "loss": 0.2306,
+      "step": 48348
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7626987725219554e-06,
+      "loss": 0.3225,
+      "step": 48349
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7623629372155897e-06,
+      "loss": 0.2229,
+      "step": 48350
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7620271019092243e-06,
+      "loss": 0.3245,
+      "step": 48351
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.761691266602858e-06,
+      "loss": 0.2776,
+      "step": 48352
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7613554312964923e-06,
+      "loss": 0.1695,
+      "step": 48353
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.761019595990127e-06,
+      "loss": 0.2911,
+      "step": 48354
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.760683760683761e-06,
+      "loss": 0.3014,
+      "step": 48355
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.760347925377395e-06,
+      "loss": 0.3113,
+      "step": 48356
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7600120900710297e-06,
+      "loss": 0.2069,
+      "step": 48357
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.759676254764664e-06,
+      "loss": 0.4,
+      "step": 48358
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7593404194582977e-06,
+      "loss": 0.1278,
+      "step": 48359
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7590045841519323e-06,
+      "loss": 0.4095,
+      "step": 48360
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7586687488455666e-06,
+      "loss": 0.1903,
+      "step": 48361
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7583329135392004e-06,
+      "loss": 0.3207,
+      "step": 48362
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.757997078232835e-06,
+      "loss": 0.1725,
+      "step": 48363
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7576612429264692e-06,
+      "loss": 0.3776,
+      "step": 48364
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.757325407620103e-06,
+      "loss": 0.2871,
+      "step": 48365
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7569895723137377e-06,
+      "loss": 0.2277,
+      "step": 48366
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.756653737007372e-06,
+      "loss": 0.3008,
+      "step": 48367
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.756317901701006e-06,
+      "loss": 0.1614,
+      "step": 48368
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7559820663946408e-06,
+      "loss": 0.2521,
+      "step": 48369
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7556462310882746e-06,
+      "loss": 0.247,
+      "step": 48370
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.755310395781909e-06,
+      "loss": 0.4963,
+      "step": 48371
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7549745604755434e-06,
+      "loss": 0.3153,
+      "step": 48372
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7546387251691772e-06,
+      "loss": 0.1865,
+      "step": 48373
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7543028898628115e-06,
+      "loss": 0.2543,
+      "step": 48374
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.753967054556446e-06,
+      "loss": 0.2182,
+      "step": 48375
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.75363121925008e-06,
+      "loss": 0.297,
+      "step": 48376
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.753295383943714e-06,
+      "loss": 0.354,
+      "step": 48377
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.752959548637349e-06,
+      "loss": 0.3939,
+      "step": 48378
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.752623713330983e-06,
+      "loss": 0.24,
+      "step": 48379
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.752287878024617e-06,
+      "loss": 0.2465,
+      "step": 48380
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7519520427182515e-06,
+      "loss": 0.2414,
+      "step": 48381
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7516162074118857e-06,
+      "loss": 0.3358,
+      "step": 48382
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7512803721055195e-06,
+      "loss": 0.1736,
+      "step": 48383
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.750944536799154e-06,
+      "loss": 0.2766,
+      "step": 48384
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7506087014927884e-06,
+      "loss": 0.2924,
+      "step": 48385
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7502728661864226e-06,
+      "loss": 0.2621,
+      "step": 48386
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7499370308800564e-06,
+      "loss": 0.2818,
+      "step": 48387
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.749601195573691e-06,
+      "loss": 0.4885,
+      "step": 48388
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7492653602673253e-06,
+      "loss": 0.288,
+      "step": 48389
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.748929524960959e-06,
+      "loss": 0.2348,
+      "step": 48390
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7485936896545937e-06,
+      "loss": 0.3244,
+      "step": 48391
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.748257854348228e-06,
+      "loss": 0.2171,
+      "step": 48392
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7479220190418617e-06,
+      "loss": 0.3295,
+      "step": 48393
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7475861837354964e-06,
+      "loss": 0.2984,
+      "step": 48394
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7472503484291306e-06,
+      "loss": 0.2575,
+      "step": 48395
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.746914513122765e-06,
+      "loss": 0.2301,
+      "step": 48396
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7465786778163995e-06,
+      "loss": 0.2525,
+      "step": 48397
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7462428425100333e-06,
+      "loss": 0.5209,
+      "step": 48398
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7459070072036675e-06,
+      "loss": 0.2388,
+      "step": 48399
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.745571171897302e-06,
+      "loss": 0.2799,
+      "step": 48400
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.745235336590936e-06,
+      "loss": 0.225,
+      "step": 48401
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.74489950128457e-06,
+      "loss": 0.3032,
+      "step": 48402
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.744563665978205e-06,
+      "loss": 0.1962,
+      "step": 48403
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7442278306718386e-06,
+      "loss": 0.2944,
+      "step": 48404
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.743891995365473e-06,
+      "loss": 0.3137,
+      "step": 48405
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7435561600591075e-06,
+      "loss": 0.3525,
+      "step": 48406
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7432203247527417e-06,
+      "loss": 0.2132,
+      "step": 48407
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7428844894463755e-06,
+      "loss": 0.3226,
+      "step": 48408
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.74254865414001e-06,
+      "loss": 0.349,
+      "step": 48409
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7422128188336444e-06,
+      "loss": 0.3422,
+      "step": 48410
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.741876983527278e-06,
+      "loss": 0.3119,
+      "step": 48411
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.741541148220913e-06,
+      "loss": 0.2759,
+      "step": 48412
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.741205312914547e-06,
+      "loss": 0.3738,
+      "step": 48413
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7408694776081813e-06,
+      "loss": 0.2414,
+      "step": 48414
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7405336423018155e-06,
+      "loss": 0.4339,
+      "step": 48415
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7401978069954497e-06,
+      "loss": 0.2166,
+      "step": 48416
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.739861971689084e-06,
+      "loss": 0.2774,
+      "step": 48417
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7395261363827186e-06,
+      "loss": 0.3391,
+      "step": 48418
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7391903010763524e-06,
+      "loss": 0.3123,
+      "step": 48419
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7388544657699866e-06,
+      "loss": 0.1935,
+      "step": 48420
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7385186304636213e-06,
+      "loss": 0.2191,
+      "step": 48421
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.738182795157255e-06,
+      "loss": 0.3465,
+      "step": 48422
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7378469598508893e-06,
+      "loss": 0.2734,
+      "step": 48423
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.737511124544524e-06,
+      "loss": 0.1966,
+      "step": 48424
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.737175289238158e-06,
+      "loss": 0.2451,
+      "step": 48425
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.736839453931792e-06,
+      "loss": 0.2633,
+      "step": 48426
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7365036186254266e-06,
+      "loss": 0.2569,
+      "step": 48427
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.736167783319061e-06,
+      "loss": 0.5104,
+      "step": 48428
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7358319480126946e-06,
+      "loss": 0.3236,
+      "step": 48429
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7354961127063293e-06,
+      "loss": 0.286,
+      "step": 48430
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7351602773999635e-06,
+      "loss": 0.2854,
+      "step": 48431
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7348244420935977e-06,
+      "loss": 0.2474,
+      "step": 48432
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7344886067872315e-06,
+      "loss": 0.2427,
+      "step": 48433
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.734152771480866e-06,
+      "loss": 0.3179,
+      "step": 48434
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7338169361745004e-06,
+      "loss": 0.2074,
+      "step": 48435
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.733481100868134e-06,
+      "loss": 0.3092,
+      "step": 48436
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.733145265561769e-06,
+      "loss": 0.3759,
+      "step": 48437
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.732809430255403e-06,
+      "loss": 0.2376,
+      "step": 48438
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.732473594949037e-06,
+      "loss": 0.1926,
+      "step": 48439
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7321377596426715e-06,
+      "loss": 0.1381,
+      "step": 48440
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7318019243363057e-06,
+      "loss": 0.3015,
+      "step": 48441
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.73146608902994e-06,
+      "loss": 0.4119,
+      "step": 48442
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7311302537235746e-06,
+      "loss": 0.2573,
+      "step": 48443
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7307944184172084e-06,
+      "loss": 0.2536,
+      "step": 48444
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7304585831108426e-06,
+      "loss": 0.2546,
+      "step": 48445
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7301227478044773e-06,
+      "loss": 0.108,
+      "step": 48446
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.729786912498111e-06,
+      "loss": 0.1472,
+      "step": 48447
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7294510771917453e-06,
+      "loss": 0.284,
+      "step": 48448
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.72911524188538e-06,
+      "loss": 0.4255,
+      "step": 48449
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7287794065790138e-06,
+      "loss": 0.329,
+      "step": 48450
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.728443571272648e-06,
+      "loss": 0.4291,
+      "step": 48451
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7281077359662826e-06,
+      "loss": 0.315,
+      "step": 48452
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.727771900659917e-06,
+      "loss": 0.3024,
+      "step": 48453
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7274360653535506e-06,
+      "loss": 0.2179,
+      "step": 48454
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7271002300471853e-06,
+      "loss": 0.174,
+      "step": 48455
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7267643947408195e-06,
+      "loss": 0.2486,
+      "step": 48456
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7264285594344533e-06,
+      "loss": 0.2145,
+      "step": 48457
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.726092724128088e-06,
+      "loss": 0.2449,
+      "step": 48458
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.725756888821722e-06,
+      "loss": 0.333,
+      "step": 48459
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7254210535153564e-06,
+      "loss": 0.3291,
+      "step": 48460
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7250852182089906e-06,
+      "loss": 0.143,
+      "step": 48461
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.724749382902625e-06,
+      "loss": 0.1364,
+      "step": 48462
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.724413547596259e-06,
+      "loss": 0.2957,
+      "step": 48463
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7240777122898937e-06,
+      "loss": 0.2921,
+      "step": 48464
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7237418769835275e-06,
+      "loss": 0.3181,
+      "step": 48465
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7234060416771618e-06,
+      "loss": 0.3232,
+      "step": 48466
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7230702063707964e-06,
+      "loss": 0.2318,
+      "step": 48467
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.72273437106443e-06,
+      "loss": 0.2322,
+      "step": 48468
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7223985357580644e-06,
+      "loss": 0.305,
+      "step": 48469
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.722062700451699e-06,
+      "loss": 0.2825,
+      "step": 48470
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7217268651453333e-06,
+      "loss": 0.2395,
+      "step": 48471
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.721391029838967e-06,
+      "loss": 0.3307,
+      "step": 48472
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7210551945326017e-06,
+      "loss": 0.1863,
+      "step": 48473
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.720719359226236e-06,
+      "loss": 0.32,
+      "step": 48474
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7203835239198698e-06,
+      "loss": 0.3033,
+      "step": 48475
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7200476886135044e-06,
+      "loss": 0.2253,
+      "step": 48476
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7197118533071386e-06,
+      "loss": 0.2368,
+      "step": 48477
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7193760180007724e-06,
+      "loss": 0.4409,
+      "step": 48478
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7190401826944067e-06,
+      "loss": 0.443,
+      "step": 48479
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7187043473880413e-06,
+      "loss": 0.3623,
+      "step": 48480
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7183685120816755e-06,
+      "loss": 0.2451,
+      "step": 48481
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7180326767753093e-06,
+      "loss": 0.2096,
+      "step": 48482
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.717696841468944e-06,
+      "loss": 0.2805,
+      "step": 48483
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7173610061625782e-06,
+      "loss": 0.3035,
+      "step": 48484
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.717025170856212e-06,
+      "loss": 0.2341,
+      "step": 48485
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7166893355498467e-06,
+      "loss": 0.2506,
+      "step": 48486
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.716353500243481e-06,
+      "loss": 0.1958,
+      "step": 48487
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.716017664937115e-06,
+      "loss": 0.301,
+      "step": 48488
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7156818296307493e-06,
+      "loss": 0.2592,
+      "step": 48489
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7153459943243836e-06,
+      "loss": 0.4201,
+      "step": 48490
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7150101590180178e-06,
+      "loss": 0.2353,
+      "step": 48491
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7146743237116524e-06,
+      "loss": 0.3725,
+      "step": 48492
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7143384884052862e-06,
+      "loss": 0.3918,
+      "step": 48493
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7140026530989205e-06,
+      "loss": 0.2123,
+      "step": 48494
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.713666817792555e-06,
+      "loss": 0.3395,
+      "step": 48495
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.713330982486189e-06,
+      "loss": 0.3844,
+      "step": 48496
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.712995147179823e-06,
+      "loss": 0.2432,
+      "step": 48497
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7126593118734578e-06,
+      "loss": 0.1937,
+      "step": 48498
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.712323476567092e-06,
+      "loss": 0.2215,
+      "step": 48499
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.711987641260726e-06,
+      "loss": 0.1393,
+      "step": 48500
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7116518059543604e-06,
+      "loss": 0.3082,
+      "step": 48501
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7113159706479947e-06,
+      "loss": 0.3724,
+      "step": 48502
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7109801353416285e-06,
+      "loss": 0.2891,
+      "step": 48503
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.710644300035263e-06,
+      "loss": 0.2679,
+      "step": 48504
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7103084647288973e-06,
+      "loss": 0.2042,
+      "step": 48505
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.709972629422531e-06,
+      "loss": 0.3693,
+      "step": 48506
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7096367941161658e-06,
+      "loss": 0.159,
+      "step": 48507
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7093009588098e-06,
+      "loss": 0.3517,
+      "step": 48508
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7089651235034342e-06,
+      "loss": 0.3471,
+      "step": 48509
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.708629288197069e-06,
+      "loss": 0.274,
+      "step": 48510
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7082934528907027e-06,
+      "loss": 0.2723,
+      "step": 48511
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.707957617584337e-06,
+      "loss": 0.2846,
+      "step": 48512
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7076217822779716e-06,
+      "loss": 0.2122,
+      "step": 48513
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7072859469716054e-06,
+      "loss": 0.163,
+      "step": 48514
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7069501116652396e-06,
+      "loss": 0.2189,
+      "step": 48515
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7066142763588742e-06,
+      "loss": 0.256,
+      "step": 48516
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.706278441052508e-06,
+      "loss": 0.2467,
+      "step": 48517
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7059426057461422e-06,
+      "loss": 0.2294,
+      "step": 48518
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.705606770439777e-06,
+      "loss": 0.2323,
+      "step": 48519
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.705270935133411e-06,
+      "loss": 0.2616,
+      "step": 48520
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.704935099827045e-06,
+      "loss": 0.205,
+      "step": 48521
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7045992645206796e-06,
+      "loss": 0.2055,
+      "step": 48522
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.704263429214314e-06,
+      "loss": 0.2252,
+      "step": 48523
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7039275939079476e-06,
+      "loss": 0.1936,
+      "step": 48524
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.703591758601582e-06,
+      "loss": 0.2935,
+      "step": 48525
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7032559232952165e-06,
+      "loss": 0.3824,
+      "step": 48526
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7029200879888507e-06,
+      "loss": 0.1827,
+      "step": 48527
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7025842526824845e-06,
+      "loss": 0.1774,
+      "step": 48528
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.702248417376119e-06,
+      "loss": 0.287,
+      "step": 48529
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7019125820697534e-06,
+      "loss": 0.3516,
+      "step": 48530
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.701576746763387e-06,
+      "loss": 0.2465,
+      "step": 48531
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.701240911457022e-06,
+      "loss": 0.2505,
+      "step": 48532
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.700905076150656e-06,
+      "loss": 0.2061,
+      "step": 48533
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.70056924084429e-06,
+      "loss": 0.2278,
+      "step": 48534
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 3.7002334055379245e-06,
+      "loss": 0.3425,
+      "step": 48535
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6998975702315587e-06,
+      "loss": 0.3998,
+      "step": 48536
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.699561734925193e-06,
+      "loss": 0.2363,
+      "step": 48537
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6992258996188276e-06,
+      "loss": 0.1823,
+      "step": 48538
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6988900643124614e-06,
+      "loss": 0.2636,
+      "step": 48539
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6985542290060956e-06,
+      "loss": 0.2161,
+      "step": 48540
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6982183936997302e-06,
+      "loss": 0.255,
+      "step": 48541
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.697882558393364e-06,
+      "loss": 0.212,
+      "step": 48542
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6975467230869983e-06,
+      "loss": 0.2358,
+      "step": 48543
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.697210887780633e-06,
+      "loss": 0.2424,
+      "step": 48544
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6968750524742667e-06,
+      "loss": 0.1855,
+      "step": 48545
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.696539217167901e-06,
+      "loss": 0.1607,
+      "step": 48546
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6962033818615356e-06,
+      "loss": 0.3297,
+      "step": 48547
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.69586754655517e-06,
+      "loss": 0.2848,
+      "step": 48548
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6955317112488036e-06,
+      "loss": 0.1207,
+      "step": 48549
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6951958759424383e-06,
+      "loss": 0.2274,
+      "step": 48550
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6948600406360725e-06,
+      "loss": 0.1961,
+      "step": 48551
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6945242053297063e-06,
+      "loss": 0.2994,
+      "step": 48552
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.694188370023341e-06,
+      "loss": 0.1007,
+      "step": 48553
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.693852534716975e-06,
+      "loss": 0.2043,
+      "step": 48554
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6935166994106094e-06,
+      "loss": 0.2566,
+      "step": 48555
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6931808641042436e-06,
+      "loss": 0.2668,
+      "step": 48556
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.692845028797878e-06,
+      "loss": 0.3012,
+      "step": 48557
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.692509193491512e-06,
+      "loss": 0.2543,
+      "step": 48558
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6921733581851467e-06,
+      "loss": 0.3434,
+      "step": 48559
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6918375228787805e-06,
+      "loss": 0.3805,
+      "step": 48560
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6915016875724147e-06,
+      "loss": 0.1113,
+      "step": 48561
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6911658522660494e-06,
+      "loss": 0.2889,
+      "step": 48562
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.690830016959683e-06,
+      "loss": 0.4104,
+      "step": 48563
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6904941816533174e-06,
+      "loss": 0.0874,
+      "step": 48564
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.690158346346952e-06,
+      "loss": 0.232,
+      "step": 48565
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6898225110405863e-06,
+      "loss": 0.1204,
+      "step": 48566
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.68948667573422e-06,
+      "loss": 0.3251,
+      "step": 48567
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6891508404278543e-06,
+      "loss": 0.3043,
+      "step": 48568
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.688815005121489e-06,
+      "loss": 0.219,
+      "step": 48569
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6884791698151227e-06,
+      "loss": 0.1456,
+      "step": 48570
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.688143334508757e-06,
+      "loss": 0.4205,
+      "step": 48571
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6878074992023916e-06,
+      "loss": 0.2799,
+      "step": 48572
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.687471663896026e-06,
+      "loss": 0.2391,
+      "step": 48573
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6871358285896596e-06,
+      "loss": 0.2902,
+      "step": 48574
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6867999932832943e-06,
+      "loss": 0.1845,
+      "step": 48575
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6864641579769285e-06,
+      "loss": 0.2178,
+      "step": 48576
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6861283226705623e-06,
+      "loss": 0.1877,
+      "step": 48577
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.685792487364197e-06,
+      "loss": 0.3968,
+      "step": 48578
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.685456652057831e-06,
+      "loss": 0.1673,
+      "step": 48579
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.685120816751465e-06,
+      "loss": 0.2829,
+      "step": 48580
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6847849814450996e-06,
+      "loss": 0.1191,
+      "step": 48581
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.684449146138734e-06,
+      "loss": 0.2514,
+      "step": 48582
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.684113310832368e-06,
+      "loss": 0.3575,
+      "step": 48583
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6837774755260027e-06,
+      "loss": 0.3212,
+      "step": 48584
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6834416402196365e-06,
+      "loss": 0.1806,
+      "step": 48585
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6831058049132707e-06,
+      "loss": 0.2688,
+      "step": 48586
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6827699696069054e-06,
+      "loss": 0.3376,
+      "step": 48587
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.682434134300539e-06,
+      "loss": 0.3567,
+      "step": 48588
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6820982989941734e-06,
+      "loss": 0.3099,
+      "step": 48589
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.681762463687808e-06,
+      "loss": 0.3312,
+      "step": 48590
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.681426628381442e-06,
+      "loss": 0.2374,
+      "step": 48591
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.681090793075076e-06,
+      "loss": 0.279,
+      "step": 48592
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6807549577687107e-06,
+      "loss": 0.3185,
+      "step": 48593
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.680419122462345e-06,
+      "loss": 0.119,
+      "step": 48594
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6800832871559788e-06,
+      "loss": 0.2565,
+      "step": 48595
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6797474518496134e-06,
+      "loss": 0.4323,
+      "step": 48596
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6794116165432476e-06,
+      "loss": 0.3462,
+      "step": 48597
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6790757812368814e-06,
+      "loss": 0.4974,
+      "step": 48598
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.678739945930516e-06,
+      "loss": 0.2963,
+      "step": 48599
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6784041106241503e-06,
+      "loss": 0.1359,
+      "step": 48600
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6780682753177845e-06,
+      "loss": 0.3137,
+      "step": 48601
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6777324400114187e-06,
+      "loss": 0.2338,
+      "step": 48602
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.677396604705053e-06,
+      "loss": 0.3418,
+      "step": 48603
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.677060769398687e-06,
+      "loss": 0.15,
+      "step": 48604
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.676724934092322e-06,
+      "loss": 0.2579,
+      "step": 48605
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6763890987859556e-06,
+      "loss": 0.4759,
+      "step": 48606
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.67605326347959e-06,
+      "loss": 0.1537,
+      "step": 48607
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6757174281732245e-06,
+      "loss": 0.3896,
+      "step": 48608
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6753815928668583e-06,
+      "loss": 0.2014,
+      "step": 48609
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6750457575604925e-06,
+      "loss": 0.2131,
+      "step": 48610
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.674709922254127e-06,
+      "loss": 0.167,
+      "step": 48611
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6743740869477614e-06,
+      "loss": 0.3183,
+      "step": 48612
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.674038251641395e-06,
+      "loss": 0.2109,
+      "step": 48613
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6737024163350294e-06,
+      "loss": 0.4933,
+      "step": 48614
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.673366581028664e-06,
+      "loss": 0.1885,
+      "step": 48615
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.673030745722298e-06,
+      "loss": 0.2717,
+      "step": 48616
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.672694910415932e-06,
+      "loss": 0.412,
+      "step": 48617
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6723590751095668e-06,
+      "loss": 0.2831,
+      "step": 48618
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6720232398032006e-06,
+      "loss": 0.3742,
+      "step": 48619
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6716874044968348e-06,
+      "loss": 0.3083,
+      "step": 48620
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6713515691904694e-06,
+      "loss": 0.2822,
+      "step": 48621
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6710157338841036e-06,
+      "loss": 0.1823,
+      "step": 48622
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6706798985777374e-06,
+      "loss": 0.219,
+      "step": 48623
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.670344063271372e-06,
+      "loss": 0.2636,
+      "step": 48624
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6700082279650063e-06,
+      "loss": 0.2306,
+      "step": 48625
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.66967239265864e-06,
+      "loss": 0.2493,
+      "step": 48626
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6693365573522748e-06,
+      "loss": 0.1153,
+      "step": 48627
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.669000722045909e-06,
+      "loss": 0.3675,
+      "step": 48628
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6686648867395432e-06,
+      "loss": 0.2915,
+      "step": 48629
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6683290514331774e-06,
+      "loss": 0.2928,
+      "step": 48630
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6679932161268117e-06,
+      "loss": 0.2822,
+      "step": 48631
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.667657380820446e-06,
+      "loss": 0.1911,
+      "step": 48632
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6673215455140805e-06,
+      "loss": 0.2227,
+      "step": 48633
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6669857102077143e-06,
+      "loss": 0.2919,
+      "step": 48634
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6666498749013486e-06,
+      "loss": 0.2348,
+      "step": 48635
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.666314039594983e-06,
+      "loss": 0.2432,
+      "step": 48636
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.665978204288617e-06,
+      "loss": 0.2394,
+      "step": 48637
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6656423689822512e-06,
+      "loss": 0.2448,
+      "step": 48638
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.665306533675886e-06,
+      "loss": 0.2611,
+      "step": 48639
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.66497069836952e-06,
+      "loss": 0.3257,
+      "step": 48640
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.664634863063154e-06,
+      "loss": 0.4212,
+      "step": 48641
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6642990277567885e-06,
+      "loss": 0.2369,
+      "step": 48642
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6639631924504228e-06,
+      "loss": 0.3055,
+      "step": 48643
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6636273571440566e-06,
+      "loss": 0.2236,
+      "step": 48644
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6632915218376912e-06,
+      "loss": 0.215,
+      "step": 48645
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6629556865313254e-06,
+      "loss": 0.2022,
+      "step": 48646
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6626198512249592e-06,
+      "loss": 0.2304,
+      "step": 48647
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.662284015918594e-06,
+      "loss": 0.2151,
+      "step": 48648
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.661948180612228e-06,
+      "loss": 0.3632,
+      "step": 48649
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6616123453058623e-06,
+      "loss": 0.2471,
+      "step": 48650
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.661276509999497e-06,
+      "loss": 0.2749,
+      "step": 48651
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6609406746931308e-06,
+      "loss": 0.2642,
+      "step": 48652
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.660604839386765e-06,
+      "loss": 0.2652,
+      "step": 48653
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6602690040803997e-06,
+      "loss": 0.2828,
+      "step": 48654
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6599331687740335e-06,
+      "loss": 0.4477,
+      "step": 48655
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6595973334676677e-06,
+      "loss": 0.2325,
+      "step": 48656
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6592614981613023e-06,
+      "loss": 0.1438,
+      "step": 48657
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.658925662854936e-06,
+      "loss": 0.4853,
+      "step": 48658
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6585898275485704e-06,
+      "loss": 0.2921,
+      "step": 48659
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6582539922422046e-06,
+      "loss": 0.2575,
+      "step": 48660
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6579181569358392e-06,
+      "loss": 0.2851,
+      "step": 48661
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.657582321629473e-06,
+      "loss": 0.2348,
+      "step": 48662
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6572464863231072e-06,
+      "loss": 0.2425,
+      "step": 48663
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.656910651016742e-06,
+      "loss": 0.2436,
+      "step": 48664
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6565748157103757e-06,
+      "loss": 0.2448,
+      "step": 48665
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.65623898040401e-06,
+      "loss": 0.2156,
+      "step": 48666
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6559031450976446e-06,
+      "loss": 0.2928,
+      "step": 48667
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.655567309791279e-06,
+      "loss": 0.3023,
+      "step": 48668
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6552314744849126e-06,
+      "loss": 0.2009,
+      "step": 48669
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6548956391785472e-06,
+      "loss": 0.3907,
+      "step": 48670
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6545598038721815e-06,
+      "loss": 0.2623,
+      "step": 48671
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6542239685658153e-06,
+      "loss": 0.2224,
+      "step": 48672
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.65388813325945e-06,
+      "loss": 0.2983,
+      "step": 48673
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.653552297953084e-06,
+      "loss": 0.3183,
+      "step": 48674
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.653216462646718e-06,
+      "loss": 0.2999,
+      "step": 48675
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6528806273403526e-06,
+      "loss": 0.1902,
+      "step": 48676
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.652544792033987e-06,
+      "loss": 0.264,
+      "step": 48677
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.652208956727621e-06,
+      "loss": 0.3339,
+      "step": 48678
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6518731214212557e-06,
+      "loss": 0.1769,
+      "step": 48679
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6515372861148895e-06,
+      "loss": 0.2868,
+      "step": 48680
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6512014508085237e-06,
+      "loss": 0.293,
+      "step": 48681
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6508656155021584e-06,
+      "loss": 0.2544,
+      "step": 48682
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.650529780195792e-06,
+      "loss": 0.2655,
+      "step": 48683
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6501939448894264e-06,
+      "loss": 0.4003,
+      "step": 48684
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.649858109583061e-06,
+      "loss": 0.3581,
+      "step": 48685
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.649522274276695e-06,
+      "loss": 0.1852,
+      "step": 48686
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.649186438970329e-06,
+      "loss": 0.2153,
+      "step": 48687
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6488506036639637e-06,
+      "loss": 0.3968,
+      "step": 48688
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.648514768357598e-06,
+      "loss": 0.2961,
+      "step": 48689
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6481789330512317e-06,
+      "loss": 0.1952,
+      "step": 48690
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6478430977448664e-06,
+      "loss": 0.3372,
+      "step": 48691
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6475072624385006e-06,
+      "loss": 0.248,
+      "step": 48692
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6471714271321344e-06,
+      "loss": 0.3157,
+      "step": 48693
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.646835591825769e-06,
+      "loss": 0.1978,
+      "step": 48694
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6464997565194033e-06,
+      "loss": 0.2577,
+      "step": 48695
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6461639212130375e-06,
+      "loss": 0.3284,
+      "step": 48696
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.645828085906672e-06,
+      "loss": 0.1875,
+      "step": 48697
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.645492250600306e-06,
+      "loss": 0.1944,
+      "step": 48698
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.64515641529394e-06,
+      "loss": 0.1434,
+      "step": 48699
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.644820579987575e-06,
+      "loss": 0.277,
+      "step": 48700
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6444847446812086e-06,
+      "loss": 0.4025,
+      "step": 48701
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.644148909374843e-06,
+      "loss": 0.1698,
+      "step": 48702
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.643813074068477e-06,
+      "loss": 0.1837,
+      "step": 48703
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6434772387621113e-06,
+      "loss": 0.4002,
+      "step": 48704
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6431414034557455e-06,
+      "loss": 0.4446,
+      "step": 48705
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6428055681493797e-06,
+      "loss": 0.3065,
+      "step": 48706
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6424697328430144e-06,
+      "loss": 0.2245,
+      "step": 48707
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.642133897536648e-06,
+      "loss": 0.3431,
+      "step": 48708
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6417980622302824e-06,
+      "loss": 0.364,
+      "step": 48709
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.641462226923917e-06,
+      "loss": 0.1883,
+      "step": 48710
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.641126391617551e-06,
+      "loss": 0.1831,
+      "step": 48711
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.640790556311185e-06,
+      "loss": 0.1813,
+      "step": 48712
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6404547210048197e-06,
+      "loss": 0.1719,
+      "step": 48713
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.640118885698454e-06,
+      "loss": 0.3548,
+      "step": 48714
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6397830503920877e-06,
+      "loss": 0.247,
+      "step": 48715
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6394472150857224e-06,
+      "loss": 0.2302,
+      "step": 48716
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6391113797793566e-06,
+      "loss": 0.127,
+      "step": 48717
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6387755444729904e-06,
+      "loss": 0.2482,
+      "step": 48718
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.638439709166625e-06,
+      "loss": 0.1539,
+      "step": 48719
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6381038738602593e-06,
+      "loss": 0.1584,
+      "step": 48720
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.637768038553893e-06,
+      "loss": 0.2254,
+      "step": 48721
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6374322032475277e-06,
+      "loss": 0.1391,
+      "step": 48722
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.637096367941162e-06,
+      "loss": 0.3766,
+      "step": 48723
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.636760532634796e-06,
+      "loss": 0.1478,
+      "step": 48724
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.636424697328431e-06,
+      "loss": 0.3091,
+      "step": 48725
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6360888620220646e-06,
+      "loss": 0.3225,
+      "step": 48726
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.635753026715699e-06,
+      "loss": 0.3474,
+      "step": 48727
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6354171914093335e-06,
+      "loss": 0.2319,
+      "step": 48728
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6350813561029673e-06,
+      "loss": 0.2093,
+      "step": 48729
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.6347455207966015e-06,
+      "loss": 0.267,
+      "step": 48730
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.634409685490236e-06,
+      "loss": 0.1564,
+      "step": 48731
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.63407385018387e-06,
+      "loss": 0.2787,
+      "step": 48732
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.633738014877504e-06,
+      "loss": 0.4357,
+      "step": 48733
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.633402179571139e-06,
+      "loss": 0.31,
+      "step": 48734
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.633066344264773e-06,
+      "loss": 0.293,
+      "step": 48735
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.632730508958407e-06,
+      "loss": 0.2764,
+      "step": 48736
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6323946736520415e-06,
+      "loss": 0.1738,
+      "step": 48737
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6320588383456757e-06,
+      "loss": 0.2958,
+      "step": 48738
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6317230030393095e-06,
+      "loss": 0.2354,
+      "step": 48739
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.631387167732944e-06,
+      "loss": 0.1662,
+      "step": 48740
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6310513324265784e-06,
+      "loss": 0.3911,
+      "step": 48741
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6307154971202126e-06,
+      "loss": 0.2831,
+      "step": 48742
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.630379661813847e-06,
+      "loss": 0.288,
+      "step": 48743
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.630043826507481e-06,
+      "loss": 0.4226,
+      "step": 48744
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6297079912011153e-06,
+      "loss": 0.1386,
+      "step": 48745
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.62937215589475e-06,
+      "loss": 0.212,
+      "step": 48746
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6290363205883837e-06,
+      "loss": 0.2416,
+      "step": 48747
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.628700485282018e-06,
+      "loss": 0.2474,
+      "step": 48748
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6283646499756518e-06,
+      "loss": 0.2296,
+      "step": 48749
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6280288146692864e-06,
+      "loss": 0.3216,
+      "step": 48750
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6276929793629206e-06,
+      "loss": 0.2267,
+      "step": 48751
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.627357144056555e-06,
+      "loss": 0.1929,
+      "step": 48752
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6270213087501895e-06,
+      "loss": 0.5151,
+      "step": 48753
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6266854734438233e-06,
+      "loss": 0.3381,
+      "step": 48754
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6263496381374575e-06,
+      "loss": 0.159,
+      "step": 48755
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.626013802831092e-06,
+      "loss": 0.2967,
+      "step": 48756
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.625677967524726e-06,
+      "loss": 0.2132,
+      "step": 48757
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.62534213221836e-06,
+      "loss": 0.2804,
+      "step": 48758
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.625006296911995e-06,
+      "loss": 0.2893,
+      "step": 48759
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6246704616056287e-06,
+      "loss": 0.1261,
+      "step": 48760
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.624334626299263e-06,
+      "loss": 0.2561,
+      "step": 48761
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6239987909928975e-06,
+      "loss": 0.4344,
+      "step": 48762
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6236629556865318e-06,
+      "loss": 0.1482,
+      "step": 48763
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6233271203801656e-06,
+      "loss": 0.2649,
+      "step": 48764
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6229912850738e-06,
+      "loss": 0.2383,
+      "step": 48765
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6226554497674344e-06,
+      "loss": 0.2652,
+      "step": 48766
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6223196144610682e-06,
+      "loss": 0.3384,
+      "step": 48767
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.621983779154703e-06,
+      "loss": 0.4357,
+      "step": 48768
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.621647943848337e-06,
+      "loss": 0.1514,
+      "step": 48769
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6213121085419713e-06,
+      "loss": 0.2698,
+      "step": 48770
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6209762732356055e-06,
+      "loss": 0.2203,
+      "step": 48771
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6206404379292398e-06,
+      "loss": 0.4184,
+      "step": 48772
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.620304602622874e-06,
+      "loss": 0.4891,
+      "step": 48773
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6199687673165086e-06,
+      "loss": 0.3639,
+      "step": 48774
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6196329320101424e-06,
+      "loss": 0.3765,
+      "step": 48775
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6192970967037767e-06,
+      "loss": 0.3055,
+      "step": 48776
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6189612613974113e-06,
+      "loss": 0.2106,
+      "step": 48777
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.618625426091045e-06,
+      "loss": 0.2097,
+      "step": 48778
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6182895907846793e-06,
+      "loss": 0.31,
+      "step": 48779
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.617953755478314e-06,
+      "loss": 0.2977,
+      "step": 48780
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.617617920171948e-06,
+      "loss": 0.4053,
+      "step": 48781
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.617282084865582e-06,
+      "loss": 0.3671,
+      "step": 48782
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6169462495592167e-06,
+      "loss": 0.4897,
+      "step": 48783
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.616610414252851e-06,
+      "loss": 0.3242,
+      "step": 48784
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6162745789464847e-06,
+      "loss": 0.3195,
+      "step": 48785
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6159387436401193e-06,
+      "loss": 0.2188,
+      "step": 48786
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6156029083337536e-06,
+      "loss": 0.3766,
+      "step": 48787
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6152670730273874e-06,
+      "loss": 0.3768,
+      "step": 48788
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.614931237721022e-06,
+      "loss": 0.2952,
+      "step": 48789
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6145954024146562e-06,
+      "loss": 0.3196,
+      "step": 48790
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6142595671082904e-06,
+      "loss": 0.2513,
+      "step": 48791
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.613923731801925e-06,
+      "loss": 0.4309,
+      "step": 48792
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.613587896495559e-06,
+      "loss": 0.3026,
+      "step": 48793
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.613252061189193e-06,
+      "loss": 0.2497,
+      "step": 48794
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.612916225882827e-06,
+      "loss": 0.1303,
+      "step": 48795
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6125803905764616e-06,
+      "loss": 0.4213,
+      "step": 48796
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.612244555270096e-06,
+      "loss": 0.2937,
+      "step": 48797
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.61190871996373e-06,
+      "loss": 0.248,
+      "step": 48798
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6115728846573642e-06,
+      "loss": 0.2339,
+      "step": 48799
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6112370493509985e-06,
+      "loss": 0.3107,
+      "step": 48800
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6109012140446327e-06,
+      "loss": 0.2656,
+      "step": 48801
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6105653787382673e-06,
+      "loss": 0.2051,
+      "step": 48802
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.610229543431901e-06,
+      "loss": 0.2291,
+      "step": 48803
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6098937081255354e-06,
+      "loss": 0.2359,
+      "step": 48804
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.60955787281917e-06,
+      "loss": 0.244,
+      "step": 48805
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.609222037512804e-06,
+      "loss": 0.2031,
+      "step": 48806
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.608886202206438e-06,
+      "loss": 0.2097,
+      "step": 48807
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6085503669000727e-06,
+      "loss": 0.2835,
+      "step": 48808
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.608214531593707e-06,
+      "loss": 0.271,
+      "step": 48809
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6078786962873407e-06,
+      "loss": 0.2486,
+      "step": 48810
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6075428609809753e-06,
+      "loss": 0.2126,
+      "step": 48811
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6072070256746096e-06,
+      "loss": 0.1662,
+      "step": 48812
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6068711903682434e-06,
+      "loss": 0.152,
+      "step": 48813
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.606535355061878e-06,
+      "loss": 0.2318,
+      "step": 48814
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6061995197555122e-06,
+      "loss": 0.1736,
+      "step": 48815
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.605863684449146e-06,
+      "loss": 0.3334,
+      "step": 48816
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6055278491427807e-06,
+      "loss": 0.2869,
+      "step": 48817
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.605192013836415e-06,
+      "loss": 0.2961,
+      "step": 48818
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.604856178530049e-06,
+      "loss": 0.2272,
+      "step": 48819
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6045203432236838e-06,
+      "loss": 0.3058,
+      "step": 48820
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6041845079173176e-06,
+      "loss": 0.3342,
+      "step": 48821
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.603848672610952e-06,
+      "loss": 0.2344,
+      "step": 48822
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6035128373045865e-06,
+      "loss": 0.2996,
+      "step": 48823
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6031770019982203e-06,
+      "loss": 0.1953,
+      "step": 48824
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6028411666918545e-06,
+      "loss": 0.2425,
+      "step": 48825
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.602505331385489e-06,
+      "loss": 0.2327,
+      "step": 48826
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.602169496079123e-06,
+      "loss": 0.277,
+      "step": 48827
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.601833660772757e-06,
+      "loss": 0.3302,
+      "step": 48828
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.601497825466392e-06,
+      "loss": 0.3418,
+      "step": 48829
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.601161990160026e-06,
+      "loss": 0.3081,
+      "step": 48830
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.60082615485366e-06,
+      "loss": 0.4975,
+      "step": 48831
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6004903195472945e-06,
+      "loss": 0.2645,
+      "step": 48832
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.6001544842409287e-06,
+      "loss": 0.327,
+      "step": 48833
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5998186489345625e-06,
+      "loss": 0.3027,
+      "step": 48834
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.599482813628197e-06,
+      "loss": 0.1995,
+      "step": 48835
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5991469783218314e-06,
+      "loss": 0.4374,
+      "step": 48836
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5988111430154656e-06,
+      "loss": 0.151,
+      "step": 48837
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5984753077090994e-06,
+      "loss": 0.1481,
+      "step": 48838
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.598139472402734e-06,
+      "loss": 0.2427,
+      "step": 48839
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5978036370963683e-06,
+      "loss": 0.1684,
+      "step": 48840
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.597467801790002e-06,
+      "loss": 0.3214,
+      "step": 48841
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5971319664836367e-06,
+      "loss": 0.2449,
+      "step": 48842
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.596796131177271e-06,
+      "loss": 0.2744,
+      "step": 48843
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.596460295870905e-06,
+      "loss": 0.2202,
+      "step": 48844
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5961244605645394e-06,
+      "loss": 0.1932,
+      "step": 48845
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5957886252581736e-06,
+      "loss": 0.2602,
+      "step": 48846
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.595452789951808e-06,
+      "loss": 0.1564,
+      "step": 48847
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5951169546454425e-06,
+      "loss": 0.2944,
+      "step": 48848
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5947811193390763e-06,
+      "loss": 0.2557,
+      "step": 48849
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5944452840327105e-06,
+      "loss": 0.2737,
+      "step": 48850
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.594109448726345e-06,
+      "loss": 0.3791,
+      "step": 48851
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.593773613419979e-06,
+      "loss": 0.2939,
+      "step": 48852
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.593437778113613e-06,
+      "loss": 0.2963,
+      "step": 48853
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.593101942807248e-06,
+      "loss": 0.262,
+      "step": 48854
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.592766107500882e-06,
+      "loss": 0.295,
+      "step": 48855
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.592430272194516e-06,
+      "loss": 0.322,
+      "step": 48856
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5920944368881505e-06,
+      "loss": 0.2294,
+      "step": 48857
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5917586015817847e-06,
+      "loss": 0.2077,
+      "step": 48858
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5914227662754185e-06,
+      "loss": 0.2273,
+      "step": 48859
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.591086930969053e-06,
+      "loss": 0.1928,
+      "step": 48860
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5907510956626874e-06,
+      "loss": 0.3542,
+      "step": 48861
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.590415260356321e-06,
+      "loss": 0.1961,
+      "step": 48862
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.590079425049956e-06,
+      "loss": 0.2728,
+      "step": 48863
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.58974358974359e-06,
+      "loss": 0.2627,
+      "step": 48864
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5894077544372243e-06,
+      "loss": 0.2244,
+      "step": 48865
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.589071919130859e-06,
+      "loss": 0.3467,
+      "step": 48866
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5887360838244927e-06,
+      "loss": 0.2936,
+      "step": 48867
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.588400248518127e-06,
+      "loss": 0.3229,
+      "step": 48868
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5880644132117616e-06,
+      "loss": 0.3152,
+      "step": 48869
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5877285779053954e-06,
+      "loss": 0.3762,
+      "step": 48870
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5873927425990296e-06,
+      "loss": 0.1817,
+      "step": 48871
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5870569072926643e-06,
+      "loss": 0.1796,
+      "step": 48872
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.586721071986298e-06,
+      "loss": 0.2117,
+      "step": 48873
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5863852366799323e-06,
+      "loss": 0.1673,
+      "step": 48874
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.586049401373567e-06,
+      "loss": 0.285,
+      "step": 48875
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.585713566067201e-06,
+      "loss": 0.2162,
+      "step": 48876
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.585377730760835e-06,
+      "loss": 0.3835,
+      "step": 48877
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5850418954544696e-06,
+      "loss": 0.2541,
+      "step": 48878
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.584706060148104e-06,
+      "loss": 0.2485,
+      "step": 48879
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5843702248417376e-06,
+      "loss": 0.395,
+      "step": 48880
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5840343895353723e-06,
+      "loss": 0.1943,
+      "step": 48881
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5836985542290065e-06,
+      "loss": 0.1494,
+      "step": 48882
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5833627189226407e-06,
+      "loss": 0.2747,
+      "step": 48883
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5830268836162745e-06,
+      "loss": 0.212,
+      "step": 48884
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.582691048309909e-06,
+      "loss": 0.1696,
+      "step": 48885
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5823552130035434e-06,
+      "loss": 0.2343,
+      "step": 48886
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.582019377697177e-06,
+      "loss": 0.2576,
+      "step": 48887
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.581683542390812e-06,
+      "loss": 0.3767,
+      "step": 48888
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.581347707084446e-06,
+      "loss": 0.1012,
+      "step": 48889
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.58101187177808e-06,
+      "loss": 0.248,
+      "step": 48890
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5806760364717145e-06,
+      "loss": 0.2883,
+      "step": 48891
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5803402011653488e-06,
+      "loss": 0.3516,
+      "step": 48892
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.580004365858983e-06,
+      "loss": 0.2772,
+      "step": 48893
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5796685305526176e-06,
+      "loss": 0.2309,
+      "step": 48894
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5793326952462514e-06,
+      "loss": 0.2557,
+      "step": 48895
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5789968599398856e-06,
+      "loss": 0.1953,
+      "step": 48896
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5786610246335203e-06,
+      "loss": 0.3675,
+      "step": 48897
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.578325189327154e-06,
+      "loss": 0.3104,
+      "step": 48898
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5779893540207883e-06,
+      "loss": 0.1882,
+      "step": 48899
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.577653518714423e-06,
+      "loss": 0.4697,
+      "step": 48900
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5773176834080568e-06,
+      "loss": 0.281,
+      "step": 48901
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.576981848101691e-06,
+      "loss": 0.318,
+      "step": 48902
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5766460127953256e-06,
+      "loss": 0.2796,
+      "step": 48903
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.57631017748896e-06,
+      "loss": 0.398,
+      "step": 48904
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5759743421825937e-06,
+      "loss": 0.2381,
+      "step": 48905
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5756385068762283e-06,
+      "loss": 0.1977,
+      "step": 48906
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5753026715698625e-06,
+      "loss": 0.2656,
+      "step": 48907
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5749668362634963e-06,
+      "loss": 0.3981,
+      "step": 48908
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.574631000957131e-06,
+      "loss": 0.3456,
+      "step": 48909
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.574295165650765e-06,
+      "loss": 0.1915,
+      "step": 48910
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5739593303443994e-06,
+      "loss": 0.3219,
+      "step": 48911
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5736234950380337e-06,
+      "loss": 0.3746,
+      "step": 48912
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.573287659731668e-06,
+      "loss": 0.2512,
+      "step": 48913
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.572951824425302e-06,
+      "loss": 0.3467,
+      "step": 48914
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5726159891189367e-06,
+      "loss": 0.1571,
+      "step": 48915
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5722801538125705e-06,
+      "loss": 0.2608,
+      "step": 48916
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5719443185062048e-06,
+      "loss": 0.252,
+      "step": 48917
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5716084831998394e-06,
+      "loss": 0.2077,
+      "step": 48918
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5712726478934732e-06,
+      "loss": 0.2162,
+      "step": 48919
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5709368125871074e-06,
+      "loss": 0.25,
+      "step": 48920
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.570600977280742e-06,
+      "loss": 0.2168,
+      "step": 48921
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5702651419743763e-06,
+      "loss": 0.3505,
+      "step": 48922
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.56992930666801e-06,
+      "loss": 0.2478,
+      "step": 48923
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5695934713616448e-06,
+      "loss": 0.1466,
+      "step": 48924
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.569257636055279e-06,
+      "loss": 0.2403,
+      "step": 48925
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5689218007489128e-06,
+      "loss": 0.3292,
+      "step": 48926
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5685859654425474e-06,
+      "loss": 0.2434,
+      "step": 48927
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5682501301361817e-06,
+      "loss": 0.3788,
+      "step": 48928
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5679142948298155e-06,
+      "loss": 0.3291,
+      "step": 48929
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5675784595234497e-06,
+      "loss": 0.2016,
+      "step": 48930
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5672426242170843e-06,
+      "loss": 0.3196,
+      "step": 48931
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 3.5669067889107186e-06,
+      "loss": 0.2843,
+      "step": 48932
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5665709536043524e-06,
+      "loss": 0.2132,
+      "step": 48933
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.566235118297987e-06,
+      "loss": 0.2452,
+      "step": 48934
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5658992829916212e-06,
+      "loss": 0.2324,
+      "step": 48935
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.565563447685255e-06,
+      "loss": 0.2497,
+      "step": 48936
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5652276123788897e-06,
+      "loss": 0.1443,
+      "step": 48937
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.564891777072524e-06,
+      "loss": 0.2273,
+      "step": 48938
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.564555941766158e-06,
+      "loss": 0.3592,
+      "step": 48939
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5642201064597923e-06,
+      "loss": 0.1021,
+      "step": 48940
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5638842711534266e-06,
+      "loss": 0.2867,
+      "step": 48941
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.563548435847061e-06,
+      "loss": 0.1779,
+      "step": 48942
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5632126005406954e-06,
+      "loss": 0.1035,
+      "step": 48943
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5628767652343292e-06,
+      "loss": 0.1844,
+      "step": 48944
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5625409299279635e-06,
+      "loss": 0.2873,
+      "step": 48945
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.562205094621598e-06,
+      "loss": 0.4003,
+      "step": 48946
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.561869259315232e-06,
+      "loss": 0.3914,
+      "step": 48947
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.561533424008866e-06,
+      "loss": 0.3409,
+      "step": 48948
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5611975887025008e-06,
+      "loss": 0.2265,
+      "step": 48949
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.560861753396135e-06,
+      "loss": 0.2452,
+      "step": 48950
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.560525918089769e-06,
+      "loss": 0.2676,
+      "step": 48951
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5601900827834035e-06,
+      "loss": 0.265,
+      "step": 48952
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5598542474770377e-06,
+      "loss": 0.2252,
+      "step": 48953
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5595184121706715e-06,
+      "loss": 0.229,
+      "step": 48954
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.559182576864306e-06,
+      "loss": 0.295,
+      "step": 48955
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5588467415579403e-06,
+      "loss": 0.231,
+      "step": 48956
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.558510906251574e-06,
+      "loss": 0.2024,
+      "step": 48957
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.558175070945209e-06,
+      "loss": 0.2335,
+      "step": 48958
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.557839235638843e-06,
+      "loss": 0.1924,
+      "step": 48959
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5575034003324772e-06,
+      "loss": 0.5408,
+      "step": 48960
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.557167565026112e-06,
+      "loss": 0.1082,
+      "step": 48961
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5568317297197457e-06,
+      "loss": 0.3384,
+      "step": 48962
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.55649589441338e-06,
+      "loss": 0.1633,
+      "step": 48963
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5561600591070146e-06,
+      "loss": 0.1568,
+      "step": 48964
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5558242238006484e-06,
+      "loss": 0.3439,
+      "step": 48965
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5554883884942826e-06,
+      "loss": 0.3873,
+      "step": 48966
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5551525531879172e-06,
+      "loss": 0.1747,
+      "step": 48967
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5548167178815515e-06,
+      "loss": 0.2405,
+      "step": 48968
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5544808825751853e-06,
+      "loss": 0.3271,
+      "step": 48969
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.55414504726882e-06,
+      "loss": 0.2645,
+      "step": 48970
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.553809211962454e-06,
+      "loss": 0.2379,
+      "step": 48971
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.553473376656088e-06,
+      "loss": 0.2427,
+      "step": 48972
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5531375413497226e-06,
+      "loss": 0.376,
+      "step": 48973
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.552801706043357e-06,
+      "loss": 0.3789,
+      "step": 48974
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5524658707369906e-06,
+      "loss": 0.3033,
+      "step": 48975
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.552130035430625e-06,
+      "loss": 0.2516,
+      "step": 48976
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5517942001242595e-06,
+      "loss": 0.3729,
+      "step": 48977
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5514583648178937e-06,
+      "loss": 0.1411,
+      "step": 48978
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5511225295115275e-06,
+      "loss": 0.3503,
+      "step": 48979
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.550786694205162e-06,
+      "loss": 0.2823,
+      "step": 48980
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5504508588987964e-06,
+      "loss": 0.2601,
+      "step": 48981
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.55011502359243e-06,
+      "loss": 0.4261,
+      "step": 48982
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.549779188286065e-06,
+      "loss": 0.1299,
+      "step": 48983
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.549443352979699e-06,
+      "loss": 0.2901,
+      "step": 48984
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5491075176733333e-06,
+      "loss": 0.2471,
+      "step": 48985
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5487716823669675e-06,
+      "loss": 0.3133,
+      "step": 48986
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5484358470606017e-06,
+      "loss": 0.338,
+      "step": 48987
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.548100011754236e-06,
+      "loss": 0.3006,
+      "step": 48988
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5477641764478706e-06,
+      "loss": 0.2996,
+      "step": 48989
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5474283411415044e-06,
+      "loss": 0.3076,
+      "step": 48990
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5470925058351386e-06,
+      "loss": 0.2863,
+      "step": 48991
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5467566705287733e-06,
+      "loss": 0.241,
+      "step": 48992
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.546420835222407e-06,
+      "loss": 0.3053,
+      "step": 48993
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5460849999160413e-06,
+      "loss": 0.2808,
+      "step": 48994
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.545749164609676e-06,
+      "loss": 0.3763,
+      "step": 48995
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.54541332930331e-06,
+      "loss": 0.2532,
+      "step": 48996
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.545077493996944e-06,
+      "loss": 0.2316,
+      "step": 48997
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5447416586905786e-06,
+      "loss": 0.2668,
+      "step": 48998
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.544405823384213e-06,
+      "loss": 0.2127,
+      "step": 48999
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5440699880778466e-06,
+      "loss": 0.1892,
+      "step": 49000
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5437341527714813e-06,
+      "loss": 0.1847,
+      "step": 49001
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5433983174651155e-06,
+      "loss": 0.1524,
+      "step": 49002
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5430624821587493e-06,
+      "loss": 0.2131,
+      "step": 49003
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.542726646852384e-06,
+      "loss": 0.244,
+      "step": 49004
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.542390811546018e-06,
+      "loss": 0.2229,
+      "step": 49005
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5420549762396524e-06,
+      "loss": 0.2915,
+      "step": 49006
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.541719140933287e-06,
+      "loss": 0.2426,
+      "step": 49007
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.541383305626921e-06,
+      "loss": 0.2675,
+      "step": 49008
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.541047470320555e-06,
+      "loss": 0.2099,
+      "step": 49009
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5407116350141897e-06,
+      "loss": 0.2497,
+      "step": 49010
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5403757997078235e-06,
+      "loss": 0.2996,
+      "step": 49011
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5400399644014577e-06,
+      "loss": 0.3038,
+      "step": 49012
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5397041290950924e-06,
+      "loss": 0.2258,
+      "step": 49013
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.539368293788726e-06,
+      "loss": 0.4054,
+      "step": 49014
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5390324584823604e-06,
+      "loss": 0.3942,
+      "step": 49015
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.538696623175995e-06,
+      "loss": 0.3747,
+      "step": 49016
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5383607878696293e-06,
+      "loss": 0.3111,
+      "step": 49017
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.538024952563263e-06,
+      "loss": 0.1963,
+      "step": 49018
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5376891172568973e-06,
+      "loss": 0.2172,
+      "step": 49019
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.537353281950532e-06,
+      "loss": 0.3318,
+      "step": 49020
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5370174466441657e-06,
+      "loss": 0.2163,
+      "step": 49021
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5366816113378e-06,
+      "loss": 0.2753,
+      "step": 49022
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5363457760314346e-06,
+      "loss": 0.1627,
+      "step": 49023
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.536009940725069e-06,
+      "loss": 0.1525,
+      "step": 49024
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5356741054187026e-06,
+      "loss": 0.3257,
+      "step": 49025
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5353382701123373e-06,
+      "loss": 0.2587,
+      "step": 49026
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5350024348059715e-06,
+      "loss": 0.1989,
+      "step": 49027
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5346665994996053e-06,
+      "loss": 0.3114,
+      "step": 49028
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.53433076419324e-06,
+      "loss": 0.1866,
+      "step": 49029
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.533994928886874e-06,
+      "loss": 0.297,
+      "step": 49030
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.533659093580508e-06,
+      "loss": 0.2694,
+      "step": 49031
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5333232582741426e-06,
+      "loss": 0.2226,
+      "step": 49032
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.532987422967777e-06,
+      "loss": 0.294,
+      "step": 49033
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.532651587661411e-06,
+      "loss": 0.3754,
+      "step": 49034
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5323157523550457e-06,
+      "loss": 0.2739,
+      "step": 49035
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5319799170486795e-06,
+      "loss": 0.2349,
+      "step": 49036
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5316440817423138e-06,
+      "loss": 0.2703,
+      "step": 49037
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5313082464359484e-06,
+      "loss": 0.2087,
+      "step": 49038
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.530972411129582e-06,
+      "loss": 0.1491,
+      "step": 49039
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5306365758232164e-06,
+      "loss": 0.3461,
+      "step": 49040
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.530300740516851e-06,
+      "loss": 0.2617,
+      "step": 49041
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.529964905210485e-06,
+      "loss": 0.3037,
+      "step": 49042
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.529629069904119e-06,
+      "loss": 0.2242,
+      "step": 49043
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5292932345977537e-06,
+      "loss": 0.3606,
+      "step": 49044
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.528957399291388e-06,
+      "loss": 0.3776,
+      "step": 49045
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5286215639850218e-06,
+      "loss": 0.3646,
+      "step": 49046
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5282857286786564e-06,
+      "loss": 0.337,
+      "step": 49047
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5279498933722906e-06,
+      "loss": 0.3469,
+      "step": 49048
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5276140580659244e-06,
+      "loss": 0.1687,
+      "step": 49049
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.527278222759559e-06,
+      "loss": 0.3607,
+      "step": 49050
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5269423874531933e-06,
+      "loss": 0.2612,
+      "step": 49051
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5266065521468275e-06,
+      "loss": 0.3052,
+      "step": 49052
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5262707168404618e-06,
+      "loss": 0.1278,
+      "step": 49053
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.525934881534096e-06,
+      "loss": 0.3074,
+      "step": 49054
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.52559904622773e-06,
+      "loss": 0.2335,
+      "step": 49055
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.525263210921365e-06,
+      "loss": 0.1602,
+      "step": 49056
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5249273756149987e-06,
+      "loss": 0.1745,
+      "step": 49057
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.524591540308633e-06,
+      "loss": 0.2684,
+      "step": 49058
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5242557050022675e-06,
+      "loss": 0.2237,
+      "step": 49059
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5239198696959013e-06,
+      "loss": 0.2298,
+      "step": 49060
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5235840343895355e-06,
+      "loss": 0.3761,
+      "step": 49061
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.52324819908317e-06,
+      "loss": 0.2647,
+      "step": 49062
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5229123637768044e-06,
+      "loss": 0.133,
+      "step": 49063
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5225765284704382e-06,
+      "loss": 0.3678,
+      "step": 49064
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5222406931640724e-06,
+      "loss": 0.2419,
+      "step": 49065
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.521904857857707e-06,
+      "loss": 0.2247,
+      "step": 49066
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.521569022551341e-06,
+      "loss": 0.262,
+      "step": 49067
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.521233187244975e-06,
+      "loss": 0.385,
+      "step": 49068
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5208973519386098e-06,
+      "loss": 0.2497,
+      "step": 49069
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5205615166322436e-06,
+      "loss": 0.2985,
+      "step": 49070
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5202256813258778e-06,
+      "loss": 0.494,
+      "step": 49071
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5198898460195124e-06,
+      "loss": 0.2784,
+      "step": 49072
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5195540107131467e-06,
+      "loss": 0.2126,
+      "step": 49073
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5192181754067805e-06,
+      "loss": 0.3461,
+      "step": 49074
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.518882340100415e-06,
+      "loss": 0.4837,
+      "step": 49075
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5185465047940493e-06,
+      "loss": 0.3528,
+      "step": 49076
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.518210669487683e-06,
+      "loss": 0.2763,
+      "step": 49077
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5178748341813178e-06,
+      "loss": 0.2342,
+      "step": 49078
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.517538998874952e-06,
+      "loss": 0.2541,
+      "step": 49079
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5172031635685862e-06,
+      "loss": 0.3289,
+      "step": 49080
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5168673282622205e-06,
+      "loss": 0.2524,
+      "step": 49081
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5165314929558547e-06,
+      "loss": 0.4772,
+      "step": 49082
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.516195657649489e-06,
+      "loss": 0.1853,
+      "step": 49083
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5158598223431235e-06,
+      "loss": 0.2787,
+      "step": 49084
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5155239870367573e-06,
+      "loss": 0.2716,
+      "step": 49085
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5151881517303916e-06,
+      "loss": 0.2708,
+      "step": 49086
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5148523164240262e-06,
+      "loss": 0.2731,
+      "step": 49087
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.51451648111766e-06,
+      "loss": 0.1662,
+      "step": 49088
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5141806458112942e-06,
+      "loss": 0.2855,
+      "step": 49089
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.513844810504929e-06,
+      "loss": 0.1995,
+      "step": 49090
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.513508975198563e-06,
+      "loss": 0.2377,
+      "step": 49091
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.513173139892197e-06,
+      "loss": 0.2864,
+      "step": 49092
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5128373045858316e-06,
+      "loss": 0.4256,
+      "step": 49093
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5125014692794658e-06,
+      "loss": 0.325,
+      "step": 49094
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5121656339730996e-06,
+      "loss": 0.5124,
+      "step": 49095
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5118297986667342e-06,
+      "loss": 0.2519,
+      "step": 49096
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5114939633603685e-06,
+      "loss": 0.2571,
+      "step": 49097
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5111581280540027e-06,
+      "loss": 0.1219,
+      "step": 49098
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.510822292747637e-06,
+      "loss": 0.3553,
+      "step": 49099
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.510486457441271e-06,
+      "loss": 0.2554,
+      "step": 49100
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5101506221349054e-06,
+      "loss": 0.2987,
+      "step": 49101
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.50981478682854e-06,
+      "loss": 0.3037,
+      "step": 49102
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.509478951522174e-06,
+      "loss": 0.2702,
+      "step": 49103
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.509143116215808e-06,
+      "loss": 0.1178,
+      "step": 49104
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5088072809094427e-06,
+      "loss": 0.3923,
+      "step": 49105
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5084714456030765e-06,
+      "loss": 0.2193,
+      "step": 49106
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5081356102967107e-06,
+      "loss": 0.1317,
+      "step": 49107
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5077997749903453e-06,
+      "loss": 0.299,
+      "step": 49108
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5074639396839796e-06,
+      "loss": 0.1784,
+      "step": 49109
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5071281043776134e-06,
+      "loss": 0.1693,
+      "step": 49110
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5067922690712476e-06,
+      "loss": 0.3715,
+      "step": 49111
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5064564337648822e-06,
+      "loss": 0.257,
+      "step": 49112
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.506120598458516e-06,
+      "loss": 0.2665,
+      "step": 49113
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5057847631521503e-06,
+      "loss": 0.2937,
+      "step": 49114
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.505448927845785e-06,
+      "loss": 0.3938,
+      "step": 49115
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5051130925394187e-06,
+      "loss": 0.392,
+      "step": 49116
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.504777257233053e-06,
+      "loss": 0.3997,
+      "step": 49117
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5044414219266876e-06,
+      "loss": 0.3506,
+      "step": 49118
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.504105586620322e-06,
+      "loss": 0.2037,
+      "step": 49119
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5037697513139556e-06,
+      "loss": 0.1578,
+      "step": 49120
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5034339160075903e-06,
+      "loss": 0.2784,
+      "step": 49121
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5030980807012245e-06,
+      "loss": 0.4069,
+      "step": 49122
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5027622453948583e-06,
+      "loss": 0.3519,
+      "step": 49123
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.502426410088493e-06,
+      "loss": 0.3334,
+      "step": 49124
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.502090574782127e-06,
+      "loss": 0.2909,
+      "step": 49125
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5017547394757614e-06,
+      "loss": 0.4564,
+      "step": 49126
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5014189041693956e-06,
+      "loss": 0.2374,
+      "step": 49127
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.50108306886303e-06,
+      "loss": 0.3627,
+      "step": 49128
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.500747233556664e-06,
+      "loss": 0.2079,
+      "step": 49129
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5004113982502987e-06,
+      "loss": 0.4475,
+      "step": 49130
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 3.5000755629439325e-06,
+      "loss": 0.171,
+      "step": 49131
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4997397276375667e-06,
+      "loss": 0.2949,
+      "step": 49132
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4994038923312014e-06,
+      "loss": 0.3192,
+      "step": 49133
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.499068057024835e-06,
+      "loss": 0.2269,
+      "step": 49134
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4987322217184694e-06,
+      "loss": 0.242,
+      "step": 49135
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.498396386412104e-06,
+      "loss": 0.2065,
+      "step": 49136
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4980605511057383e-06,
+      "loss": 0.3024,
+      "step": 49137
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.497724715799372e-06,
+      "loss": 0.2216,
+      "step": 49138
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4973888804930067e-06,
+      "loss": 0.2533,
+      "step": 49139
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.497053045186641e-06,
+      "loss": 0.1835,
+      "step": 49140
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4967172098802747e-06,
+      "loss": 0.2325,
+      "step": 49141
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4963813745739094e-06,
+      "loss": 0.5421,
+      "step": 49142
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4960455392675436e-06,
+      "loss": 0.3215,
+      "step": 49143
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4957097039611774e-06,
+      "loss": 0.3,
+      "step": 49144
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.495373868654812e-06,
+      "loss": 0.385,
+      "step": 49145
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4950380333484463e-06,
+      "loss": 0.2723,
+      "step": 49146
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4947021980420805e-06,
+      "loss": 0.2384,
+      "step": 49147
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.494366362735715e-06,
+      "loss": 0.193,
+      "step": 49148
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.494030527429349e-06,
+      "loss": 0.2223,
+      "step": 49149
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.493694692122983e-06,
+      "loss": 0.295,
+      "step": 49150
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.493358856816618e-06,
+      "loss": 0.2613,
+      "step": 49151
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4930230215102516e-06,
+      "loss": 0.3038,
+      "step": 49152
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.492687186203886e-06,
+      "loss": 0.2313,
+      "step": 49153
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.49235135089752e-06,
+      "loss": 0.2111,
+      "step": 49154
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4920155155911543e-06,
+      "loss": 0.4004,
+      "step": 49155
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4916796802847885e-06,
+      "loss": 0.3783,
+      "step": 49156
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4913438449784227e-06,
+      "loss": 0.1682,
+      "step": 49157
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4910080096720574e-06,
+      "loss": 0.3751,
+      "step": 49158
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.490672174365691e-06,
+      "loss": 0.2754,
+      "step": 49159
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4903363390593254e-06,
+      "loss": 0.4325,
+      "step": 49160
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.49000050375296e-06,
+      "loss": 0.2642,
+      "step": 49161
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.489664668446594e-06,
+      "loss": 0.3614,
+      "step": 49162
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.489328833140228e-06,
+      "loss": 0.3461,
+      "step": 49163
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4889929978338627e-06,
+      "loss": 0.3504,
+      "step": 49164
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.488657162527497e-06,
+      "loss": 0.2096,
+      "step": 49165
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4883213272211307e-06,
+      "loss": 0.2182,
+      "step": 49166
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4879854919147654e-06,
+      "loss": 0.2411,
+      "step": 49167
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4876496566083996e-06,
+      "loss": 0.321,
+      "step": 49168
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4873138213020334e-06,
+      "loss": 0.2827,
+      "step": 49169
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.486977985995668e-06,
+      "loss": 0.1621,
+      "step": 49170
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4866421506893023e-06,
+      "loss": 0.1691,
+      "step": 49171
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.486306315382936e-06,
+      "loss": 0.2106,
+      "step": 49172
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4859704800765707e-06,
+      "loss": 0.3025,
+      "step": 49173
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.485634644770205e-06,
+      "loss": 0.236,
+      "step": 49174
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.485298809463839e-06,
+      "loss": 0.3552,
+      "step": 49175
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.484962974157474e-06,
+      "loss": 0.2435,
+      "step": 49176
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4846271388511076e-06,
+      "loss": 0.3048,
+      "step": 49177
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.484291303544742e-06,
+      "loss": 0.2828,
+      "step": 49178
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4839554682383765e-06,
+      "loss": 0.2482,
+      "step": 49179
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4836196329320103e-06,
+      "loss": 0.2891,
+      "step": 49180
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4832837976256445e-06,
+      "loss": 0.2167,
+      "step": 49181
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.482947962319279e-06,
+      "loss": 0.3286,
+      "step": 49182
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.482612127012913e-06,
+      "loss": 0.2292,
+      "step": 49183
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.482276291706547e-06,
+      "loss": 0.1305,
+      "step": 49184
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.481940456400182e-06,
+      "loss": 0.1792,
+      "step": 49185
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.481604621093816e-06,
+      "loss": 0.2442,
+      "step": 49186
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.48126878578745e-06,
+      "loss": 0.3068,
+      "step": 49187
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4809329504810845e-06,
+      "loss": 0.2324,
+      "step": 49188
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4805971151747187e-06,
+      "loss": 0.2967,
+      "step": 49189
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4802612798683525e-06,
+      "loss": 0.2254,
+      "step": 49190
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.479925444561987e-06,
+      "loss": 0.28,
+      "step": 49191
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4795896092556214e-06,
+      "loss": 0.2388,
+      "step": 49192
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4792537739492556e-06,
+      "loss": 0.2385,
+      "step": 49193
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.47891793864289e-06,
+      "loss": 0.1597,
+      "step": 49194
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.478582103336524e-06,
+      "loss": 0.2146,
+      "step": 49195
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4782462680301583e-06,
+      "loss": 0.4439,
+      "step": 49196
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.477910432723793e-06,
+      "loss": 0.3189,
+      "step": 49197
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4775745974174268e-06,
+      "loss": 0.2056,
+      "step": 49198
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.477238762111061e-06,
+      "loss": 0.2814,
+      "step": 49199
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4769029268046948e-06,
+      "loss": 0.273,
+      "step": 49200
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4765670914983294e-06,
+      "loss": 0.2092,
+      "step": 49201
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4762312561919637e-06,
+      "loss": 0.1296,
+      "step": 49202
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.475895420885598e-06,
+      "loss": 0.3245,
+      "step": 49203
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4755595855792325e-06,
+      "loss": 0.3203,
+      "step": 49204
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4752237502728663e-06,
+      "loss": 0.1659,
+      "step": 49205
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4748879149665006e-06,
+      "loss": 0.2219,
+      "step": 49206
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.474552079660135e-06,
+      "loss": 0.198,
+      "step": 49207
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.474216244353769e-06,
+      "loss": 0.2772,
+      "step": 49208
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4738804090474032e-06,
+      "loss": 0.2013,
+      "step": 49209
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.473544573741038e-06,
+      "loss": 0.1592,
+      "step": 49210
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4732087384346717e-06,
+      "loss": 0.2331,
+      "step": 49211
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.472872903128306e-06,
+      "loss": 0.3877,
+      "step": 49212
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4725370678219405e-06,
+      "loss": 0.2382,
+      "step": 49213
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4722012325155748e-06,
+      "loss": 0.2534,
+      "step": 49214
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4718653972092086e-06,
+      "loss": 0.216,
+      "step": 49215
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4715295619028432e-06,
+      "loss": 0.2375,
+      "step": 49216
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4711937265964774e-06,
+      "loss": 0.1563,
+      "step": 49217
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4708578912901112e-06,
+      "loss": 0.2057,
+      "step": 49218
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.470522055983746e-06,
+      "loss": 0.2279,
+      "step": 49219
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.47018622067738e-06,
+      "loss": 0.3736,
+      "step": 49220
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4698503853710143e-06,
+      "loss": 0.2739,
+      "step": 49221
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4695145500646486e-06,
+      "loss": 0.2121,
+      "step": 49222
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4691787147582828e-06,
+      "loss": 0.219,
+      "step": 49223
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.468842879451917e-06,
+      "loss": 0.1596,
+      "step": 49224
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4685070441455517e-06,
+      "loss": 0.4227,
+      "step": 49225
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4681712088391855e-06,
+      "loss": 0.1816,
+      "step": 49226
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4678353735328197e-06,
+      "loss": 0.2323,
+      "step": 49227
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4674995382264543e-06,
+      "loss": 0.2134,
+      "step": 49228
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.467163702920088e-06,
+      "loss": 0.3543,
+      "step": 49229
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4668278676137223e-06,
+      "loss": 0.2509,
+      "step": 49230
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.466492032307357e-06,
+      "loss": 0.1734,
+      "step": 49231
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4661561970009912e-06,
+      "loss": 0.2638,
+      "step": 49232
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.465820361694625e-06,
+      "loss": 0.2468,
+      "step": 49233
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4654845263882597e-06,
+      "loss": 0.2744,
+      "step": 49234
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.465148691081894e-06,
+      "loss": 0.2857,
+      "step": 49235
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4648128557755277e-06,
+      "loss": 0.3111,
+      "step": 49236
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4644770204691623e-06,
+      "loss": 0.2581,
+      "step": 49237
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4641411851627966e-06,
+      "loss": 0.3789,
+      "step": 49238
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4638053498564308e-06,
+      "loss": 0.2329,
+      "step": 49239
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.463469514550065e-06,
+      "loss": 0.1613,
+      "step": 49240
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4631336792436992e-06,
+      "loss": 0.3184,
+      "step": 49241
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4627978439373335e-06,
+      "loss": 0.3423,
+      "step": 49242
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.462462008630968e-06,
+      "loss": 0.2341,
+      "step": 49243
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.462126173324602e-06,
+      "loss": 0.3328,
+      "step": 49244
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.461790338018236e-06,
+      "loss": 0.1321,
+      "step": 49245
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.46145450271187e-06,
+      "loss": 0.2244,
+      "step": 49246
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4611186674055046e-06,
+      "loss": 0.1928,
+      "step": 49247
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.460782832099139e-06,
+      "loss": 0.2071,
+      "step": 49248
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.460446996792773e-06,
+      "loss": 0.1373,
+      "step": 49249
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4601111614864077e-06,
+      "loss": 0.2805,
+      "step": 49250
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4597753261800415e-06,
+      "loss": 0.3819,
+      "step": 49251
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4594394908736757e-06,
+      "loss": 0.3585,
+      "step": 49252
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4591036555673103e-06,
+      "loss": 0.2832,
+      "step": 49253
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.458767820260944e-06,
+      "loss": 0.1127,
+      "step": 49254
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4584319849545784e-06,
+      "loss": 0.4777,
+      "step": 49255
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.458096149648213e-06,
+      "loss": 0.1915,
+      "step": 49256
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.457760314341847e-06,
+      "loss": 0.2718,
+      "step": 49257
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.457424479035481e-06,
+      "loss": 0.175,
+      "step": 49258
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4570886437291157e-06,
+      "loss": 0.3091,
+      "step": 49259
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.45675280842275e-06,
+      "loss": 0.4249,
+      "step": 49260
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4564169731163837e-06,
+      "loss": 0.2966,
+      "step": 49261
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4560811378100184e-06,
+      "loss": 0.118,
+      "step": 49262
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4557453025036526e-06,
+      "loss": 0.2135,
+      "step": 49263
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4554094671972864e-06,
+      "loss": 0.1784,
+      "step": 49264
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.455073631890921e-06,
+      "loss": 0.2908,
+      "step": 49265
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4547377965845553e-06,
+      "loss": 0.3276,
+      "step": 49266
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4544019612781895e-06,
+      "loss": 0.2475,
+      "step": 49267
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4540661259718237e-06,
+      "loss": 0.3077,
+      "step": 49268
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.453730290665458e-06,
+      "loss": 0.2753,
+      "step": 49269
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.453394455359092e-06,
+      "loss": 0.4336,
+      "step": 49270
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.453058620052727e-06,
+      "loss": 0.4163,
+      "step": 49271
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4527227847463606e-06,
+      "loss": 0.1206,
+      "step": 49272
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.452386949439995e-06,
+      "loss": 0.2852,
+      "step": 49273
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4520511141336295e-06,
+      "loss": 0.2711,
+      "step": 49274
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4517152788272633e-06,
+      "loss": 0.1544,
+      "step": 49275
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4513794435208975e-06,
+      "loss": 0.2793,
+      "step": 49276
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.451043608214532e-06,
+      "loss": 0.5366,
+      "step": 49277
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4507077729081664e-06,
+      "loss": 0.3769,
+      "step": 49278
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4503719376018e-06,
+      "loss": 0.2971,
+      "step": 49279
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.450036102295435e-06,
+      "loss": 0.2168,
+      "step": 49280
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.449700266989069e-06,
+      "loss": 0.2533,
+      "step": 49281
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.449364431682703e-06,
+      "loss": 0.2925,
+      "step": 49282
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4490285963763375e-06,
+      "loss": 0.2599,
+      "step": 49283
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4486927610699717e-06,
+      "loss": 0.2079,
+      "step": 49284
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4483569257636055e-06,
+      "loss": 0.3232,
+      "step": 49285
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.44802109045724e-06,
+      "loss": 0.3093,
+      "step": 49286
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4476852551508744e-06,
+      "loss": 0.2646,
+      "step": 49287
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4473494198445086e-06,
+      "loss": 0.2205,
+      "step": 49288
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4470135845381433e-06,
+      "loss": 0.2422,
+      "step": 49289
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.446677749231777e-06,
+      "loss": 0.2221,
+      "step": 49290
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4463419139254113e-06,
+      "loss": 0.2169,
+      "step": 49291
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.446006078619045e-06,
+      "loss": 0.2616,
+      "step": 49292
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4456702433126797e-06,
+      "loss": 0.3152,
+      "step": 49293
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.445334408006314e-06,
+      "loss": 0.2649,
+      "step": 49294
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.444998572699948e-06,
+      "loss": 0.2181,
+      "step": 49295
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4446627373935824e-06,
+      "loss": 0.16,
+      "step": 49296
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4443269020872166e-06,
+      "loss": 0.4618,
+      "step": 49297
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.443991066780851e-06,
+      "loss": 0.2922,
+      "step": 49298
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4436552314744855e-06,
+      "loss": 0.2153,
+      "step": 49299
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4433193961681193e-06,
+      "loss": 0.2912,
+      "step": 49300
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4429835608617535e-06,
+      "loss": 0.357,
+      "step": 49301
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.442647725555388e-06,
+      "loss": 0.2505,
+      "step": 49302
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.442311890249022e-06,
+      "loss": 0.2426,
+      "step": 49303
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.441976054942656e-06,
+      "loss": 0.1347,
+      "step": 49304
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.441640219636291e-06,
+      "loss": 0.3047,
+      "step": 49305
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.441304384329925e-06,
+      "loss": 0.3816,
+      "step": 49306
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.440968549023559e-06,
+      "loss": 0.2969,
+      "step": 49307
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4406327137171935e-06,
+      "loss": 0.2604,
+      "step": 49308
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4402968784108277e-06,
+      "loss": 0.1219,
+      "step": 49309
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4399610431044615e-06,
+      "loss": 0.265,
+      "step": 49310
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.439625207798096e-06,
+      "loss": 0.2369,
+      "step": 49311
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4392893724917304e-06,
+      "loss": 0.3542,
+      "step": 49312
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.438953537185364e-06,
+      "loss": 0.3097,
+      "step": 49313
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.438617701878999e-06,
+      "loss": 0.4844,
+      "step": 49314
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.438281866572633e-06,
+      "loss": 0.1788,
+      "step": 49315
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4379460312662673e-06,
+      "loss": 0.2322,
+      "step": 49316
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.437610195959902e-06,
+      "loss": 0.2951,
+      "step": 49317
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4372743606535357e-06,
+      "loss": 0.2221,
+      "step": 49318
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.43693852534717e-06,
+      "loss": 0.327,
+      "step": 49319
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4366026900408046e-06,
+      "loss": 0.3544,
+      "step": 49320
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4362668547344384e-06,
+      "loss": 0.3284,
+      "step": 49321
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4359310194280726e-06,
+      "loss": 0.2594,
+      "step": 49322
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4355951841217073e-06,
+      "loss": 0.2967,
+      "step": 49323
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.435259348815341e-06,
+      "loss": 0.2458,
+      "step": 49324
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4349235135089753e-06,
+      "loss": 0.1271,
+      "step": 49325
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.43458767820261e-06,
+      "loss": 0.3942,
+      "step": 49326
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.434251842896244e-06,
+      "loss": 0.2804,
+      "step": 49327
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.433916007589878e-06,
+      "loss": 0.1567,
+      "step": 49328
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.4335801722835126e-06,
+      "loss": 0.2124,
+      "step": 49329
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.433244336977147e-06,
+      "loss": 0.3103,
+      "step": 49330
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4329085016707807e-06,
+      "loss": 0.2611,
+      "step": 49331
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4325726663644153e-06,
+      "loss": 0.3417,
+      "step": 49332
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4322368310580495e-06,
+      "loss": 0.3354,
+      "step": 49333
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4319009957516837e-06,
+      "loss": 0.3515,
+      "step": 49334
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4315651604453175e-06,
+      "loss": 0.4568,
+      "step": 49335
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.431229325138952e-06,
+      "loss": 0.3444,
+      "step": 49336
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4308934898325864e-06,
+      "loss": 0.3892,
+      "step": 49337
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4305576545262202e-06,
+      "loss": 0.3225,
+      "step": 49338
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.430221819219855e-06,
+      "loss": 0.2331,
+      "step": 49339
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.429885983913489e-06,
+      "loss": 0.2896,
+      "step": 49340
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.429550148607123e-06,
+      "loss": 0.3206,
+      "step": 49341
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4292143133007575e-06,
+      "loss": 0.2644,
+      "step": 49342
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4288784779943918e-06,
+      "loss": 0.2016,
+      "step": 49343
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.428542642688026e-06,
+      "loss": 0.3351,
+      "step": 49344
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4282068073816606e-06,
+      "loss": 0.1902,
+      "step": 49345
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4278709720752944e-06,
+      "loss": 0.2351,
+      "step": 49346
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4275351367689287e-06,
+      "loss": 0.2028,
+      "step": 49347
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4271993014625633e-06,
+      "loss": 0.2492,
+      "step": 49348
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.426863466156197e-06,
+      "loss": 0.2647,
+      "step": 49349
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4265276308498313e-06,
+      "loss": 0.2454,
+      "step": 49350
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.426191795543466e-06,
+      "loss": 0.3996,
+      "step": 49351
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4258559602370998e-06,
+      "loss": 0.3575,
+      "step": 49352
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.425520124930734e-06,
+      "loss": 0.3158,
+      "step": 49353
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4251842896243686e-06,
+      "loss": 0.1758,
+      "step": 49354
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.424848454318003e-06,
+      "loss": 0.172,
+      "step": 49355
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4245126190116367e-06,
+      "loss": 0.1738,
+      "step": 49356
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4241767837052713e-06,
+      "loss": 0.2167,
+      "step": 49357
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4238409483989055e-06,
+      "loss": 0.1901,
+      "step": 49358
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4235051130925393e-06,
+      "loss": 0.323,
+      "step": 49359
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.423169277786174e-06,
+      "loss": 0.2994,
+      "step": 49360
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4228334424798082e-06,
+      "loss": 0.3799,
+      "step": 49361
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4224976071734424e-06,
+      "loss": 0.4153,
+      "step": 49362
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.422161771867077e-06,
+      "loss": 0.1878,
+      "step": 49363
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.421825936560711e-06,
+      "loss": 0.277,
+      "step": 49364
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.421490101254345e-06,
+      "loss": 0.2318,
+      "step": 49365
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4211542659479798e-06,
+      "loss": 0.3094,
+      "step": 49366
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4208184306416136e-06,
+      "loss": 0.2545,
+      "step": 49367
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4204825953352478e-06,
+      "loss": 0.182,
+      "step": 49368
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4201467600288824e-06,
+      "loss": 0.1155,
+      "step": 49369
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4198109247225162e-06,
+      "loss": 0.2137,
+      "step": 49370
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4194750894161505e-06,
+      "loss": 0.2636,
+      "step": 49371
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.419139254109785e-06,
+      "loss": 0.2055,
+      "step": 49372
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4188034188034193e-06,
+      "loss": 0.2095,
+      "step": 49373
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.418467583497053e-06,
+      "loss": 0.1997,
+      "step": 49374
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4181317481906878e-06,
+      "loss": 0.1641,
+      "step": 49375
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.417795912884322e-06,
+      "loss": 0.1904,
+      "step": 49376
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.417460077577956e-06,
+      "loss": 0.1849,
+      "step": 49377
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4171242422715904e-06,
+      "loss": 0.1816,
+      "step": 49378
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4167884069652247e-06,
+      "loss": 0.3268,
+      "step": 49379
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.416452571658859e-06,
+      "loss": 0.1653,
+      "step": 49380
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4161167363524927e-06,
+      "loss": 0.1896,
+      "step": 49381
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4157809010461273e-06,
+      "loss": 0.1963,
+      "step": 49382
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4154450657397616e-06,
+      "loss": 0.3307,
+      "step": 49383
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4151092304333954e-06,
+      "loss": 0.342,
+      "step": 49384
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.41477339512703e-06,
+      "loss": 0.2696,
+      "step": 49385
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4144375598206642e-06,
+      "loss": 0.3271,
+      "step": 49386
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.414101724514298e-06,
+      "loss": 0.2045,
+      "step": 49387
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4137658892079327e-06,
+      "loss": 0.3507,
+      "step": 49388
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.413430053901567e-06,
+      "loss": 0.2287,
+      "step": 49389
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.413094218595201e-06,
+      "loss": 0.3346,
+      "step": 49390
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4127583832888358e-06,
+      "loss": 0.2431,
+      "step": 49391
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4124225479824696e-06,
+      "loss": 0.2738,
+      "step": 49392
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.412086712676104e-06,
+      "loss": 0.2467,
+      "step": 49393
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4117508773697385e-06,
+      "loss": 0.2816,
+      "step": 49394
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4114150420633723e-06,
+      "loss": 0.2784,
+      "step": 49395
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4110792067570065e-06,
+      "loss": 0.1341,
+      "step": 49396
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.410743371450641e-06,
+      "loss": 0.2687,
+      "step": 49397
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.410407536144275e-06,
+      "loss": 0.2667,
+      "step": 49398
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.410071700837909e-06,
+      "loss": 0.23,
+      "step": 49399
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.409735865531544e-06,
+      "loss": 0.2638,
+      "step": 49400
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.409400030225178e-06,
+      "loss": 0.3425,
+      "step": 49401
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.409064194918812e-06,
+      "loss": 0.3868,
+      "step": 49402
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4087283596124465e-06,
+      "loss": 0.2717,
+      "step": 49403
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4083925243060807e-06,
+      "loss": 0.3647,
+      "step": 49404
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4080566889997145e-06,
+      "loss": 0.2439,
+      "step": 49405
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.407720853693349e-06,
+      "loss": 0.267,
+      "step": 49406
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4073850183869834e-06,
+      "loss": 0.3287,
+      "step": 49407
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4070491830806176e-06,
+      "loss": 0.2316,
+      "step": 49408
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.406713347774252e-06,
+      "loss": 0.1849,
+      "step": 49409
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.406377512467886e-06,
+      "loss": 0.33,
+      "step": 49410
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4060416771615203e-06,
+      "loss": 0.391,
+      "step": 49411
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.405705841855155e-06,
+      "loss": 0.2369,
+      "step": 49412
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4053700065487887e-06,
+      "loss": 0.3396,
+      "step": 49413
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.405034171242423e-06,
+      "loss": 0.2637,
+      "step": 49414
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4046983359360576e-06,
+      "loss": 0.3136,
+      "step": 49415
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4043625006296914e-06,
+      "loss": 0.3007,
+      "step": 49416
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4040266653233256e-06,
+      "loss": 0.3228,
+      "step": 49417
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4036908300169602e-06,
+      "loss": 0.5147,
+      "step": 49418
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4033549947105945e-06,
+      "loss": 0.2923,
+      "step": 49419
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4030191594042283e-06,
+      "loss": 0.2945,
+      "step": 49420
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.402683324097863e-06,
+      "loss": 0.3298,
+      "step": 49421
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.402347488791497e-06,
+      "loss": 0.1834,
+      "step": 49422
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.402011653485131e-06,
+      "loss": 0.2487,
+      "step": 49423
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4016758181787656e-06,
+      "loss": 0.2233,
+      "step": 49424
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4013399828724e-06,
+      "loss": 0.193,
+      "step": 49425
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4010041475660336e-06,
+      "loss": 0.4936,
+      "step": 49426
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.400668312259668e-06,
+      "loss": 0.4836,
+      "step": 49427
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.4003324769533025e-06,
+      "loss": 0.3968,
+      "step": 49428
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3999966416469367e-06,
+      "loss": 0.2355,
+      "step": 49429
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3996608063405705e-06,
+      "loss": 0.2352,
+      "step": 49430
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.399324971034205e-06,
+      "loss": 0.2693,
+      "step": 49431
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3989891357278394e-06,
+      "loss": 0.2843,
+      "step": 49432
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.398653300421473e-06,
+      "loss": 0.3634,
+      "step": 49433
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.398317465115108e-06,
+      "loss": 0.2639,
+      "step": 49434
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.397981629808742e-06,
+      "loss": 0.2098,
+      "step": 49435
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3976457945023763e-06,
+      "loss": 0.2046,
+      "step": 49436
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3973099591960105e-06,
+      "loss": 0.1231,
+      "step": 49437
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3969741238896447e-06,
+      "loss": 0.299,
+      "step": 49438
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.396638288583279e-06,
+      "loss": 0.3168,
+      "step": 49439
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3963024532769136e-06,
+      "loss": 0.5247,
+      "step": 49440
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3959666179705474e-06,
+      "loss": 0.3024,
+      "step": 49441
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3956307826641816e-06,
+      "loss": 0.2306,
+      "step": 49442
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3952949473578163e-06,
+      "loss": 0.241,
+      "step": 49443
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.39495911205145e-06,
+      "loss": 0.2665,
+      "step": 49444
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3946232767450843e-06,
+      "loss": 0.3187,
+      "step": 49445
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.394287441438719e-06,
+      "loss": 0.2079,
+      "step": 49446
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.393951606132353e-06,
+      "loss": 0.2214,
+      "step": 49447
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.393615770825987e-06,
+      "loss": 0.3419,
+      "step": 49448
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3932799355196216e-06,
+      "loss": 0.1796,
+      "step": 49449
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.392944100213256e-06,
+      "loss": 0.2027,
+      "step": 49450
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3926082649068896e-06,
+      "loss": 0.2079,
+      "step": 49451
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3922724296005243e-06,
+      "loss": 0.2913,
+      "step": 49452
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3919365942941585e-06,
+      "loss": 0.2939,
+      "step": 49453
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3916007589877923e-06,
+      "loss": 0.2124,
+      "step": 49454
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.391264923681427e-06,
+      "loss": 0.3089,
+      "step": 49455
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.390929088375061e-06,
+      "loss": 0.3385,
+      "step": 49456
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3905932530686954e-06,
+      "loss": 0.3649,
+      "step": 49457
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.39025741776233e-06,
+      "loss": 0.4207,
+      "step": 49458
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.389921582455964e-06,
+      "loss": 0.4434,
+      "step": 49459
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.389585747149598e-06,
+      "loss": 0.3968,
+      "step": 49460
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3892499118432327e-06,
+      "loss": 0.2011,
+      "step": 49461
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3889140765368665e-06,
+      "loss": 0.2029,
+      "step": 49462
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3885782412305007e-06,
+      "loss": 0.3373,
+      "step": 49463
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3882424059241354e-06,
+      "loss": 0.2939,
+      "step": 49464
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.387906570617769e-06,
+      "loss": 0.1011,
+      "step": 49465
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3875707353114034e-06,
+      "loss": 0.2103,
+      "step": 49466
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.387234900005038e-06,
+      "loss": 0.3226,
+      "step": 49467
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3868990646986723e-06,
+      "loss": 0.3077,
+      "step": 49468
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.386563229392306e-06,
+      "loss": 0.219,
+      "step": 49469
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3862273940859403e-06,
+      "loss": 0.3755,
+      "step": 49470
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.385891558779575e-06,
+      "loss": 0.2866,
+      "step": 49471
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3855557234732088e-06,
+      "loss": 0.2591,
+      "step": 49472
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.385219888166843e-06,
+      "loss": 0.2043,
+      "step": 49473
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3848840528604776e-06,
+      "loss": 0.1928,
+      "step": 49474
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.384548217554112e-06,
+      "loss": 0.1275,
+      "step": 49475
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3842123822477457e-06,
+      "loss": 0.3705,
+      "step": 49476
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3838765469413803e-06,
+      "loss": 0.2424,
+      "step": 49477
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3835407116350145e-06,
+      "loss": 0.3441,
+      "step": 49478
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3832048763286483e-06,
+      "loss": 0.1961,
+      "step": 49479
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.382869041022283e-06,
+      "loss": 0.1822,
+      "step": 49480
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.382533205715917e-06,
+      "loss": 0.1989,
+      "step": 49481
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.382197370409551e-06,
+      "loss": 0.4045,
+      "step": 49482
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3818615351031856e-06,
+      "loss": 0.223,
+      "step": 49483
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.38152569979682e-06,
+      "loss": 0.1765,
+      "step": 49484
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.381189864490454e-06,
+      "loss": 0.3108,
+      "step": 49485
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3808540291840887e-06,
+      "loss": 0.1613,
+      "step": 49486
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3805181938777225e-06,
+      "loss": 0.3236,
+      "step": 49487
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3801823585713568e-06,
+      "loss": 0.2342,
+      "step": 49488
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3798465232649914e-06,
+      "loss": 0.3105,
+      "step": 49489
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3795106879586252e-06,
+      "loss": 0.3976,
+      "step": 49490
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3791748526522594e-06,
+      "loss": 0.2917,
+      "step": 49491
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.378839017345894e-06,
+      "loss": 0.4241,
+      "step": 49492
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.378503182039528e-06,
+      "loss": 0.2448,
+      "step": 49493
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.378167346733162e-06,
+      "loss": 0.1803,
+      "step": 49494
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3778315114267968e-06,
+      "loss": 0.2463,
+      "step": 49495
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.377495676120431e-06,
+      "loss": 0.2032,
+      "step": 49496
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3771598408140648e-06,
+      "loss": 0.2985,
+      "step": 49497
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3768240055076994e-06,
+      "loss": 0.1746,
+      "step": 49498
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3764881702013337e-06,
+      "loss": 0.3504,
+      "step": 49499
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3761523348949675e-06,
+      "loss": 0.2718,
+      "step": 49500
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.375816499588602e-06,
+      "loss": 0.142,
+      "step": 49501
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3754806642822363e-06,
+      "loss": 0.1853,
+      "step": 49502
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3751448289758705e-06,
+      "loss": 0.1867,
+      "step": 49503
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.374808993669505e-06,
+      "loss": 0.2848,
+      "step": 49504
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.374473158363139e-06,
+      "loss": 0.253,
+      "step": 49505
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3741373230567732e-06,
+      "loss": 0.2116,
+      "step": 49506
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.373801487750408e-06,
+      "loss": 0.2668,
+      "step": 49507
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3734656524440417e-06,
+      "loss": 0.2884,
+      "step": 49508
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.373129817137676e-06,
+      "loss": 0.3007,
+      "step": 49509
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3727939818313105e-06,
+      "loss": 0.327,
+      "step": 49510
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3724581465249443e-06,
+      "loss": 0.3935,
+      "step": 49511
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3721223112185786e-06,
+      "loss": 0.3923,
+      "step": 49512
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.371786475912213e-06,
+      "loss": 0.3682,
+      "step": 49513
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3714506406058474e-06,
+      "loss": 0.3003,
+      "step": 49514
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3711148052994812e-06,
+      "loss": 0.3198,
+      "step": 49515
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3707789699931155e-06,
+      "loss": 0.2427,
+      "step": 49516
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.37044313468675e-06,
+      "loss": 0.1616,
+      "step": 49517
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.370107299380384e-06,
+      "loss": 0.2594,
+      "step": 49518
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.369771464074018e-06,
+      "loss": 0.2314,
+      "step": 49519
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3694356287676528e-06,
+      "loss": 0.1659,
+      "step": 49520
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.369099793461287e-06,
+      "loss": 0.2053,
+      "step": 49521
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.368763958154921e-06,
+      "loss": 0.3375,
+      "step": 49522
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3684281228485554e-06,
+      "loss": 0.2516,
+      "step": 49523
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3680922875421897e-06,
+      "loss": 0.2457,
+      "step": 49524
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3677564522358235e-06,
+      "loss": 0.2719,
+      "step": 49525
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.367420616929458e-06,
+      "loss": 0.2363,
+      "step": 49526
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.3670847816230923e-06,
+      "loss": 0.2309,
+      "step": 49527
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 3.366748946316726e-06,
+      "loss": 0.2445,
+      "step": 49528
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.366413111010361e-06,
+      "loss": 0.1973,
+      "step": 49529
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.366077275703995e-06,
+      "loss": 0.2056,
+      "step": 49530
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3657414403976292e-06,
+      "loss": 0.3175,
+      "step": 49531
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.365405605091264e-06,
+      "loss": 0.3289,
+      "step": 49532
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3650697697848977e-06,
+      "loss": 0.3763,
+      "step": 49533
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.364733934478532e-06,
+      "loss": 0.2972,
+      "step": 49534
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3643980991721666e-06,
+      "loss": 0.3129,
+      "step": 49535
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3640622638658004e-06,
+      "loss": 0.345,
+      "step": 49536
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3637264285594346e-06,
+      "loss": 0.3688,
+      "step": 49537
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3633905932530692e-06,
+      "loss": 0.3405,
+      "step": 49538
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.363054757946703e-06,
+      "loss": 0.2784,
+      "step": 49539
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3627189226403373e-06,
+      "loss": 0.3782,
+      "step": 49540
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.362383087333972e-06,
+      "loss": 0.5436,
+      "step": 49541
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.362047252027606e-06,
+      "loss": 0.1361,
+      "step": 49542
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.36171141672124e-06,
+      "loss": 0.1424,
+      "step": 49543
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3613755814148746e-06,
+      "loss": 0.3419,
+      "step": 49544
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.361039746108509e-06,
+      "loss": 0.3299,
+      "step": 49545
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3607039108021426e-06,
+      "loss": 0.502,
+      "step": 49546
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3603680754957772e-06,
+      "loss": 0.2836,
+      "step": 49547
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3600322401894115e-06,
+      "loss": 0.4534,
+      "step": 49548
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3596964048830457e-06,
+      "loss": 0.2,
+      "step": 49549
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.35936056957668e-06,
+      "loss": 0.2292,
+      "step": 49550
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.359024734270314e-06,
+      "loss": 0.4693,
+      "step": 49551
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3586888989639484e-06,
+      "loss": 0.2427,
+      "step": 49552
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.358353063657583e-06,
+      "loss": 0.2877,
+      "step": 49553
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.358017228351217e-06,
+      "loss": 0.24,
+      "step": 49554
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.357681393044851e-06,
+      "loss": 0.3435,
+      "step": 49555
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3573455577384857e-06,
+      "loss": 0.4055,
+      "step": 49556
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3570097224321195e-06,
+      "loss": 0.2679,
+      "step": 49557
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3566738871257537e-06,
+      "loss": 0.3846,
+      "step": 49558
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3563380518193884e-06,
+      "loss": 0.2032,
+      "step": 49559
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3560022165130226e-06,
+      "loss": 0.1947,
+      "step": 49560
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3556663812066564e-06,
+      "loss": 0.3449,
+      "step": 49561
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3553305459002906e-06,
+      "loss": 0.3997,
+      "step": 49562
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3549947105939252e-06,
+      "loss": 0.262,
+      "step": 49563
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.354658875287559e-06,
+      "loss": 0.167,
+      "step": 49564
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3543230399811933e-06,
+      "loss": 0.3125,
+      "step": 49565
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.353987204674828e-06,
+      "loss": 0.2362,
+      "step": 49566
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3536513693684617e-06,
+      "loss": 0.3026,
+      "step": 49567
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.353315534062096e-06,
+      "loss": 0.2364,
+      "step": 49568
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3529796987557306e-06,
+      "loss": 0.3746,
+      "step": 49569
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.352643863449365e-06,
+      "loss": 0.2713,
+      "step": 49570
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3523080281429986e-06,
+      "loss": 0.3766,
+      "step": 49571
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3519721928366333e-06,
+      "loss": 0.3145,
+      "step": 49572
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3516363575302675e-06,
+      "loss": 0.3216,
+      "step": 49573
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3513005222239013e-06,
+      "loss": 0.2511,
+      "step": 49574
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.350964686917536e-06,
+      "loss": 0.2155,
+      "step": 49575
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.35062885161117e-06,
+      "loss": 0.197,
+      "step": 49576
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3502930163048044e-06,
+      "loss": 0.3693,
+      "step": 49577
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3499571809984386e-06,
+      "loss": 0.3182,
+      "step": 49578
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.349621345692073e-06,
+      "loss": 0.2506,
+      "step": 49579
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.349285510385707e-06,
+      "loss": 0.3568,
+      "step": 49580
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3489496750793417e-06,
+      "loss": 0.2263,
+      "step": 49581
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3486138397729755e-06,
+      "loss": 0.2376,
+      "step": 49582
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3482780044666097e-06,
+      "loss": 0.3682,
+      "step": 49583
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3479421691602444e-06,
+      "loss": 0.1235,
+      "step": 49584
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.347606333853878e-06,
+      "loss": 0.2807,
+      "step": 49585
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3472704985475124e-06,
+      "loss": 0.3332,
+      "step": 49586
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.346934663241147e-06,
+      "loss": 0.3565,
+      "step": 49587
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3465988279347813e-06,
+      "loss": 0.1591,
+      "step": 49588
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.346262992628415e-06,
+      "loss": 0.1514,
+      "step": 49589
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3459271573220497e-06,
+      "loss": 0.3049,
+      "step": 49590
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.345591322015684e-06,
+      "loss": 0.2446,
+      "step": 49591
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3452554867093177e-06,
+      "loss": 0.2012,
+      "step": 49592
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3449196514029524e-06,
+      "loss": 0.2822,
+      "step": 49593
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3445838160965866e-06,
+      "loss": 0.2074,
+      "step": 49594
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3442479807902204e-06,
+      "loss": 0.2794,
+      "step": 49595
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.343912145483855e-06,
+      "loss": 0.1818,
+      "step": 49596
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3435763101774893e-06,
+      "loss": 0.1883,
+      "step": 49597
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3432404748711235e-06,
+      "loss": 0.2154,
+      "step": 49598
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.342904639564758e-06,
+      "loss": 0.4444,
+      "step": 49599
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.342568804258392e-06,
+      "loss": 0.2058,
+      "step": 49600
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.342232968952026e-06,
+      "loss": 0.3005,
+      "step": 49601
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.341897133645661e-06,
+      "loss": 0.1719,
+      "step": 49602
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3415612983392946e-06,
+      "loss": 0.2163,
+      "step": 49603
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.341225463032929e-06,
+      "loss": 0.1839,
+      "step": 49604
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.340889627726563e-06,
+      "loss": 0.2971,
+      "step": 49605
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3405537924201973e-06,
+      "loss": 0.3521,
+      "step": 49606
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3402179571138315e-06,
+      "loss": 0.3447,
+      "step": 49607
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3398821218074657e-06,
+      "loss": 0.2078,
+      "step": 49608
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3395462865011004e-06,
+      "loss": 0.3722,
+      "step": 49609
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.339210451194734e-06,
+      "loss": 0.2829,
+      "step": 49610
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3388746158883684e-06,
+      "loss": 0.2305,
+      "step": 49611
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.338538780582003e-06,
+      "loss": 0.3893,
+      "step": 49612
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.338202945275637e-06,
+      "loss": 0.1909,
+      "step": 49613
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.337867109969271e-06,
+      "loss": 0.3908,
+      "step": 49614
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3375312746629057e-06,
+      "loss": 0.246,
+      "step": 49615
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.33719543935654e-06,
+      "loss": 0.24,
+      "step": 49616
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3368596040501738e-06,
+      "loss": 0.2552,
+      "step": 49617
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3365237687438084e-06,
+      "loss": 0.2793,
+      "step": 49618
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3361879334374426e-06,
+      "loss": 0.2617,
+      "step": 49619
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3358520981310764e-06,
+      "loss": 0.1653,
+      "step": 49620
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.335516262824711e-06,
+      "loss": 0.2882,
+      "step": 49621
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3351804275183453e-06,
+      "loss": 0.187,
+      "step": 49622
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.334844592211979e-06,
+      "loss": 0.1923,
+      "step": 49623
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3345087569056138e-06,
+      "loss": 0.2512,
+      "step": 49624
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.334172921599248e-06,
+      "loss": 0.2563,
+      "step": 49625
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.333837086292882e-06,
+      "loss": 0.1994,
+      "step": 49626
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.333501250986517e-06,
+      "loss": 0.2943,
+      "step": 49627
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3331654156801506e-06,
+      "loss": 0.205,
+      "step": 49628
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.332829580373785e-06,
+      "loss": 0.2599,
+      "step": 49629
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3324937450674195e-06,
+      "loss": 0.1883,
+      "step": 49630
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3321579097610533e-06,
+      "loss": 0.2303,
+      "step": 49631
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3318220744546875e-06,
+      "loss": 0.2814,
+      "step": 49632
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.331486239148322e-06,
+      "loss": 0.344,
+      "step": 49633
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3311504038419564e-06,
+      "loss": 0.416,
+      "step": 49634
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3308145685355902e-06,
+      "loss": 0.204,
+      "step": 49635
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.330478733229225e-06,
+      "loss": 0.2363,
+      "step": 49636
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.330142897922859e-06,
+      "loss": 0.2667,
+      "step": 49637
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.329807062616493e-06,
+      "loss": 0.315,
+      "step": 49638
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3294712273101275e-06,
+      "loss": 0.1824,
+      "step": 49639
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3291353920037618e-06,
+      "loss": 0.262,
+      "step": 49640
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3287995566973956e-06,
+      "loss": 0.3062,
+      "step": 49641
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.32846372139103e-06,
+      "loss": 0.321,
+      "step": 49642
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3281278860846644e-06,
+      "loss": 0.1797,
+      "step": 49643
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3277920507782987e-06,
+      "loss": 0.1337,
+      "step": 49644
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3274562154719333e-06,
+      "loss": 0.3121,
+      "step": 49645
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.327120380165567e-06,
+      "loss": 0.4222,
+      "step": 49646
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3267845448592013e-06,
+      "loss": 0.1715,
+      "step": 49647
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.326448709552836e-06,
+      "loss": 0.3287,
+      "step": 49648
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3261128742464698e-06,
+      "loss": 0.3814,
+      "step": 49649
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.325777038940104e-06,
+      "loss": 0.224,
+      "step": 49650
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3254412036337382e-06,
+      "loss": 0.2465,
+      "step": 49651
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3251053683273724e-06,
+      "loss": 0.1323,
+      "step": 49652
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3247695330210067e-06,
+      "loss": 0.2905,
+      "step": 49653
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.324433697714641e-06,
+      "loss": 0.323,
+      "step": 49654
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3240978624082755e-06,
+      "loss": 0.2815,
+      "step": 49655
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3237620271019093e-06,
+      "loss": 0.3586,
+      "step": 49656
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3234261917955436e-06,
+      "loss": 0.2292,
+      "step": 49657
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.323090356489178e-06,
+      "loss": 0.3589,
+      "step": 49658
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.322754521182812e-06,
+      "loss": 0.1836,
+      "step": 49659
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3224186858764462e-06,
+      "loss": 0.314,
+      "step": 49660
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.322082850570081e-06,
+      "loss": 0.1658,
+      "step": 49661
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.321747015263715e-06,
+      "loss": 0.3429,
+      "step": 49662
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.321411179957349e-06,
+      "loss": 0.2451,
+      "step": 49663
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3210753446509836e-06,
+      "loss": 0.1806,
+      "step": 49664
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3207395093446178e-06,
+      "loss": 0.226,
+      "step": 49665
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3204036740382516e-06,
+      "loss": 0.27,
+      "step": 49666
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3200678387318862e-06,
+      "loss": 0.3166,
+      "step": 49667
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3197320034255204e-06,
+      "loss": 0.25,
+      "step": 49668
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3193961681191542e-06,
+      "loss": 0.2413,
+      "step": 49669
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.319060332812789e-06,
+      "loss": 0.3055,
+      "step": 49670
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.318724497506423e-06,
+      "loss": 0.3283,
+      "step": 49671
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3183886622000573e-06,
+      "loss": 0.2993,
+      "step": 49672
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.318052826893692e-06,
+      "loss": 0.3818,
+      "step": 49673
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.317716991587326e-06,
+      "loss": 0.352,
+      "step": 49674
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.31738115628096e-06,
+      "loss": 0.1741,
+      "step": 49675
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3170453209745947e-06,
+      "loss": 0.2642,
+      "step": 49676
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3167094856682285e-06,
+      "loss": 0.2964,
+      "step": 49677
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3163736503618627e-06,
+      "loss": 0.3892,
+      "step": 49678
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3160378150554973e-06,
+      "loss": 0.2809,
+      "step": 49679
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.315701979749131e-06,
+      "loss": 0.2062,
+      "step": 49680
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3153661444427654e-06,
+      "loss": 0.1663,
+      "step": 49681
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3150303091364e-06,
+      "loss": 0.1699,
+      "step": 49682
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3146944738300342e-06,
+      "loss": 0.2651,
+      "step": 49683
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.314358638523668e-06,
+      "loss": 0.2833,
+      "step": 49684
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3140228032173027e-06,
+      "loss": 0.3447,
+      "step": 49685
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.313686967910937e-06,
+      "loss": 0.2825,
+      "step": 49686
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3133511326045707e-06,
+      "loss": 0.3842,
+      "step": 49687
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3130152972982054e-06,
+      "loss": 0.2723,
+      "step": 49688
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3126794619918396e-06,
+      "loss": 0.2386,
+      "step": 49689
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.312343626685474e-06,
+      "loss": 0.3074,
+      "step": 49690
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.312007791379108e-06,
+      "loss": 0.3152,
+      "step": 49691
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3116719560727422e-06,
+      "loss": 0.399,
+      "step": 49692
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3113361207663765e-06,
+      "loss": 0.1922,
+      "step": 49693
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.311000285460011e-06,
+      "loss": 0.2777,
+      "step": 49694
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.310664450153645e-06,
+      "loss": 0.3537,
+      "step": 49695
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.310328614847279e-06,
+      "loss": 0.1565,
+      "step": 49696
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.309992779540913e-06,
+      "loss": 0.2003,
+      "step": 49697
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3096569442345476e-06,
+      "loss": 0.1814,
+      "step": 49698
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.309321108928182e-06,
+      "loss": 0.3028,
+      "step": 49699
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.308985273621816e-06,
+      "loss": 0.1896,
+      "step": 49700
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3086494383154507e-06,
+      "loss": 0.3017,
+      "step": 49701
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3083136030090845e-06,
+      "loss": 0.1809,
+      "step": 49702
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3079777677027187e-06,
+      "loss": 0.2946,
+      "step": 49703
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3076419323963534e-06,
+      "loss": 0.3269,
+      "step": 49704
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.307306097089987e-06,
+      "loss": 0.3694,
+      "step": 49705
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3069702617836214e-06,
+      "loss": 0.4391,
+      "step": 49706
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.306634426477256e-06,
+      "loss": 0.4082,
+      "step": 49707
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.30629859117089e-06,
+      "loss": 0.2738,
+      "step": 49708
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.305962755864524e-06,
+      "loss": 0.3193,
+      "step": 49709
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3056269205581587e-06,
+      "loss": 0.2319,
+      "step": 49710
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.305291085251793e-06,
+      "loss": 0.3573,
+      "step": 49711
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3049552499454267e-06,
+      "loss": 0.4288,
+      "step": 49712
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3046194146390614e-06,
+      "loss": 0.2524,
+      "step": 49713
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3042835793326956e-06,
+      "loss": 0.3932,
+      "step": 49714
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3039477440263294e-06,
+      "loss": 0.1921,
+      "step": 49715
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.303611908719964e-06,
+      "loss": 0.4158,
+      "step": 49716
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3032760734135983e-06,
+      "loss": 0.1657,
+      "step": 49717
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3029402381072325e-06,
+      "loss": 0.2498,
+      "step": 49718
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3026044028008667e-06,
+      "loss": 0.4349,
+      "step": 49719
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.302268567494501e-06,
+      "loss": 0.1899,
+      "step": 49720
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.301932732188135e-06,
+      "loss": 0.1737,
+      "step": 49721
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.30159689688177e-06,
+      "loss": 0.4225,
+      "step": 49722
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3012610615754036e-06,
+      "loss": 0.3056,
+      "step": 49723
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.300925226269038e-06,
+      "loss": 0.25,
+      "step": 49724
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3005893909626725e-06,
+      "loss": 0.1807,
+      "step": 49725
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3002535556563063e-06,
+      "loss": 0.2386,
+      "step": 49726
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2999177203499405e-06,
+      "loss": 0.2217,
+      "step": 49727
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.299581885043575e-06,
+      "loss": 0.38,
+      "step": 49728
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2992460497372094e-06,
+      "loss": 0.1905,
+      "step": 49729
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.298910214430843e-06,
+      "loss": 0.16,
+      "step": 49730
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.298574379124478e-06,
+      "loss": 0.3262,
+      "step": 49731
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.298238543818112e-06,
+      "loss": 0.3484,
+      "step": 49732
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.297902708511746e-06,
+      "loss": 0.2354,
+      "step": 49733
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2975668732053805e-06,
+      "loss": 0.4121,
+      "step": 49734
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2972310378990147e-06,
+      "loss": 0.1268,
+      "step": 49735
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2968952025926485e-06,
+      "loss": 0.3114,
+      "step": 49736
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.296559367286283e-06,
+      "loss": 0.3163,
+      "step": 49737
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2962235319799174e-06,
+      "loss": 0.1186,
+      "step": 49738
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2958876966735516e-06,
+      "loss": 0.3123,
+      "step": 49739
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2955518613671863e-06,
+      "loss": 0.4574,
+      "step": 49740
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.29521602606082e-06,
+      "loss": 0.2821,
+      "step": 49741
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2948801907544543e-06,
+      "loss": 0.3097,
+      "step": 49742
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.294544355448088e-06,
+      "loss": 0.2678,
+      "step": 49743
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2942085201417227e-06,
+      "loss": 0.248,
+      "step": 49744
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.293872684835357e-06,
+      "loss": 0.1852,
+      "step": 49745
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.293536849528991e-06,
+      "loss": 0.4514,
+      "step": 49746
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2932010142226254e-06,
+      "loss": 0.3207,
+      "step": 49747
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2928651789162596e-06,
+      "loss": 0.2238,
+      "step": 49748
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.292529343609894e-06,
+      "loss": 0.1497,
+      "step": 49749
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2921935083035285e-06,
+      "loss": 0.2437,
+      "step": 49750
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2918576729971623e-06,
+      "loss": 0.2942,
+      "step": 49751
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2915218376907965e-06,
+      "loss": 0.2044,
+      "step": 49752
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.291186002384431e-06,
+      "loss": 0.2424,
+      "step": 49753
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.290850167078065e-06,
+      "loss": 0.1491,
+      "step": 49754
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.290514331771699e-06,
+      "loss": 0.2758,
+      "step": 49755
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.290178496465334e-06,
+      "loss": 0.229,
+      "step": 49756
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.289842661158968e-06,
+      "loss": 0.359,
+      "step": 49757
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.289506825852602e-06,
+      "loss": 0.3407,
+      "step": 49758
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2891709905462365e-06,
+      "loss": 0.0928,
+      "step": 49759
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2888351552398707e-06,
+      "loss": 0.229,
+      "step": 49760
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2884993199335045e-06,
+      "loss": 0.2496,
+      "step": 49761
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.288163484627139e-06,
+      "loss": 0.4729,
+      "step": 49762
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2878276493207734e-06,
+      "loss": 0.2548,
+      "step": 49763
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2874918140144076e-06,
+      "loss": 0.5629,
+      "step": 49764
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.287155978708042e-06,
+      "loss": 0.2255,
+      "step": 49765
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.286820143401676e-06,
+      "loss": 0.3516,
+      "step": 49766
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2864843080953103e-06,
+      "loss": 0.5862,
+      "step": 49767
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.286148472788945e-06,
+      "loss": 0.1732,
+      "step": 49768
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2858126374825788e-06,
+      "loss": 0.1636,
+      "step": 49769
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.285476802176213e-06,
+      "loss": 0.224,
+      "step": 49770
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2851409668698476e-06,
+      "loss": 0.1893,
+      "step": 49771
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2848051315634814e-06,
+      "loss": 0.1882,
+      "step": 49772
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2844692962571156e-06,
+      "loss": 0.2311,
+      "step": 49773
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2841334609507503e-06,
+      "loss": 0.4601,
+      "step": 49774
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2837976256443845e-06,
+      "loss": 0.3639,
+      "step": 49775
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2834617903380183e-06,
+      "loss": 0.2881,
+      "step": 49776
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.283125955031653e-06,
+      "loss": 0.2346,
+      "step": 49777
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.282790119725287e-06,
+      "loss": 0.1206,
+      "step": 49778
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.282454284418921e-06,
+      "loss": 0.2448,
+      "step": 49779
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2821184491125556e-06,
+      "loss": 0.2398,
+      "step": 49780
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.28178261380619e-06,
+      "loss": 0.2647,
+      "step": 49781
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2814467784998237e-06,
+      "loss": 0.2202,
+      "step": 49782
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2811109431934583e-06,
+      "loss": 0.1153,
+      "step": 49783
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2807751078870925e-06,
+      "loss": 0.3434,
+      "step": 49784
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2804392725807268e-06,
+      "loss": 0.1631,
+      "step": 49785
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2801034372743606e-06,
+      "loss": 0.2303,
+      "step": 49786
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.279767601967995e-06,
+      "loss": 0.2609,
+      "step": 49787
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2794317666616294e-06,
+      "loss": 0.4752,
+      "step": 49788
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2790959313552632e-06,
+      "loss": 0.2591,
+      "step": 49789
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.278760096048898e-06,
+      "loss": 0.2215,
+      "step": 49790
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.278424260742532e-06,
+      "loss": 0.2745,
+      "step": 49791
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2780884254361663e-06,
+      "loss": 0.243,
+      "step": 49792
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2777525901298006e-06,
+      "loss": 0.3519,
+      "step": 49793
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2774167548234348e-06,
+      "loss": 0.2036,
+      "step": 49794
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.277080919517069e-06,
+      "loss": 0.3613,
+      "step": 49795
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2767450842107036e-06,
+      "loss": 0.2213,
+      "step": 49796
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2764092489043374e-06,
+      "loss": 0.3237,
+      "step": 49797
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2760734135979717e-06,
+      "loss": 0.2379,
+      "step": 49798
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2757375782916063e-06,
+      "loss": 0.1599,
+      "step": 49799
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.27540174298524e-06,
+      "loss": 0.194,
+      "step": 49800
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2750659076788743e-06,
+      "loss": 0.1751,
+      "step": 49801
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.274730072372509e-06,
+      "loss": 0.2482,
+      "step": 49802
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2743942370661432e-06,
+      "loss": 0.2516,
+      "step": 49803
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.274058401759777e-06,
+      "loss": 0.2126,
+      "step": 49804
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2737225664534117e-06,
+      "loss": 0.2994,
+      "step": 49805
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.273386731147046e-06,
+      "loss": 0.3063,
+      "step": 49806
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2730508958406797e-06,
+      "loss": 0.3658,
+      "step": 49807
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2727150605343143e-06,
+      "loss": 0.301,
+      "step": 49808
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2723792252279486e-06,
+      "loss": 0.2234,
+      "step": 49809
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2720433899215824e-06,
+      "loss": 0.296,
+      "step": 49810
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.271707554615217e-06,
+      "loss": 0.3231,
+      "step": 49811
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2713717193088512e-06,
+      "loss": 0.2194,
+      "step": 49812
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2710358840024855e-06,
+      "loss": 0.3257,
+      "step": 49813
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.27070004869612e-06,
+      "loss": 0.3106,
+      "step": 49814
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.270364213389754e-06,
+      "loss": 0.4327,
+      "step": 49815
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.270028378083388e-06,
+      "loss": 0.4289,
+      "step": 49816
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2696925427770228e-06,
+      "loss": 0.1619,
+      "step": 49817
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2693567074706566e-06,
+      "loss": 0.345,
+      "step": 49818
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.269020872164291e-06,
+      "loss": 0.2356,
+      "step": 49819
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2686850368579254e-06,
+      "loss": 0.2951,
+      "step": 49820
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2683492015515592e-06,
+      "loss": 0.3994,
+      "step": 49821
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2680133662451935e-06,
+      "loss": 0.2798,
+      "step": 49822
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.267677530938828e-06,
+      "loss": 0.2236,
+      "step": 49823
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2673416956324623e-06,
+      "loss": 0.3188,
+      "step": 49824
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.267005860326096e-06,
+      "loss": 0.2493,
+      "step": 49825
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2666700250197308e-06,
+      "loss": 0.1705,
+      "step": 49826
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.266334189713365e-06,
+      "loss": 0.3003,
+      "step": 49827
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.265998354406999e-06,
+      "loss": 0.2938,
+      "step": 49828
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2656625191006335e-06,
+      "loss": 0.2724,
+      "step": 49829
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2653266837942677e-06,
+      "loss": 0.1072,
+      "step": 49830
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.264990848487902e-06,
+      "loss": 0.1789,
+      "step": 49831
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2646550131815357e-06,
+      "loss": 0.1536,
+      "step": 49832
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2643191778751704e-06,
+      "loss": 0.2501,
+      "step": 49833
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2639833425688046e-06,
+      "loss": 0.372,
+      "step": 49834
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2636475072624384e-06,
+      "loss": 0.5106,
+      "step": 49835
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.263311671956073e-06,
+      "loss": 0.3692,
+      "step": 49836
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2629758366497072e-06,
+      "loss": 0.3568,
+      "step": 49837
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.262640001343341e-06,
+      "loss": 0.2236,
+      "step": 49838
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2623041660369757e-06,
+      "loss": 0.2736,
+      "step": 49839
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.26196833073061e-06,
+      "loss": 0.1548,
+      "step": 49840
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.261632495424244e-06,
+      "loss": 0.1833,
+      "step": 49841
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.261296660117879e-06,
+      "loss": 0.3334,
+      "step": 49842
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2609608248115126e-06,
+      "loss": 0.1603,
+      "step": 49843
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.260624989505147e-06,
+      "loss": 0.2678,
+      "step": 49844
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2602891541987815e-06,
+      "loss": 0.2792,
+      "step": 49845
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2599533188924153e-06,
+      "loss": 0.2544,
+      "step": 49846
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2596174835860495e-06,
+      "loss": 0.2473,
+      "step": 49847
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.259281648279684e-06,
+      "loss": 0.302,
+      "step": 49848
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.258945812973318e-06,
+      "loss": 0.2691,
+      "step": 49849
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.258609977666952e-06,
+      "loss": 0.2013,
+      "step": 49850
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.258274142360587e-06,
+      "loss": 0.2623,
+      "step": 49851
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.257938307054221e-06,
+      "loss": 0.3534,
+      "step": 49852
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.257602471747855e-06,
+      "loss": 0.2393,
+      "step": 49853
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2572666364414895e-06,
+      "loss": 0.2224,
+      "step": 49854
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2569308011351237e-06,
+      "loss": 0.2984,
+      "step": 49855
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2565949658287575e-06,
+      "loss": 0.3252,
+      "step": 49856
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.256259130522392e-06,
+      "loss": 0.1573,
+      "step": 49857
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2559232952160264e-06,
+      "loss": 0.3138,
+      "step": 49858
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2555874599096606e-06,
+      "loss": 0.3555,
+      "step": 49859
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.255251624603295e-06,
+      "loss": 0.2319,
+      "step": 49860
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.254915789296929e-06,
+      "loss": 0.2832,
+      "step": 49861
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2545799539905633e-06,
+      "loss": 0.3078,
+      "step": 49862
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.254244118684198e-06,
+      "loss": 0.3007,
+      "step": 49863
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2539082833778317e-06,
+      "loss": 0.2445,
+      "step": 49864
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.253572448071466e-06,
+      "loss": 0.3487,
+      "step": 49865
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2532366127651006e-06,
+      "loss": 0.2574,
+      "step": 49866
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2529007774587344e-06,
+      "loss": 0.1814,
+      "step": 49867
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2525649421523686e-06,
+      "loss": 0.2665,
+      "step": 49868
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2522291068460033e-06,
+      "loss": 0.2245,
+      "step": 49869
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2518932715396375e-06,
+      "loss": 0.2989,
+      "step": 49870
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2515574362332713e-06,
+      "loss": 0.2074,
+      "step": 49871
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.251221600926906e-06,
+      "loss": 0.243,
+      "step": 49872
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.25088576562054e-06,
+      "loss": 0.2797,
+      "step": 49873
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.250549930314174e-06,
+      "loss": 0.2948,
+      "step": 49874
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2502140950078086e-06,
+      "loss": 0.3582,
+      "step": 49875
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.249878259701443e-06,
+      "loss": 0.3019,
+      "step": 49876
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2495424243950766e-06,
+      "loss": 0.2885,
+      "step": 49877
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.249206589088711e-06,
+      "loss": 0.1662,
+      "step": 49878
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2488707537823455e-06,
+      "loss": 0.3397,
+      "step": 49879
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2485349184759797e-06,
+      "loss": 0.5971,
+      "step": 49880
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2481990831696135e-06,
+      "loss": 0.2832,
+      "step": 49881
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.247863247863248e-06,
+      "loss": 0.1899,
+      "step": 49882
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2475274125568824e-06,
+      "loss": 0.3086,
+      "step": 49883
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.247191577250516e-06,
+      "loss": 0.3146,
+      "step": 49884
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.246855741944151e-06,
+      "loss": 0.3054,
+      "step": 49885
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.246519906637785e-06,
+      "loss": 0.1617,
+      "step": 49886
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2461840713314193e-06,
+      "loss": 0.3947,
+      "step": 49887
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2458482360250535e-06,
+      "loss": 0.3465,
+      "step": 49888
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2455124007186877e-06,
+      "loss": 0.28,
+      "step": 49889
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.245176565412322e-06,
+      "loss": 0.3318,
+      "step": 49890
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2448407301059566e-06,
+      "loss": 0.098,
+      "step": 49891
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2445048947995904e-06,
+      "loss": 0.4554,
+      "step": 49892
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2441690594932246e-06,
+      "loss": 0.1576,
+      "step": 49893
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2438332241868593e-06,
+      "loss": 0.2572,
+      "step": 49894
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.243497388880493e-06,
+      "loss": 0.1996,
+      "step": 49895
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2431615535741273e-06,
+      "loss": 0.2035,
+      "step": 49896
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.242825718267762e-06,
+      "loss": 0.3911,
+      "step": 49897
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.242489882961396e-06,
+      "loss": 0.2769,
+      "step": 49898
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.24215404765503e-06,
+      "loss": 0.3288,
+      "step": 49899
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2418182123486646e-06,
+      "loss": 0.2981,
+      "step": 49900
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.241482377042299e-06,
+      "loss": 0.2555,
+      "step": 49901
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2411465417359326e-06,
+      "loss": 0.2901,
+      "step": 49902
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2408107064295673e-06,
+      "loss": 0.2702,
+      "step": 49903
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2404748711232015e-06,
+      "loss": 0.2355,
+      "step": 49904
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2401390358168357e-06,
+      "loss": 0.2918,
+      "step": 49905
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.23980320051047e-06,
+      "loss": 0.2706,
+      "step": 49906
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.239467365204104e-06,
+      "loss": 0.186,
+      "step": 49907
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2391315298977384e-06,
+      "loss": 0.3074,
+      "step": 49908
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.238795694591373e-06,
+      "loss": 0.3737,
+      "step": 49909
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.238459859285007e-06,
+      "loss": 0.3018,
+      "step": 49910
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.238124023978641e-06,
+      "loss": 0.2508,
+      "step": 49911
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2377881886722757e-06,
+      "loss": 0.2596,
+      "step": 49912
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2374523533659095e-06,
+      "loss": 0.2498,
+      "step": 49913
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2371165180595438e-06,
+      "loss": 0.4709,
+      "step": 49914
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2367806827531784e-06,
+      "loss": 0.2177,
+      "step": 49915
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2364448474468126e-06,
+      "loss": 0.1376,
+      "step": 49916
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2361090121404464e-06,
+      "loss": 0.1797,
+      "step": 49917
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.235773176834081e-06,
+      "loss": 0.2783,
+      "step": 49918
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2354373415277153e-06,
+      "loss": 0.2326,
+      "step": 49919
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.235101506221349e-06,
+      "loss": 0.2338,
+      "step": 49920
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2347656709149833e-06,
+      "loss": 0.4018,
+      "step": 49921
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.234429835608618e-06,
+      "loss": 0.4531,
+      "step": 49922
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2340940003022518e-06,
+      "loss": 0.1767,
+      "step": 49923
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.233758164995886e-06,
+      "loss": 0.2805,
+      "step": 49924
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 3.2334223296895206e-06,
+      "loss": 0.308,
+      "step": 49925
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.233086494383155e-06,
+      "loss": 0.194,
+      "step": 49926
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2327506590767887e-06,
+      "loss": 0.256,
+      "step": 49927
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2324148237704233e-06,
+      "loss": 0.4133,
+      "step": 49928
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2320789884640575e-06,
+      "loss": 0.1627,
+      "step": 49929
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2317431531576913e-06,
+      "loss": 0.1408,
+      "step": 49930
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.231407317851326e-06,
+      "loss": 0.1284,
+      "step": 49931
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.23107148254496e-06,
+      "loss": 0.3018,
+      "step": 49932
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2307356472385944e-06,
+      "loss": 0.2541,
+      "step": 49933
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2303998119322287e-06,
+      "loss": 0.2484,
+      "step": 49934
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.230063976625863e-06,
+      "loss": 0.2325,
+      "step": 49935
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.229728141319497e-06,
+      "loss": 0.199,
+      "step": 49936
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2293923060131318e-06,
+      "loss": 0.2487,
+      "step": 49937
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2290564707067656e-06,
+      "loss": 0.3037,
+      "step": 49938
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2287206354003998e-06,
+      "loss": 0.3369,
+      "step": 49939
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2283848000940344e-06,
+      "loss": 0.3147,
+      "step": 49940
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2280489647876682e-06,
+      "loss": 0.1819,
+      "step": 49941
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2277131294813024e-06,
+      "loss": 0.2304,
+      "step": 49942
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.227377294174937e-06,
+      "loss": 0.2269,
+      "step": 49943
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2270414588685713e-06,
+      "loss": 0.4106,
+      "step": 49944
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.226705623562205e-06,
+      "loss": 0.2269,
+      "step": 49945
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2263697882558398e-06,
+      "loss": 0.2078,
+      "step": 49946
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.226033952949474e-06,
+      "loss": 0.269,
+      "step": 49947
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.225698117643108e-06,
+      "loss": 0.3827,
+      "step": 49948
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2253622823367424e-06,
+      "loss": 0.3551,
+      "step": 49949
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2250264470303767e-06,
+      "loss": 0.4778,
+      "step": 49950
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2246906117240105e-06,
+      "loss": 0.2551,
+      "step": 49951
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.224354776417645e-06,
+      "loss": 0.3875,
+      "step": 49952
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2240189411112793e-06,
+      "loss": 0.1601,
+      "step": 49953
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2236831058049136e-06,
+      "loss": 0.2296,
+      "step": 49954
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.223347270498548e-06,
+      "loss": 0.2738,
+      "step": 49955
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.223011435192182e-06,
+      "loss": 0.1913,
+      "step": 49956
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2226755998858162e-06,
+      "loss": 0.3719,
+      "step": 49957
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.222339764579451e-06,
+      "loss": 0.3769,
+      "step": 49958
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2220039292730847e-06,
+      "loss": 0.2066,
+      "step": 49959
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.221668093966719e-06,
+      "loss": 0.2588,
+      "step": 49960
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2213322586603535e-06,
+      "loss": 0.2835,
+      "step": 49961
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2209964233539873e-06,
+      "loss": 0.2466,
+      "step": 49962
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2206605880476216e-06,
+      "loss": 0.1649,
+      "step": 49963
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2203247527412562e-06,
+      "loss": 0.3751,
+      "step": 49964
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2199889174348904e-06,
+      "loss": 0.3573,
+      "step": 49965
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2196530821285242e-06,
+      "loss": 0.2936,
+      "step": 49966
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2193172468221585e-06,
+      "loss": 0.2412,
+      "step": 49967
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.218981411515793e-06,
+      "loss": 0.2653,
+      "step": 49968
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.218645576209427e-06,
+      "loss": 0.1962,
+      "step": 49969
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.218309740903061e-06,
+      "loss": 0.3734,
+      "step": 49970
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2179739055966958e-06,
+      "loss": 0.1687,
+      "step": 49971
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.21763807029033e-06,
+      "loss": 0.4196,
+      "step": 49972
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.217302234983964e-06,
+      "loss": 0.1232,
+      "step": 49973
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2169663996775985e-06,
+      "loss": 0.3049,
+      "step": 49974
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2166305643712327e-06,
+      "loss": 0.2745,
+      "step": 49975
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2162947290648665e-06,
+      "loss": 0.2,
+      "step": 49976
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.215958893758501e-06,
+      "loss": 0.287,
+      "step": 49977
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2156230584521354e-06,
+      "loss": 0.1977,
+      "step": 49978
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.215287223145769e-06,
+      "loss": 0.1975,
+      "step": 49979
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.214951387839404e-06,
+      "loss": 0.3333,
+      "step": 49980
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.214615552533038e-06,
+      "loss": 0.2729,
+      "step": 49981
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2142797172266722e-06,
+      "loss": 0.2441,
+      "step": 49982
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.213943881920307e-06,
+      "loss": 0.3442,
+      "step": 49983
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2136080466139407e-06,
+      "loss": 0.1503,
+      "step": 49984
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.213272211307575e-06,
+      "loss": 0.1906,
+      "step": 49985
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2129363760012096e-06,
+      "loss": 0.2551,
+      "step": 49986
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2126005406948434e-06,
+      "loss": 0.3585,
+      "step": 49987
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2122647053884776e-06,
+      "loss": 0.3298,
+      "step": 49988
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2119288700821122e-06,
+      "loss": 0.1546,
+      "step": 49989
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.211593034775746e-06,
+      "loss": 0.2157,
+      "step": 49990
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2112571994693803e-06,
+      "loss": 0.3525,
+      "step": 49991
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.210921364163015e-06,
+      "loss": 0.238,
+      "step": 49992
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.210585528856649e-06,
+      "loss": 0.3393,
+      "step": 49993
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.210249693550283e-06,
+      "loss": 0.1852,
+      "step": 49994
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2099138582439176e-06,
+      "loss": 0.2693,
+      "step": 49995
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.209578022937552e-06,
+      "loss": 0.1683,
+      "step": 49996
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2092421876311856e-06,
+      "loss": 0.2407,
+      "step": 49997
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2089063523248203e-06,
+      "loss": 0.2634,
+      "step": 49998
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2085705170184545e-06,
+      "loss": 0.2186,
+      "step": 49999
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2082346817120887e-06,
+      "loss": 0.1856,
+      "step": 50000
+    },
+    {
+      "epoch": 2.52,
+      "eval_accuracy": 0.8842667078071393,
+      "eval_f1_microaverage": 0.8842667078071393,
+      "eval_loss": 0.30753660202026367,
+      "eval_precision_macroaverage": 0.7649704368678851,
+      "eval_precision_microaverage": 0.8842667078071393,
+      "eval_recall_macroaverage": 0.6898798373146766,
+      "eval_recall_microaverage": 0.8842667078071393,
+      "eval_runtime": 1174.0166,
+      "eval_samples_per_second": 231.886,
+      "eval_steps_per_second": 3.623,
+      "step": 50000
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.207898846405723e-06,
+      "loss": 0.2333,
+      "step": 50001
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.207563011099357e-06,
+      "loss": 0.3772,
+      "step": 50002
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2072271757929914e-06,
+      "loss": 0.2086,
+      "step": 50003
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.206891340486626e-06,
+      "loss": 0.2384,
+      "step": 50004
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.20655550518026e-06,
+      "loss": 0.2689,
+      "step": 50005
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.206219669873894e-06,
+      "loss": 0.252,
+      "step": 50006
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2058838345675287e-06,
+      "loss": 0.1543,
+      "step": 50007
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2055479992611625e-06,
+      "loss": 0.3775,
+      "step": 50008
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2052121639547967e-06,
+      "loss": 0.2233,
+      "step": 50009
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2048763286484314e-06,
+      "loss": 0.2259,
+      "step": 50010
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2045404933420656e-06,
+      "loss": 0.3397,
+      "step": 50011
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2042046580356994e-06,
+      "loss": 0.2463,
+      "step": 50012
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2038688227293336e-06,
+      "loss": 0.2964,
+      "step": 50013
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2035329874229683e-06,
+      "loss": 0.2717,
+      "step": 50014
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.203197152116602e-06,
+      "loss": 0.2522,
+      "step": 50015
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2028613168102363e-06,
+      "loss": 0.2408,
+      "step": 50016
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.202525481503871e-06,
+      "loss": 0.2685,
+      "step": 50017
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2021896461975047e-06,
+      "loss": 0.2246,
+      "step": 50018
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.201853810891139e-06,
+      "loss": 0.3204,
+      "step": 50019
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2015179755847736e-06,
+      "loss": 0.4989,
+      "step": 50020
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.201182140278408e-06,
+      "loss": 0.3176,
+      "step": 50021
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2008463049720416e-06,
+      "loss": 0.3882,
+      "step": 50022
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2005104696656763e-06,
+      "loss": 0.2537,
+      "step": 50023
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.2001746343593105e-06,
+      "loss": 0.2785,
+      "step": 50024
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1998387990529443e-06,
+      "loss": 0.2579,
+      "step": 50025
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.199502963746579e-06,
+      "loss": 0.2374,
+      "step": 50026
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.199167128440213e-06,
+      "loss": 0.2583,
+      "step": 50027
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1988312931338474e-06,
+      "loss": 0.275,
+      "step": 50028
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.198495457827482e-06,
+      "loss": 0.476,
+      "step": 50029
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.198159622521116e-06,
+      "loss": 0.2545,
+      "step": 50030
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.19782378721475e-06,
+      "loss": 0.2294,
+      "step": 50031
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1974879519083847e-06,
+      "loss": 0.202,
+      "step": 50032
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1971521166020185e-06,
+      "loss": 0.3165,
+      "step": 50033
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1968162812956527e-06,
+      "loss": 0.1156,
+      "step": 50034
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1964804459892874e-06,
+      "loss": 0.2317,
+      "step": 50035
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.196144610682921e-06,
+      "loss": 0.2954,
+      "step": 50036
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1958087753765554e-06,
+      "loss": 0.1158,
+      "step": 50037
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.19547294007019e-06,
+      "loss": 0.2187,
+      "step": 50038
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1951371047638243e-06,
+      "loss": 0.2252,
+      "step": 50039
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.194801269457458e-06,
+      "loss": 0.2582,
+      "step": 50040
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1944654341510927e-06,
+      "loss": 0.2837,
+      "step": 50041
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.194129598844727e-06,
+      "loss": 0.2264,
+      "step": 50042
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1937937635383608e-06,
+      "loss": 0.1561,
+      "step": 50043
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1934579282319954e-06,
+      "loss": 0.2783,
+      "step": 50044
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1931220929256296e-06,
+      "loss": 0.1714,
+      "step": 50045
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.192786257619264e-06,
+      "loss": 0.3269,
+      "step": 50046
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.192450422312898e-06,
+      "loss": 0.2529,
+      "step": 50047
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1921145870065323e-06,
+      "loss": 0.2199,
+      "step": 50048
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1917787517001665e-06,
+      "loss": 0.2398,
+      "step": 50049
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.191442916393801e-06,
+      "loss": 0.4662,
+      "step": 50050
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.191107081087435e-06,
+      "loss": 0.3877,
+      "step": 50051
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.190771245781069e-06,
+      "loss": 0.2553,
+      "step": 50052
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.190435410474704e-06,
+      "loss": 0.3911,
+      "step": 50053
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1900995751683376e-06,
+      "loss": 0.2704,
+      "step": 50054
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.189763739861972e-06,
+      "loss": 0.3289,
+      "step": 50055
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1894279045556065e-06,
+      "loss": 0.2389,
+      "step": 50056
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1890920692492407e-06,
+      "loss": 0.1453,
+      "step": 50057
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1887562339428745e-06,
+      "loss": 0.307,
+      "step": 50058
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1884203986365088e-06,
+      "loss": 0.353,
+      "step": 50059
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1880845633301434e-06,
+      "loss": 0.3235,
+      "step": 50060
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.187748728023777e-06,
+      "loss": 0.2198,
+      "step": 50061
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1874128927174114e-06,
+      "loss": 0.2967,
+      "step": 50062
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.187077057411046e-06,
+      "loss": 0.3156,
+      "step": 50063
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.18674122210468e-06,
+      "loss": 0.2601,
+      "step": 50064
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.186405386798314e-06,
+      "loss": 0.1809,
+      "step": 50065
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1860695514919487e-06,
+      "loss": 0.1557,
+      "step": 50066
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.185733716185583e-06,
+      "loss": 0.3839,
+      "step": 50067
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1853978808792168e-06,
+      "loss": 0.2516,
+      "step": 50068
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1850620455728514e-06,
+      "loss": 0.2877,
+      "step": 50069
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1847262102664856e-06,
+      "loss": 0.1492,
+      "step": 50070
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1843903749601194e-06,
+      "loss": 0.2549,
+      "step": 50071
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.184054539653754e-06,
+      "loss": 0.287,
+      "step": 50072
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1837187043473883e-06,
+      "loss": 0.2525,
+      "step": 50073
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1833828690410225e-06,
+      "loss": 0.4064,
+      "step": 50074
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1830470337346568e-06,
+      "loss": 0.3676,
+      "step": 50075
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.182711198428291e-06,
+      "loss": 0.1761,
+      "step": 50076
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.182375363121925e-06,
+      "loss": 0.1226,
+      "step": 50077
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.18203952781556e-06,
+      "loss": 0.3703,
+      "step": 50078
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1817036925091937e-06,
+      "loss": 0.2778,
+      "step": 50079
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.181367857202828e-06,
+      "loss": 0.2093,
+      "step": 50080
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1810320218964625e-06,
+      "loss": 0.1506,
+      "step": 50081
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1806961865900963e-06,
+      "loss": 0.3044,
+      "step": 50082
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1803603512837306e-06,
+      "loss": 0.1829,
+      "step": 50083
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.180024515977365e-06,
+      "loss": 0.431,
+      "step": 50084
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1796886806709994e-06,
+      "loss": 0.2706,
+      "step": 50085
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1793528453646332e-06,
+      "loss": 0.1944,
+      "step": 50086
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.179017010058268e-06,
+      "loss": 0.2379,
+      "step": 50087
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.178681174751902e-06,
+      "loss": 0.3094,
+      "step": 50088
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.178345339445536e-06,
+      "loss": 0.1659,
+      "step": 50089
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1780095041391705e-06,
+      "loss": 0.1946,
+      "step": 50090
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1776736688328048e-06,
+      "loss": 0.2746,
+      "step": 50091
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1773378335264386e-06,
+      "loss": 0.3885,
+      "step": 50092
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1770019982200732e-06,
+      "loss": 0.155,
+      "step": 50093
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1766661629137074e-06,
+      "loss": 0.2039,
+      "step": 50094
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1763303276073417e-06,
+      "loss": 0.2866,
+      "step": 50095
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1759944923009763e-06,
+      "loss": 0.2766,
+      "step": 50096
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.17565865699461e-06,
+      "loss": 0.3962,
+      "step": 50097
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1753228216882443e-06,
+      "loss": 0.168,
+      "step": 50098
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.174986986381879e-06,
+      "loss": 0.171,
+      "step": 50099
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1746511510755128e-06,
+      "loss": 0.1859,
+      "step": 50100
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.174315315769147e-06,
+      "loss": 0.237,
+      "step": 50101
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1739794804627812e-06,
+      "loss": 0.3224,
+      "step": 50102
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1736436451564155e-06,
+      "loss": 0.1976,
+      "step": 50103
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1733078098500497e-06,
+      "loss": 0.3171,
+      "step": 50104
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.172971974543684e-06,
+      "loss": 0.3463,
+      "step": 50105
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1726361392373186e-06,
+      "loss": 0.4722,
+      "step": 50106
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1723003039309524e-06,
+      "loss": 0.2459,
+      "step": 50107
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1719644686245866e-06,
+      "loss": 0.2354,
+      "step": 50108
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1716286333182212e-06,
+      "loss": 0.3266,
+      "step": 50109
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.171292798011855e-06,
+      "loss": 0.2143,
+      "step": 50110
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1709569627054892e-06,
+      "loss": 0.1781,
+      "step": 50111
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.170621127399124e-06,
+      "loss": 0.2157,
+      "step": 50112
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.170285292092758e-06,
+      "loss": 0.2667,
+      "step": 50113
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.169949456786392e-06,
+      "loss": 0.2633,
+      "step": 50114
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1696136214800266e-06,
+      "loss": 0.2937,
+      "step": 50115
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.169277786173661e-06,
+      "loss": 0.4186,
+      "step": 50116
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1689419508672946e-06,
+      "loss": 0.18,
+      "step": 50117
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1686061155609292e-06,
+      "loss": 0.2526,
+      "step": 50118
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1682702802545635e-06,
+      "loss": 0.1601,
+      "step": 50119
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1679344449481973e-06,
+      "loss": 0.1811,
+      "step": 50120
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.167598609641832e-06,
+      "loss": 0.2517,
+      "step": 50121
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.167262774335466e-06,
+      "loss": 0.2271,
+      "step": 50122
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 3.1669269390291004e-06,
+      "loss": 0.1073,
+      "step": 50123
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.166591103722735e-06,
+      "loss": 0.1659,
+      "step": 50124
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.166255268416369e-06,
+      "loss": 0.2609,
+      "step": 50125
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.165919433110003e-06,
+      "loss": 0.1191,
+      "step": 50126
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1655835978036377e-06,
+      "loss": 0.2776,
+      "step": 50127
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1652477624972715e-06,
+      "loss": 0.4198,
+      "step": 50128
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1649119271909057e-06,
+      "loss": 0.2888,
+      "step": 50129
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1645760918845403e-06,
+      "loss": 0.3448,
+      "step": 50130
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.164240256578174e-06,
+      "loss": 0.327,
+      "step": 50131
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1639044212718084e-06,
+      "loss": 0.2531,
+      "step": 50132
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.163568585965443e-06,
+      "loss": 0.2515,
+      "step": 50133
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1632327506590772e-06,
+      "loss": 0.225,
+      "step": 50134
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.162896915352711e-06,
+      "loss": 0.1355,
+      "step": 50135
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1625610800463457e-06,
+      "loss": 0.3056,
+      "step": 50136
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.16222524473998e-06,
+      "loss": 0.299,
+      "step": 50137
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1618894094336137e-06,
+      "loss": 0.3494,
+      "step": 50138
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1615535741272484e-06,
+      "loss": 0.2388,
+      "step": 50139
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1612177388208826e-06,
+      "loss": 0.1799,
+      "step": 50140
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.160881903514517e-06,
+      "loss": 0.3356,
+      "step": 50141
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.160546068208151e-06,
+      "loss": 0.3345,
+      "step": 50142
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1602102329017853e-06,
+      "loss": 0.3521,
+      "step": 50143
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1598743975954195e-06,
+      "loss": 0.1494,
+      "step": 50144
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.159538562289054e-06,
+      "loss": 0.2373,
+      "step": 50145
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.159202726982688e-06,
+      "loss": 0.2204,
+      "step": 50146
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.158866891676322e-06,
+      "loss": 0.3589,
+      "step": 50147
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.158531056369956e-06,
+      "loss": 0.1487,
+      "step": 50148
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1581952210635906e-06,
+      "loss": 0.1795,
+      "step": 50149
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.157859385757225e-06,
+      "loss": 0.3559,
+      "step": 50150
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.157523550450859e-06,
+      "loss": 0.1013,
+      "step": 50151
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1571877151444937e-06,
+      "loss": 0.2677,
+      "step": 50152
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1568518798381275e-06,
+      "loss": 0.3194,
+      "step": 50153
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1565160445317617e-06,
+      "loss": 0.3633,
+      "step": 50154
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1561802092253964e-06,
+      "loss": 0.1892,
+      "step": 50155
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.15584437391903e-06,
+      "loss": 0.4217,
+      "step": 50156
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1555085386126644e-06,
+      "loss": 0.2815,
+      "step": 50157
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.155172703306299e-06,
+      "loss": 0.2455,
+      "step": 50158
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.154836867999933e-06,
+      "loss": 0.2016,
+      "step": 50159
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.154501032693567e-06,
+      "loss": 0.2162,
+      "step": 50160
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1541651973872017e-06,
+      "loss": 0.2822,
+      "step": 50161
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.153829362080836e-06,
+      "loss": 0.322,
+      "step": 50162
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1534935267744697e-06,
+      "loss": 0.2757,
+      "step": 50163
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1531576914681044e-06,
+      "loss": 0.3307,
+      "step": 50164
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1528218561617386e-06,
+      "loss": 0.2565,
+      "step": 50165
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1524860208553724e-06,
+      "loss": 0.2884,
+      "step": 50166
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.152150185549007e-06,
+      "loss": 0.2446,
+      "step": 50167
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1518143502426413e-06,
+      "loss": 0.4265,
+      "step": 50168
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1514785149362755e-06,
+      "loss": 0.3242,
+      "step": 50169
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.15114267962991e-06,
+      "loss": 0.356,
+      "step": 50170
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.150806844323544e-06,
+      "loss": 0.2239,
+      "step": 50171
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.150471009017178e-06,
+      "loss": 0.1825,
+      "step": 50172
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.150135173710813e-06,
+      "loss": 0.2669,
+      "step": 50173
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1497993384044466e-06,
+      "loss": 0.2168,
+      "step": 50174
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.149463503098081e-06,
+      "loss": 0.4308,
+      "step": 50175
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1491276677917155e-06,
+      "loss": 0.2563,
+      "step": 50176
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1487918324853493e-06,
+      "loss": 0.3579,
+      "step": 50177
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1484559971789835e-06,
+      "loss": 0.2303,
+      "step": 50178
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.148120161872618e-06,
+      "loss": 0.3437,
+      "step": 50179
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1477843265662524e-06,
+      "loss": 0.3567,
+      "step": 50180
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.147448491259886e-06,
+      "loss": 0.422,
+      "step": 50181
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.147112655953521e-06,
+      "loss": 0.1859,
+      "step": 50182
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.146776820647155e-06,
+      "loss": 0.2838,
+      "step": 50183
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.146440985340789e-06,
+      "loss": 0.3033,
+      "step": 50184
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1461051500344235e-06,
+      "loss": 0.2041,
+      "step": 50185
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1457693147280577e-06,
+      "loss": 0.2563,
+      "step": 50186
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.145433479421692e-06,
+      "loss": 0.1956,
+      "step": 50187
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.145097644115326e-06,
+      "loss": 0.3056,
+      "step": 50188
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1447618088089604e-06,
+      "loss": 0.135,
+      "step": 50189
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1444259735025946e-06,
+      "loss": 0.2373,
+      "step": 50190
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1440901381962293e-06,
+      "loss": 0.2279,
+      "step": 50191
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.143754302889863e-06,
+      "loss": 0.4162,
+      "step": 50192
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1434184675834973e-06,
+      "loss": 0.2689,
+      "step": 50193
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.143082632277131e-06,
+      "loss": 0.1476,
+      "step": 50194
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1427467969707657e-06,
+      "loss": 0.4286,
+      "step": 50195
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1424109616644e-06,
+      "loss": 0.292,
+      "step": 50196
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.142075126358034e-06,
+      "loss": 0.2471,
+      "step": 50197
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.141739291051669e-06,
+      "loss": 0.3392,
+      "step": 50198
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1414034557453026e-06,
+      "loss": 0.2681,
+      "step": 50199
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.141067620438937e-06,
+      "loss": 0.261,
+      "step": 50200
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1407317851325715e-06,
+      "loss": 0.3021,
+      "step": 50201
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1403959498262053e-06,
+      "loss": 0.3384,
+      "step": 50202
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1400601145198395e-06,
+      "loss": 0.2025,
+      "step": 50203
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.139724279213474e-06,
+      "loss": 0.2321,
+      "step": 50204
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.139388443907108e-06,
+      "loss": 0.2853,
+      "step": 50205
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.139052608600742e-06,
+      "loss": 0.5031,
+      "step": 50206
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.138716773294377e-06,
+      "loss": 0.1207,
+      "step": 50207
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.138380937988011e-06,
+      "loss": 0.226,
+      "step": 50208
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.138045102681645e-06,
+      "loss": 0.3975,
+      "step": 50209
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1377092673752795e-06,
+      "loss": 0.2606,
+      "step": 50210
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1373734320689138e-06,
+      "loss": 0.3295,
+      "step": 50211
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1370375967625476e-06,
+      "loss": 0.306,
+      "step": 50212
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.136701761456182e-06,
+      "loss": 0.2277,
+      "step": 50213
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1363659261498164e-06,
+      "loss": 0.4993,
+      "step": 50214
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1360300908434506e-06,
+      "loss": 0.2586,
+      "step": 50215
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.135694255537085e-06,
+      "loss": 0.3179,
+      "step": 50216
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.135358420230719e-06,
+      "loss": 0.2293,
+      "step": 50217
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1350225849243533e-06,
+      "loss": 0.2642,
+      "step": 50218
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.134686749617988e-06,
+      "loss": 0.2933,
+      "step": 50219
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1343509143116218e-06,
+      "loss": 0.3801,
+      "step": 50220
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.134015079005256e-06,
+      "loss": 0.292,
+      "step": 50221
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1336792436988906e-06,
+      "loss": 0.2368,
+      "step": 50222
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1333434083925244e-06,
+      "loss": 0.2828,
+      "step": 50223
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1330075730861587e-06,
+      "loss": 0.2348,
+      "step": 50224
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1326717377797933e-06,
+      "loss": 0.2377,
+      "step": 50225
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1323359024734275e-06,
+      "loss": 0.3603,
+      "step": 50226
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1320000671670613e-06,
+      "loss": 0.4641,
+      "step": 50227
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.131664231860696e-06,
+      "loss": 0.4595,
+      "step": 50228
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.13132839655433e-06,
+      "loss": 0.4453,
+      "step": 50229
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.130992561247964e-06,
+      "loss": 0.272,
+      "step": 50230
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1306567259415987e-06,
+      "loss": 0.2769,
+      "step": 50231
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.130320890635233e-06,
+      "loss": 0.2623,
+      "step": 50232
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1299850553288667e-06,
+      "loss": 0.1834,
+      "step": 50233
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1296492200225013e-06,
+      "loss": 0.1394,
+      "step": 50234
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1293133847161355e-06,
+      "loss": 0.3234,
+      "step": 50235
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1289775494097698e-06,
+      "loss": 0.2851,
+      "step": 50236
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1286417141034036e-06,
+      "loss": 0.1315,
+      "step": 50237
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1283058787970382e-06,
+      "loss": 0.1781,
+      "step": 50238
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1279700434906724e-06,
+      "loss": 0.2292,
+      "step": 50239
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1276342081843062e-06,
+      "loss": 0.1714,
+      "step": 50240
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.127298372877941e-06,
+      "loss": 0.2181,
+      "step": 50241
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.126962537571575e-06,
+      "loss": 0.2797,
+      "step": 50242
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1266267022652093e-06,
+      "loss": 0.3528,
+      "step": 50243
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1262908669588436e-06,
+      "loss": 0.3233,
+      "step": 50244
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1259550316524778e-06,
+      "loss": 0.2654,
+      "step": 50245
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.125619196346112e-06,
+      "loss": 0.2517,
+      "step": 50246
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1252833610397467e-06,
+      "loss": 0.1695,
+      "step": 50247
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1249475257333805e-06,
+      "loss": 0.2289,
+      "step": 50248
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1246116904270147e-06,
+      "loss": 0.4107,
+      "step": 50249
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1242758551206493e-06,
+      "loss": 0.2455,
+      "step": 50250
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.123940019814283e-06,
+      "loss": 0.3421,
+      "step": 50251
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1236041845079174e-06,
+      "loss": 0.1664,
+      "step": 50252
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.123268349201552e-06,
+      "loss": 0.1838,
+      "step": 50253
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1229325138951862e-06,
+      "loss": 0.1689,
+      "step": 50254
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.12259667858882e-06,
+      "loss": 0.2831,
+      "step": 50255
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1222608432824547e-06,
+      "loss": 0.2135,
+      "step": 50256
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.121925007976089e-06,
+      "loss": 0.2644,
+      "step": 50257
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1215891726697227e-06,
+      "loss": 0.2636,
+      "step": 50258
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1212533373633573e-06,
+      "loss": 0.2715,
+      "step": 50259
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1209175020569916e-06,
+      "loss": 0.3134,
+      "step": 50260
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1205816667506254e-06,
+      "loss": 0.2319,
+      "step": 50261
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.12024583144426e-06,
+      "loss": 0.2015,
+      "step": 50262
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1199099961378942e-06,
+      "loss": 0.2767,
+      "step": 50263
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1195741608315285e-06,
+      "loss": 0.2847,
+      "step": 50264
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.119238325525163e-06,
+      "loss": 0.1844,
+      "step": 50265
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.118902490218797e-06,
+      "loss": 0.4556,
+      "step": 50266
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.118566654912431e-06,
+      "loss": 0.3277,
+      "step": 50267
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1182308196060658e-06,
+      "loss": 0.2714,
+      "step": 50268
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1178949842996996e-06,
+      "loss": 0.1747,
+      "step": 50269
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.117559148993334e-06,
+      "loss": 0.1768,
+      "step": 50270
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1172233136869685e-06,
+      "loss": 0.3069,
+      "step": 50271
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1168874783806023e-06,
+      "loss": 0.2332,
+      "step": 50272
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1165516430742365e-06,
+      "loss": 0.3689,
+      "step": 50273
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.116215807767871e-06,
+      "loss": 0.2688,
+      "step": 50274
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1158799724615053e-06,
+      "loss": 0.2272,
+      "step": 50275
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.115544137155139e-06,
+      "loss": 0.1474,
+      "step": 50276
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.115208301848774e-06,
+      "loss": 0.374,
+      "step": 50277
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.114872466542408e-06,
+      "loss": 0.2227,
+      "step": 50278
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.114536631236042e-06,
+      "loss": 0.3722,
+      "step": 50279
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1142007959296765e-06,
+      "loss": 0.1916,
+      "step": 50280
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1138649606233107e-06,
+      "loss": 0.1385,
+      "step": 50281
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.113529125316945e-06,
+      "loss": 0.3186,
+      "step": 50282
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1131932900105787e-06,
+      "loss": 0.2152,
+      "step": 50283
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1128574547042134e-06,
+      "loss": 0.2601,
+      "step": 50284
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1125216193978476e-06,
+      "loss": 0.4795,
+      "step": 50285
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1121857840914814e-06,
+      "loss": 0.29,
+      "step": 50286
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.111849948785116e-06,
+      "loss": 0.2955,
+      "step": 50287
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1115141134787503e-06,
+      "loss": 0.1877,
+      "step": 50288
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.111178278172384e-06,
+      "loss": 0.2365,
+      "step": 50289
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1108424428660187e-06,
+      "loss": 0.3535,
+      "step": 50290
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.110506607559653e-06,
+      "loss": 0.2358,
+      "step": 50291
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.110170772253287e-06,
+      "loss": 0.4646,
+      "step": 50292
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.109834936946922e-06,
+      "loss": 0.2156,
+      "step": 50293
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1094991016405556e-06,
+      "loss": 0.2996,
+      "step": 50294
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.10916326633419e-06,
+      "loss": 0.3402,
+      "step": 50295
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1088274310278245e-06,
+      "loss": 0.3672,
+      "step": 50296
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1084915957214583e-06,
+      "loss": 0.1345,
+      "step": 50297
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1081557604150925e-06,
+      "loss": 0.1731,
+      "step": 50298
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.107819925108727e-06,
+      "loss": 0.2149,
+      "step": 50299
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1074840898023614e-06,
+      "loss": 0.1816,
+      "step": 50300
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.107148254495995e-06,
+      "loss": 0.3748,
+      "step": 50301
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.10681241918963e-06,
+      "loss": 0.2889,
+      "step": 50302
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.106476583883264e-06,
+      "loss": 0.1961,
+      "step": 50303
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.106140748576898e-06,
+      "loss": 0.306,
+      "step": 50304
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1058049132705325e-06,
+      "loss": 0.1587,
+      "step": 50305
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1054690779641667e-06,
+      "loss": 0.3418,
+      "step": 50306
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1051332426578005e-06,
+      "loss": 0.2137,
+      "step": 50307
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.104797407351435e-06,
+      "loss": 0.3744,
+      "step": 50308
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1044615720450694e-06,
+      "loss": 0.3578,
+      "step": 50309
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1041257367387036e-06,
+      "loss": 0.4466,
+      "step": 50310
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1037899014323383e-06,
+      "loss": 0.2814,
+      "step": 50311
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.103454066125972e-06,
+      "loss": 0.3208,
+      "step": 50312
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1031182308196063e-06,
+      "loss": 0.16,
+      "step": 50313
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.102782395513241e-06,
+      "loss": 0.124,
+      "step": 50314
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1024465602068747e-06,
+      "loss": 0.1936,
+      "step": 50315
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.102110724900509e-06,
+      "loss": 0.3234,
+      "step": 50316
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1017748895941436e-06,
+      "loss": 0.2911,
+      "step": 50317
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1014390542877774e-06,
+      "loss": 0.3093,
+      "step": 50318
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1011032189814116e-06,
+      "loss": 0.2423,
+      "step": 50319
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1007673836750463e-06,
+      "loss": 0.4153,
+      "step": 50320
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1004315483686805e-06,
+      "loss": 0.2448,
+      "step": 50321
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1000957130623143e-06,
+      "loss": 0.1765,
+      "step": 50322
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.099759877755949e-06,
+      "loss": 0.1703,
+      "step": 50323
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.099424042449583e-06,
+      "loss": 0.266,
+      "step": 50324
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.099088207143217e-06,
+      "loss": 0.2413,
+      "step": 50325
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0987523718368516e-06,
+      "loss": 0.2866,
+      "step": 50326
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.098416536530486e-06,
+      "loss": 0.2369,
+      "step": 50327
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.09808070122412e-06,
+      "loss": 0.2394,
+      "step": 50328
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.097744865917754e-06,
+      "loss": 0.2033,
+      "step": 50329
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0974090306113885e-06,
+      "loss": 0.2267,
+      "step": 50330
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0970731953050227e-06,
+      "loss": 0.1327,
+      "step": 50331
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0967373599986565e-06,
+      "loss": 0.2267,
+      "step": 50332
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.096401524692291e-06,
+      "loss": 0.2811,
+      "step": 50333
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0960656893859254e-06,
+      "loss": 0.3374,
+      "step": 50334
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.095729854079559e-06,
+      "loss": 0.2597,
+      "step": 50335
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.095394018773194e-06,
+      "loss": 0.2009,
+      "step": 50336
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.095058183466828e-06,
+      "loss": 0.2192,
+      "step": 50337
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0947223481604623e-06,
+      "loss": 0.2044,
+      "step": 50338
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.094386512854097e-06,
+      "loss": 0.148,
+      "step": 50339
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0940506775477307e-06,
+      "loss": 0.2924,
+      "step": 50340
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.093714842241365e-06,
+      "loss": 0.1697,
+      "step": 50341
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0933790069349996e-06,
+      "loss": 0.2647,
+      "step": 50342
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0930431716286334e-06,
+      "loss": 0.3048,
+      "step": 50343
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0927073363222676e-06,
+      "loss": 0.4302,
+      "step": 50344
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0923715010159023e-06,
+      "loss": 0.176,
+      "step": 50345
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.092035665709536e-06,
+      "loss": 0.3161,
+      "step": 50346
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0916998304031703e-06,
+      "loss": 0.2855,
+      "step": 50347
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.091363995096805e-06,
+      "loss": 0.4397,
+      "step": 50348
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.091028159790439e-06,
+      "loss": 0.1538,
+      "step": 50349
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.090692324484073e-06,
+      "loss": 0.2929,
+      "step": 50350
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0903564891777076e-06,
+      "loss": 0.2373,
+      "step": 50351
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.090020653871342e-06,
+      "loss": 0.1582,
+      "step": 50352
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0896848185649757e-06,
+      "loss": 0.1733,
+      "step": 50353
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0893489832586103e-06,
+      "loss": 0.1563,
+      "step": 50354
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0890131479522445e-06,
+      "loss": 0.3836,
+      "step": 50355
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0886773126458788e-06,
+      "loss": 0.222,
+      "step": 50356
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.088341477339513e-06,
+      "loss": 0.2975,
+      "step": 50357
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.088005642033147e-06,
+      "loss": 0.2251,
+      "step": 50358
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0876698067267814e-06,
+      "loss": 0.2958,
+      "step": 50359
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.087333971420416e-06,
+      "loss": 0.1907,
+      "step": 50360
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.08699813611405e-06,
+      "loss": 0.3708,
+      "step": 50361
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.086662300807684e-06,
+      "loss": 0.1431,
+      "step": 50362
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0863264655013187e-06,
+      "loss": 0.4795,
+      "step": 50363
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0859906301949525e-06,
+      "loss": 0.4534,
+      "step": 50364
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0856547948885868e-06,
+      "loss": 0.2751,
+      "step": 50365
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0853189595822214e-06,
+      "loss": 0.2265,
+      "step": 50366
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0849831242758556e-06,
+      "loss": 0.2873,
+      "step": 50367
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0846472889694894e-06,
+      "loss": 0.2319,
+      "step": 50368
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.084311453663124e-06,
+      "loss": 0.3481,
+      "step": 50369
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0839756183567583e-06,
+      "loss": 0.2907,
+      "step": 50370
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.083639783050392e-06,
+      "loss": 0.1681,
+      "step": 50371
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0833039477440263e-06,
+      "loss": 0.2506,
+      "step": 50372
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.082968112437661e-06,
+      "loss": 0.3503,
+      "step": 50373
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0826322771312948e-06,
+      "loss": 0.1682,
+      "step": 50374
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.082296441824929e-06,
+      "loss": 0.2769,
+      "step": 50375
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0819606065185637e-06,
+      "loss": 0.1603,
+      "step": 50376
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.081624771212198e-06,
+      "loss": 0.344,
+      "step": 50377
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0812889359058317e-06,
+      "loss": 0.3036,
+      "step": 50378
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0809531005994663e-06,
+      "loss": 0.4194,
+      "step": 50379
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0806172652931005e-06,
+      "loss": 0.1932,
+      "step": 50380
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0802814299867343e-06,
+      "loss": 0.324,
+      "step": 50381
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.079945594680369e-06,
+      "loss": 0.1346,
+      "step": 50382
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0796097593740032e-06,
+      "loss": 0.2801,
+      "step": 50383
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0792739240676374e-06,
+      "loss": 0.267,
+      "step": 50384
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0789380887612717e-06,
+      "loss": 0.3785,
+      "step": 50385
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.078602253454906e-06,
+      "loss": 0.3306,
+      "step": 50386
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.07826641814854e-06,
+      "loss": 0.318,
+      "step": 50387
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0779305828421748e-06,
+      "loss": 0.1654,
+      "step": 50388
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0775947475358086e-06,
+      "loss": 0.2575,
+      "step": 50389
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0772589122294428e-06,
+      "loss": 0.3273,
+      "step": 50390
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0769230769230774e-06,
+      "loss": 0.2792,
+      "step": 50391
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0765872416167112e-06,
+      "loss": 0.2663,
+      "step": 50392
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0762514063103455e-06,
+      "loss": 0.2888,
+      "step": 50393
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.07591557100398e-06,
+      "loss": 0.2384,
+      "step": 50394
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0755797356976143e-06,
+      "loss": 0.4123,
+      "step": 50395
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.075243900391248e-06,
+      "loss": 0.3647,
+      "step": 50396
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0749080650848828e-06,
+      "loss": 0.4046,
+      "step": 50397
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.074572229778517e-06,
+      "loss": 0.2618,
+      "step": 50398
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.074236394472151e-06,
+      "loss": 0.2383,
+      "step": 50399
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0739005591657855e-06,
+      "loss": 0.2632,
+      "step": 50400
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0735647238594197e-06,
+      "loss": 0.1166,
+      "step": 50401
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0732288885530535e-06,
+      "loss": 0.3864,
+      "step": 50402
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.072893053246688e-06,
+      "loss": 0.4773,
+      "step": 50403
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0725572179403223e-06,
+      "loss": 0.1635,
+      "step": 50404
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0722213826339566e-06,
+      "loss": 0.2831,
+      "step": 50405
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0718855473275912e-06,
+      "loss": 0.184,
+      "step": 50406
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.071549712021225e-06,
+      "loss": 0.2898,
+      "step": 50407
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0712138767148592e-06,
+      "loss": 0.3404,
+      "step": 50408
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.070878041408494e-06,
+      "loss": 0.2101,
+      "step": 50409
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0705422061021277e-06,
+      "loss": 0.1794,
+      "step": 50410
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.070206370795762e-06,
+      "loss": 0.2297,
+      "step": 50411
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0698705354893966e-06,
+      "loss": 0.2491,
+      "step": 50412
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0695347001830304e-06,
+      "loss": 0.2882,
+      "step": 50413
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0691988648766646e-06,
+      "loss": 0.2909,
+      "step": 50414
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0688630295702992e-06,
+      "loss": 0.2365,
+      "step": 50415
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0685271942639335e-06,
+      "loss": 0.1939,
+      "step": 50416
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0681913589575673e-06,
+      "loss": 0.2794,
+      "step": 50417
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0678555236512015e-06,
+      "loss": 0.2923,
+      "step": 50418
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.067519688344836e-06,
+      "loss": 0.255,
+      "step": 50419
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.06718385303847e-06,
+      "loss": 0.3278,
+      "step": 50420
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.066848017732104e-06,
+      "loss": 0.2562,
+      "step": 50421
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.066512182425739e-06,
+      "loss": 0.2517,
+      "step": 50422
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.066176347119373e-06,
+      "loss": 0.3253,
+      "step": 50423
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.065840511813007e-06,
+      "loss": 0.3289,
+      "step": 50424
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0655046765066415e-06,
+      "loss": 0.2807,
+      "step": 50425
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0651688412002757e-06,
+      "loss": 0.3154,
+      "step": 50426
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0648330058939095e-06,
+      "loss": 0.3419,
+      "step": 50427
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.064497170587544e-06,
+      "loss": 0.2803,
+      "step": 50428
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0641613352811784e-06,
+      "loss": 0.3147,
+      "step": 50429
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0638254999748126e-06,
+      "loss": 0.2137,
+      "step": 50430
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.063489664668447e-06,
+      "loss": 0.3187,
+      "step": 50431
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.063153829362081e-06,
+      "loss": 0.1349,
+      "step": 50432
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0628179940557153e-06,
+      "loss": 0.2606,
+      "step": 50433
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.06248215874935e-06,
+      "loss": 0.2713,
+      "step": 50434
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0621463234429837e-06,
+      "loss": 0.2812,
+      "step": 50435
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.061810488136618e-06,
+      "loss": 0.2582,
+      "step": 50436
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0614746528302526e-06,
+      "loss": 0.1719,
+      "step": 50437
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0611388175238864e-06,
+      "loss": 0.2511,
+      "step": 50438
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0608029822175206e-06,
+      "loss": 0.2251,
+      "step": 50439
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0604671469111553e-06,
+      "loss": 0.4171,
+      "step": 50440
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0601313116047895e-06,
+      "loss": 0.2327,
+      "step": 50441
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0597954762984233e-06,
+      "loss": 0.2981,
+      "step": 50442
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.059459640992058e-06,
+      "loss": 0.4578,
+      "step": 50443
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.059123805685692e-06,
+      "loss": 0.2537,
+      "step": 50444
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.058787970379326e-06,
+      "loss": 0.1924,
+      "step": 50445
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0584521350729606e-06,
+      "loss": 0.2244,
+      "step": 50446
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.058116299766595e-06,
+      "loss": 0.3751,
+      "step": 50447
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0577804644602286e-06,
+      "loss": 0.5191,
+      "step": 50448
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0574446291538633e-06,
+      "loss": 0.1423,
+      "step": 50449
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0571087938474975e-06,
+      "loss": 0.312,
+      "step": 50450
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0567729585411317e-06,
+      "loss": 0.4361,
+      "step": 50451
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0564371232347664e-06,
+      "loss": 0.4205,
+      "step": 50452
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0561012879284e-06,
+      "loss": 0.3526,
+      "step": 50453
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0557654526220344e-06,
+      "loss": 0.3421,
+      "step": 50454
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.055429617315669e-06,
+      "loss": 0.1973,
+      "step": 50455
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.055093782009303e-06,
+      "loss": 0.2465,
+      "step": 50456
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.054757946702937e-06,
+      "loss": 0.2949,
+      "step": 50457
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0544221113965717e-06,
+      "loss": 0.2343,
+      "step": 50458
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0540862760902055e-06,
+      "loss": 0.2251,
+      "step": 50459
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0537504407838397e-06,
+      "loss": 0.2353,
+      "step": 50460
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0534146054774744e-06,
+      "loss": 0.2527,
+      "step": 50461
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0530787701711086e-06,
+      "loss": 0.2403,
+      "step": 50462
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0527429348647424e-06,
+      "loss": 0.2814,
+      "step": 50463
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0524070995583766e-06,
+      "loss": 0.1206,
+      "step": 50464
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0520712642520113e-06,
+      "loss": 0.2188,
+      "step": 50465
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.051735428945645e-06,
+      "loss": 0.2414,
+      "step": 50466
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0513995936392793e-06,
+      "loss": 0.3084,
+      "step": 50467
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.051063758332914e-06,
+      "loss": 0.2628,
+      "step": 50468
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.050727923026548e-06,
+      "loss": 0.2404,
+      "step": 50469
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.050392087720182e-06,
+      "loss": 0.2305,
+      "step": 50470
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0500562524138166e-06,
+      "loss": 0.3152,
+      "step": 50471
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.049720417107451e-06,
+      "loss": 0.2227,
+      "step": 50472
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0493845818010846e-06,
+      "loss": 0.2815,
+      "step": 50473
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0490487464947193e-06,
+      "loss": 0.1655,
+      "step": 50474
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0487129111883535e-06,
+      "loss": 0.2375,
+      "step": 50475
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0483770758819873e-06,
+      "loss": 0.1939,
+      "step": 50476
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.048041240575622e-06,
+      "loss": 0.31,
+      "step": 50477
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.047705405269256e-06,
+      "loss": 0.1189,
+      "step": 50478
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0473695699628904e-06,
+      "loss": 0.093,
+      "step": 50479
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.047033734656525e-06,
+      "loss": 0.2299,
+      "step": 50480
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.046697899350159e-06,
+      "loss": 0.2599,
+      "step": 50481
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.046362064043793e-06,
+      "loss": 0.3534,
+      "step": 50482
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0460262287374277e-06,
+      "loss": 0.1838,
+      "step": 50483
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0456903934310615e-06,
+      "loss": 0.1681,
+      "step": 50484
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0453545581246957e-06,
+      "loss": 0.3961,
+      "step": 50485
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0450187228183304e-06,
+      "loss": 0.2512,
+      "step": 50486
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.044682887511964e-06,
+      "loss": 0.3518,
+      "step": 50487
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0443470522055984e-06,
+      "loss": 0.2635,
+      "step": 50488
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.044011216899233e-06,
+      "loss": 0.2779,
+      "step": 50489
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0436753815928673e-06,
+      "loss": 0.1711,
+      "step": 50490
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.043339546286501e-06,
+      "loss": 0.3128,
+      "step": 50491
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0430037109801357e-06,
+      "loss": 0.2138,
+      "step": 50492
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.04266787567377e-06,
+      "loss": 0.1968,
+      "step": 50493
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0423320403674038e-06,
+      "loss": 0.2997,
+      "step": 50494
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0419962050610384e-06,
+      "loss": 0.175,
+      "step": 50495
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0416603697546726e-06,
+      "loss": 0.4382,
+      "step": 50496
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.041324534448307e-06,
+      "loss": 0.3508,
+      "step": 50497
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.040988699141941e-06,
+      "loss": 0.2015,
+      "step": 50498
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0406528638355753e-06,
+      "loss": 0.2911,
+      "step": 50499
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0403170285292095e-06,
+      "loss": 0.209,
+      "step": 50500
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.039981193222844e-06,
+      "loss": 0.2203,
+      "step": 50501
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.039645357916478e-06,
+      "loss": 0.2633,
+      "step": 50502
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.039309522610112e-06,
+      "loss": 0.3065,
+      "step": 50503
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.038973687303747e-06,
+      "loss": 0.2322,
+      "step": 50504
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0386378519973807e-06,
+      "loss": 0.2948,
+      "step": 50505
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.038302016691015e-06,
+      "loss": 0.4252,
+      "step": 50506
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0379661813846495e-06,
+      "loss": 0.2456,
+      "step": 50507
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0376303460782837e-06,
+      "loss": 0.3803,
+      "step": 50508
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0372945107719175e-06,
+      "loss": 0.3207,
+      "step": 50509
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0369586754655518e-06,
+      "loss": 0.2124,
+      "step": 50510
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0366228401591864e-06,
+      "loss": 0.161,
+      "step": 50511
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0362870048528202e-06,
+      "loss": 0.321,
+      "step": 50512
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0359511695464544e-06,
+      "loss": 0.2531,
+      "step": 50513
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.035615334240089e-06,
+      "loss": 0.1966,
+      "step": 50514
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.035279498933723e-06,
+      "loss": 0.1749,
+      "step": 50515
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.034943663627357e-06,
+      "loss": 0.2011,
+      "step": 50516
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0346078283209918e-06,
+      "loss": 0.2372,
+      "step": 50517
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.034271993014626e-06,
+      "loss": 0.1577,
+      "step": 50518
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0339361577082598e-06,
+      "loss": 0.2443,
+      "step": 50519
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 3.0336003224018944e-06,
+      "loss": 0.2493,
+      "step": 50520
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0332644870955287e-06,
+      "loss": 0.3057,
+      "step": 50521
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0329286517891625e-06,
+      "loss": 0.2981,
+      "step": 50522
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.032592816482797e-06,
+      "loss": 0.1908,
+      "step": 50523
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0322569811764313e-06,
+      "loss": 0.2731,
+      "step": 50524
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0319211458700656e-06,
+      "loss": 0.2444,
+      "step": 50525
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0315853105636998e-06,
+      "loss": 0.1867,
+      "step": 50526
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.031249475257334e-06,
+      "loss": 0.3049,
+      "step": 50527
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0309136399509682e-06,
+      "loss": 0.2414,
+      "step": 50528
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.030577804644603e-06,
+      "loss": 0.3454,
+      "step": 50529
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0302419693382367e-06,
+      "loss": 0.2884,
+      "step": 50530
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.029906134031871e-06,
+      "loss": 0.3052,
+      "step": 50531
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0295702987255055e-06,
+      "loss": 0.3634,
+      "step": 50532
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0292344634191393e-06,
+      "loss": 0.1804,
+      "step": 50533
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0288986281127736e-06,
+      "loss": 0.1107,
+      "step": 50534
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0285627928064082e-06,
+      "loss": 0.104,
+      "step": 50535
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0282269575000424e-06,
+      "loss": 0.2702,
+      "step": 50536
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0278911221936762e-06,
+      "loss": 0.2095,
+      "step": 50537
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.027555286887311e-06,
+      "loss": 0.3622,
+      "step": 50538
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.027219451580945e-06,
+      "loss": 0.3636,
+      "step": 50539
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.026883616274579e-06,
+      "loss": 0.3071,
+      "step": 50540
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0265477809682136e-06,
+      "loss": 0.2891,
+      "step": 50541
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0262119456618478e-06,
+      "loss": 0.3314,
+      "step": 50542
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0258761103554816e-06,
+      "loss": 0.4219,
+      "step": 50543
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0255402750491162e-06,
+      "loss": 0.2508,
+      "step": 50544
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0252044397427505e-06,
+      "loss": 0.1828,
+      "step": 50545
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0248686044363847e-06,
+      "loss": 0.2834,
+      "step": 50546
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0245327691300193e-06,
+      "loss": 0.344,
+      "step": 50547
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.024196933823653e-06,
+      "loss": 0.329,
+      "step": 50548
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0238610985172873e-06,
+      "loss": 0.2841,
+      "step": 50549
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.023525263210922e-06,
+      "loss": 0.2537,
+      "step": 50550
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.023189427904556e-06,
+      "loss": 0.213,
+      "step": 50551
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.02285359259819e-06,
+      "loss": 0.2353,
+      "step": 50552
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0225177572918242e-06,
+      "loss": 0.197,
+      "step": 50553
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0221819219854585e-06,
+      "loss": 0.151,
+      "step": 50554
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0218460866790927e-06,
+      "loss": 0.2969,
+      "step": 50555
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.021510251372727e-06,
+      "loss": 0.2436,
+      "step": 50556
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0211744160663616e-06,
+      "loss": 0.3028,
+      "step": 50557
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0208385807599954e-06,
+      "loss": 0.1446,
+      "step": 50558
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0205027454536296e-06,
+      "loss": 0.2574,
+      "step": 50559
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0201669101472642e-06,
+      "loss": 0.28,
+      "step": 50560
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.019831074840898e-06,
+      "loss": 0.2778,
+      "step": 50561
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0194952395345323e-06,
+      "loss": 0.1854,
+      "step": 50562
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.019159404228167e-06,
+      "loss": 0.4152,
+      "step": 50563
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.018823568921801e-06,
+      "loss": 0.2924,
+      "step": 50564
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.018487733615435e-06,
+      "loss": 0.3139,
+      "step": 50565
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0181518983090696e-06,
+      "loss": 0.1767,
+      "step": 50566
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.017816063002704e-06,
+      "loss": 0.2136,
+      "step": 50567
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0174802276963376e-06,
+      "loss": 0.2267,
+      "step": 50568
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0171443923899722e-06,
+      "loss": 0.2123,
+      "step": 50569
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0168085570836065e-06,
+      "loss": 0.2018,
+      "step": 50570
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0164727217772407e-06,
+      "loss": 0.274,
+      "step": 50571
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.016136886470875e-06,
+      "loss": 0.1604,
+      "step": 50572
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.015801051164509e-06,
+      "loss": 0.3259,
+      "step": 50573
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0154652158581434e-06,
+      "loss": 0.1308,
+      "step": 50574
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.015129380551778e-06,
+      "loss": 0.2282,
+      "step": 50575
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.014793545245412e-06,
+      "loss": 0.2922,
+      "step": 50576
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.014457709939046e-06,
+      "loss": 0.3345,
+      "step": 50577
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0141218746326807e-06,
+      "loss": 0.4172,
+      "step": 50578
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0137860393263145e-06,
+      "loss": 0.3548,
+      "step": 50579
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0134502040199487e-06,
+      "loss": 0.3255,
+      "step": 50580
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0131143687135834e-06,
+      "loss": 0.2604,
+      "step": 50581
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0127785334072176e-06,
+      "loss": 0.2322,
+      "step": 50582
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0124426981008514e-06,
+      "loss": 0.2261,
+      "step": 50583
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.012106862794486e-06,
+      "loss": 0.2513,
+      "step": 50584
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0117710274881203e-06,
+      "loss": 0.1082,
+      "step": 50585
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.011435192181754e-06,
+      "loss": 0.1833,
+      "step": 50586
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0110993568753887e-06,
+      "loss": 0.1478,
+      "step": 50587
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.010763521569023e-06,
+      "loss": 0.4166,
+      "step": 50588
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0104276862626567e-06,
+      "loss": 0.331,
+      "step": 50589
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0100918509562914e-06,
+      "loss": 0.208,
+      "step": 50590
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0097560156499256e-06,
+      "loss": 0.2549,
+      "step": 50591
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.00942018034356e-06,
+      "loss": 0.2773,
+      "step": 50592
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0090843450371945e-06,
+      "loss": 0.3426,
+      "step": 50593
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0087485097308283e-06,
+      "loss": 0.1628,
+      "step": 50594
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0084126744244625e-06,
+      "loss": 0.2623,
+      "step": 50595
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.008076839118097e-06,
+      "loss": 0.27,
+      "step": 50596
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.007741003811731e-06,
+      "loss": 0.2127,
+      "step": 50597
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.007405168505365e-06,
+      "loss": 0.413,
+      "step": 50598
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0070693331989994e-06,
+      "loss": 0.3611,
+      "step": 50599
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0067334978926336e-06,
+      "loss": 0.2548,
+      "step": 50600
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.006397662586268e-06,
+      "loss": 0.1564,
+      "step": 50601
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.006061827279902e-06,
+      "loss": 0.252,
+      "step": 50602
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0057259919735367e-06,
+      "loss": 0.1662,
+      "step": 50603
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0053901566671705e-06,
+      "loss": 0.2359,
+      "step": 50604
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0050543213608047e-06,
+      "loss": 0.1845,
+      "step": 50605
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0047184860544394e-06,
+      "loss": 0.1154,
+      "step": 50606
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.004382650748073e-06,
+      "loss": 0.2,
+      "step": 50607
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0040468154417074e-06,
+      "loss": 0.229,
+      "step": 50608
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.003710980135342e-06,
+      "loss": 0.3162,
+      "step": 50609
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0033751448289763e-06,
+      "loss": 0.2963,
+      "step": 50610
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.00303930952261e-06,
+      "loss": 0.2655,
+      "step": 50611
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0027034742162447e-06,
+      "loss": 0.342,
+      "step": 50612
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.002367638909879e-06,
+      "loss": 0.302,
+      "step": 50613
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0020318036035127e-06,
+      "loss": 0.1782,
+      "step": 50614
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0016959682971474e-06,
+      "loss": 0.1648,
+      "step": 50615
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0013601329907816e-06,
+      "loss": 0.3951,
+      "step": 50616
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0010242976844154e-06,
+      "loss": 0.1254,
+      "step": 50617
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.00068846237805e-06,
+      "loss": 0.1794,
+      "step": 50618
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0003526270716843e-06,
+      "loss": 0.2394,
+      "step": 50619
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 3.0000167917653185e-06,
+      "loss": 0.2052,
+      "step": 50620
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.999680956458953e-06,
+      "loss": 0.319,
+      "step": 50621
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.999345121152587e-06,
+      "loss": 0.3086,
+      "step": 50622
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.999009285846221e-06,
+      "loss": 0.155,
+      "step": 50623
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.998673450539856e-06,
+      "loss": 0.4011,
+      "step": 50624
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9983376152334896e-06,
+      "loss": 0.1966,
+      "step": 50625
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.998001779927124e-06,
+      "loss": 0.4684,
+      "step": 50626
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9976659446207585e-06,
+      "loss": 0.1692,
+      "step": 50627
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9973301093143923e-06,
+      "loss": 0.1391,
+      "step": 50628
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9969942740080265e-06,
+      "loss": 0.3157,
+      "step": 50629
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.996658438701661e-06,
+      "loss": 0.2659,
+      "step": 50630
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9963226033952954e-06,
+      "loss": 0.167,
+      "step": 50631
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.995986768088929e-06,
+      "loss": 0.2172,
+      "step": 50632
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.995650932782564e-06,
+      "loss": 0.3074,
+      "step": 50633
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.995315097476198e-06,
+      "loss": 0.4018,
+      "step": 50634
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.994979262169832e-06,
+      "loss": 0.3031,
+      "step": 50635
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9946434268634665e-06,
+      "loss": 0.1214,
+      "step": 50636
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9943075915571007e-06,
+      "loss": 0.246,
+      "step": 50637
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.993971756250735e-06,
+      "loss": 0.2404,
+      "step": 50638
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.993635920944369e-06,
+      "loss": 0.388,
+      "step": 50639
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9933000856380034e-06,
+      "loss": 0.3453,
+      "step": 50640
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9929642503316376e-06,
+      "loss": 0.2516,
+      "step": 50641
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9926284150252723e-06,
+      "loss": 0.493,
+      "step": 50642
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.992292579718906e-06,
+      "loss": 0.2017,
+      "step": 50643
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9919567444125403e-06,
+      "loss": 0.3087,
+      "step": 50644
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.991620909106174e-06,
+      "loss": 0.2338,
+      "step": 50645
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9912850737998088e-06,
+      "loss": 0.3361,
+      "step": 50646
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.990949238493443e-06,
+      "loss": 0.321,
+      "step": 50647
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.990613403187077e-06,
+      "loss": 0.2175,
+      "step": 50648
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.990277567880712e-06,
+      "loss": 0.2535,
+      "step": 50649
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9899417325743457e-06,
+      "loss": 0.2989,
+      "step": 50650
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.98960589726798e-06,
+      "loss": 0.2316,
+      "step": 50651
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9892700619616145e-06,
+      "loss": 0.3293,
+      "step": 50652
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9889342266552483e-06,
+      "loss": 0.4369,
+      "step": 50653
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9885983913488825e-06,
+      "loss": 0.3213,
+      "step": 50654
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.988262556042517e-06,
+      "loss": 0.2809,
+      "step": 50655
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.987926720736151e-06,
+      "loss": 0.2531,
+      "step": 50656
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9875908854297852e-06,
+      "loss": 0.2154,
+      "step": 50657
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.98725505012342e-06,
+      "loss": 0.4075,
+      "step": 50658
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.986919214817054e-06,
+      "loss": 0.3295,
+      "step": 50659
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.986583379510688e-06,
+      "loss": 0.177,
+      "step": 50660
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9862475442043225e-06,
+      "loss": 0.265,
+      "step": 50661
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9859117088979568e-06,
+      "loss": 0.3962,
+      "step": 50662
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9855758735915906e-06,
+      "loss": 0.3018,
+      "step": 50663
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.985240038285225e-06,
+      "loss": 0.3537,
+      "step": 50664
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9849042029788594e-06,
+      "loss": 0.2134,
+      "step": 50665
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9845683676724937e-06,
+      "loss": 0.4116,
+      "step": 50666
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.984232532366128e-06,
+      "loss": 0.258,
+      "step": 50667
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.983896697059762e-06,
+      "loss": 0.2065,
+      "step": 50668
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9835608617533963e-06,
+      "loss": 0.2574,
+      "step": 50669
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.983225026447031e-06,
+      "loss": 0.2824,
+      "step": 50670
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9828891911406648e-06,
+      "loss": 0.2161,
+      "step": 50671
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.982553355834299e-06,
+      "loss": 0.1918,
+      "step": 50672
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9822175205279336e-06,
+      "loss": 0.1668,
+      "step": 50673
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9818816852215674e-06,
+      "loss": 0.3628,
+      "step": 50674
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9815458499152017e-06,
+      "loss": 0.1849,
+      "step": 50675
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9812100146088363e-06,
+      "loss": 0.2676,
+      "step": 50676
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9808741793024705e-06,
+      "loss": 0.1965,
+      "step": 50677
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9805383439961043e-06,
+      "loss": 0.171,
+      "step": 50678
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.980202508689739e-06,
+      "loss": 0.1634,
+      "step": 50679
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9798666733833732e-06,
+      "loss": 0.132,
+      "step": 50680
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.979530838077007e-06,
+      "loss": 0.1644,
+      "step": 50681
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9791950027706417e-06,
+      "loss": 0.2759,
+      "step": 50682
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.978859167464276e-06,
+      "loss": 0.2466,
+      "step": 50683
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9785233321579097e-06,
+      "loss": 0.201,
+      "step": 50684
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9781874968515443e-06,
+      "loss": 0.1923,
+      "step": 50685
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9778516615451786e-06,
+      "loss": 0.2749,
+      "step": 50686
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9775158262388128e-06,
+      "loss": 0.15,
+      "step": 50687
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9771799909324466e-06,
+      "loss": 0.2262,
+      "step": 50688
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9768441556260812e-06,
+      "loss": 0.2847,
+      "step": 50689
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9765083203197155e-06,
+      "loss": 0.3295,
+      "step": 50690
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9761724850133493e-06,
+      "loss": 0.3863,
+      "step": 50691
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.975836649706984e-06,
+      "loss": 0.1752,
+      "step": 50692
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.975500814400618e-06,
+      "loss": 0.2889,
+      "step": 50693
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9751649790942523e-06,
+      "loss": 0.2895,
+      "step": 50694
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.974829143787887e-06,
+      "loss": 0.3319,
+      "step": 50695
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.974493308481521e-06,
+      "loss": 0.2565,
+      "step": 50696
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.974157473175155e-06,
+      "loss": 0.1893,
+      "step": 50697
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9738216378687897e-06,
+      "loss": 0.2624,
+      "step": 50698
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9734858025624235e-06,
+      "loss": 0.1374,
+      "step": 50699
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9731499672560577e-06,
+      "loss": 0.1993,
+      "step": 50700
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9728141319496923e-06,
+      "loss": 0.3768,
+      "step": 50701
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.972478296643326e-06,
+      "loss": 0.359,
+      "step": 50702
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9721424613369604e-06,
+      "loss": 0.1784,
+      "step": 50703
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.971806626030595e-06,
+      "loss": 0.1919,
+      "step": 50704
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9714707907242292e-06,
+      "loss": 0.1401,
+      "step": 50705
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.971134955417863e-06,
+      "loss": 0.3545,
+      "step": 50706
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9707991201114977e-06,
+      "loss": 0.3199,
+      "step": 50707
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.970463284805132e-06,
+      "loss": 0.1513,
+      "step": 50708
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9701274494987657e-06,
+      "loss": 0.2387,
+      "step": 50709
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9697916141924004e-06,
+      "loss": 0.4325,
+      "step": 50710
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9694557788860346e-06,
+      "loss": 0.2709,
+      "step": 50711
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.969119943579669e-06,
+      "loss": 0.2458,
+      "step": 50712
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.968784108273303e-06,
+      "loss": 0.3516,
+      "step": 50713
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9684482729669373e-06,
+      "loss": 0.311,
+      "step": 50714
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.9681124376605715e-06,
+      "loss": 0.3549,
+      "step": 50715
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.967776602354206e-06,
+      "loss": 0.2141,
+      "step": 50716
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.96744076704784e-06,
+      "loss": 0.3549,
+      "step": 50717
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.967104931741474e-06,
+      "loss": 0.3595,
+      "step": 50718
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.966769096435109e-06,
+      "loss": 0.3778,
+      "step": 50719
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9664332611287426e-06,
+      "loss": 0.31,
+      "step": 50720
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.966097425822377e-06,
+      "loss": 0.2188,
+      "step": 50721
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9657615905160115e-06,
+      "loss": 0.2584,
+      "step": 50722
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9654257552096457e-06,
+      "loss": 0.2926,
+      "step": 50723
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9650899199032795e-06,
+      "loss": 0.2717,
+      "step": 50724
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.964754084596914e-06,
+      "loss": 0.1919,
+      "step": 50725
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9644182492905484e-06,
+      "loss": 0.2994,
+      "step": 50726
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.964082413984182e-06,
+      "loss": 0.2165,
+      "step": 50727
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.963746578677817e-06,
+      "loss": 0.3713,
+      "step": 50728
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.963410743371451e-06,
+      "loss": 0.2835,
+      "step": 50729
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.963074908065085e-06,
+      "loss": 0.292,
+      "step": 50730
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9627390727587195e-06,
+      "loss": 0.2602,
+      "step": 50731
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9624032374523537e-06,
+      "loss": 0.2288,
+      "step": 50732
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.962067402145988e-06,
+      "loss": 0.3264,
+      "step": 50733
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9617315668396217e-06,
+      "loss": 0.29,
+      "step": 50734
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9613957315332564e-06,
+      "loss": 0.3018,
+      "step": 50735
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9610598962268906e-06,
+      "loss": 0.3687,
+      "step": 50736
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9607240609205244e-06,
+      "loss": 0.2531,
+      "step": 50737
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.960388225614159e-06,
+      "loss": 0.2275,
+      "step": 50738
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9600523903077933e-06,
+      "loss": 0.1326,
+      "step": 50739
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9597165550014275e-06,
+      "loss": 0.2822,
+      "step": 50740
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9593807196950617e-06,
+      "loss": 0.2942,
+      "step": 50741
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.959044884388696e-06,
+      "loss": 0.1712,
+      "step": 50742
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.95870904908233e-06,
+      "loss": 0.3146,
+      "step": 50743
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.958373213775965e-06,
+      "loss": 0.3107,
+      "step": 50744
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9580373784695986e-06,
+      "loss": 0.204,
+      "step": 50745
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.957701543163233e-06,
+      "loss": 0.3344,
+      "step": 50746
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9573657078568675e-06,
+      "loss": 0.1646,
+      "step": 50747
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9570298725505013e-06,
+      "loss": 0.3543,
+      "step": 50748
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9566940372441355e-06,
+      "loss": 0.387,
+      "step": 50749
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.95635820193777e-06,
+      "loss": 0.1194,
+      "step": 50750
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9560223666314044e-06,
+      "loss": 0.3864,
+      "step": 50751
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.955686531325038e-06,
+      "loss": 0.2349,
+      "step": 50752
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.955350696018673e-06,
+      "loss": 0.2346,
+      "step": 50753
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.955014860712307e-06,
+      "loss": 0.3976,
+      "step": 50754
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.954679025405941e-06,
+      "loss": 0.2791,
+      "step": 50755
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9543431900995755e-06,
+      "loss": 0.4662,
+      "step": 50756
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9540073547932097e-06,
+      "loss": 0.3293,
+      "step": 50757
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9536715194868435e-06,
+      "loss": 0.2989,
+      "step": 50758
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.953335684180478e-06,
+      "loss": 0.2198,
+      "step": 50759
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9529998488741124e-06,
+      "loss": 0.2647,
+      "step": 50760
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9526640135677466e-06,
+      "loss": 0.2424,
+      "step": 50761
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9523281782613813e-06,
+      "loss": 0.3271,
+      "step": 50762
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.951992342955015e-06,
+      "loss": 0.3456,
+      "step": 50763
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9516565076486493e-06,
+      "loss": 0.2026,
+      "step": 50764
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.951320672342284e-06,
+      "loss": 0.4956,
+      "step": 50765
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9509848370359177e-06,
+      "loss": 0.2226,
+      "step": 50766
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.950649001729552e-06,
+      "loss": 0.3856,
+      "step": 50767
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9503131664231866e-06,
+      "loss": 0.2586,
+      "step": 50768
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9499773311168204e-06,
+      "loss": 0.4322,
+      "step": 50769
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9496414958104546e-06,
+      "loss": 0.2357,
+      "step": 50770
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9493056605040893e-06,
+      "loss": 0.2818,
+      "step": 50771
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9489698251977235e-06,
+      "loss": 0.4614,
+      "step": 50772
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9486339898913573e-06,
+      "loss": 0.1504,
+      "step": 50773
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.948298154584992e-06,
+      "loss": 0.4243,
+      "step": 50774
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.947962319278626e-06,
+      "loss": 0.1979,
+      "step": 50775
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.94762648397226e-06,
+      "loss": 0.2704,
+      "step": 50776
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9472906486658946e-06,
+      "loss": 0.3787,
+      "step": 50777
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.946954813359529e-06,
+      "loss": 0.3764,
+      "step": 50778
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.946618978053163e-06,
+      "loss": 0.2314,
+      "step": 50779
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.946283142746797e-06,
+      "loss": 0.2395,
+      "step": 50780
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9459473074404315e-06,
+      "loss": 0.479,
+      "step": 50781
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9456114721340657e-06,
+      "loss": 0.1384,
+      "step": 50782
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9452756368276995e-06,
+      "loss": 0.26,
+      "step": 50783
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.944939801521334e-06,
+      "loss": 0.246,
+      "step": 50784
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9446039662149684e-06,
+      "loss": 0.3907,
+      "step": 50785
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9442681309086022e-06,
+      "loss": 0.2533,
+      "step": 50786
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.943932295602237e-06,
+      "loss": 0.2131,
+      "step": 50787
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.943596460295871e-06,
+      "loss": 0.2294,
+      "step": 50788
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9432606249895053e-06,
+      "loss": 0.2649,
+      "step": 50789
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.94292478968314e-06,
+      "loss": 0.2019,
+      "step": 50790
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9425889543767738e-06,
+      "loss": 0.1994,
+      "step": 50791
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.942253119070408e-06,
+      "loss": 0.1391,
+      "step": 50792
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9419172837640426e-06,
+      "loss": 0.3032,
+      "step": 50793
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9415814484576764e-06,
+      "loss": 0.453,
+      "step": 50794
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9412456131513107e-06,
+      "loss": 0.5631,
+      "step": 50795
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9409097778449453e-06,
+      "loss": 0.3281,
+      "step": 50796
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.940573942538579e-06,
+      "loss": 0.2531,
+      "step": 50797
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9402381072322133e-06,
+      "loss": 0.3941,
+      "step": 50798
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.939902271925848e-06,
+      "loss": 0.2914,
+      "step": 50799
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.939566436619482e-06,
+      "loss": 0.2339,
+      "step": 50800
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.939230601313116e-06,
+      "loss": 0.1854,
+      "step": 50801
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9388947660067506e-06,
+      "loss": 0.1545,
+      "step": 50802
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.938558930700385e-06,
+      "loss": 0.4274,
+      "step": 50803
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9382230953940187e-06,
+      "loss": 0.2106,
+      "step": 50804
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9378872600876533e-06,
+      "loss": 0.337,
+      "step": 50805
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9375514247812875e-06,
+      "loss": 0.2994,
+      "step": 50806
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9372155894749218e-06,
+      "loss": 0.3101,
+      "step": 50807
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.936879754168556e-06,
+      "loss": 0.3756,
+      "step": 50808
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9365439188621902e-06,
+      "loss": 0.1572,
+      "step": 50809
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9362080835558244e-06,
+      "loss": 0.3505,
+      "step": 50810
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.935872248249459e-06,
+      "loss": 0.2901,
+      "step": 50811
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.935536412943093e-06,
+      "loss": 0.2783,
+      "step": 50812
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.935200577636727e-06,
+      "loss": 0.3017,
+      "step": 50813
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9348647423303618e-06,
+      "loss": 0.1521,
+      "step": 50814
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9345289070239956e-06,
+      "loss": 0.3086,
+      "step": 50815
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9341930717176298e-06,
+      "loss": 0.1706,
+      "step": 50816
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9338572364112644e-06,
+      "loss": 0.3381,
+      "step": 50817
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9335214011048987e-06,
+      "loss": 0.3566,
+      "step": 50818
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9331855657985325e-06,
+      "loss": 0.3213,
+      "step": 50819
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.932849730492167e-06,
+      "loss": 0.3802,
+      "step": 50820
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9325138951858013e-06,
+      "loss": 0.1536,
+      "step": 50821
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.932178059879435e-06,
+      "loss": 0.2875,
+      "step": 50822
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9318422245730698e-06,
+      "loss": 0.3183,
+      "step": 50823
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.931506389266704e-06,
+      "loss": 0.2436,
+      "step": 50824
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.931170553960338e-06,
+      "loss": 0.2159,
+      "step": 50825
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.930834718653972e-06,
+      "loss": 0.1682,
+      "step": 50826
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9304988833476067e-06,
+      "loss": 0.2287,
+      "step": 50827
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.930163048041241e-06,
+      "loss": 0.2754,
+      "step": 50828
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9298272127348747e-06,
+      "loss": 0.1929,
+      "step": 50829
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9294913774285093e-06,
+      "loss": 0.2378,
+      "step": 50830
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9291555421221436e-06,
+      "loss": 0.329,
+      "step": 50831
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9288197068157774e-06,
+      "loss": 0.4192,
+      "step": 50832
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.928483871509412e-06,
+      "loss": 0.1694,
+      "step": 50833
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9281480362030462e-06,
+      "loss": 0.2012,
+      "step": 50834
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9278122008966805e-06,
+      "loss": 0.3133,
+      "step": 50835
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.927476365590315e-06,
+      "loss": 0.2668,
+      "step": 50836
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.927140530283949e-06,
+      "loss": 0.2568,
+      "step": 50837
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.926804694977583e-06,
+      "loss": 0.2874,
+      "step": 50838
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9264688596712178e-06,
+      "loss": 0.257,
+      "step": 50839
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9261330243648516e-06,
+      "loss": 0.3197,
+      "step": 50840
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.925797189058486e-06,
+      "loss": 0.1873,
+      "step": 50841
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9254613537521204e-06,
+      "loss": 0.2065,
+      "step": 50842
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9251255184457542e-06,
+      "loss": 0.2369,
+      "step": 50843
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9247896831393885e-06,
+      "loss": 0.368,
+      "step": 50844
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.924453847833023e-06,
+      "loss": 0.1288,
+      "step": 50845
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9241180125266573e-06,
+      "loss": 0.3818,
+      "step": 50846
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.923782177220291e-06,
+      "loss": 0.2213,
+      "step": 50847
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.923446341913926e-06,
+      "loss": 0.1806,
+      "step": 50848
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.92311050660756e-06,
+      "loss": 0.2192,
+      "step": 50849
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.922774671301194e-06,
+      "loss": 0.2256,
+      "step": 50850
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9224388359948285e-06,
+      "loss": 0.1934,
+      "step": 50851
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9221030006884627e-06,
+      "loss": 0.2099,
+      "step": 50852
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.921767165382097e-06,
+      "loss": 0.2717,
+      "step": 50853
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.921431330075731e-06,
+      "loss": 0.3988,
+      "step": 50854
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9210954947693654e-06,
+      "loss": 0.1441,
+      "step": 50855
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9207596594629996e-06,
+      "loss": 0.1669,
+      "step": 50856
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9204238241566342e-06,
+      "loss": 0.2551,
+      "step": 50857
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.920087988850268e-06,
+      "loss": 0.305,
+      "step": 50858
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9197521535439023e-06,
+      "loss": 0.2802,
+      "step": 50859
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.919416318237537e-06,
+      "loss": 0.2948,
+      "step": 50860
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9190804829311707e-06,
+      "loss": 0.2272,
+      "step": 50861
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.918744647624805e-06,
+      "loss": 0.23,
+      "step": 50862
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9184088123184396e-06,
+      "loss": 0.2713,
+      "step": 50863
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.918072977012074e-06,
+      "loss": 0.3881,
+      "step": 50864
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9177371417057076e-06,
+      "loss": 0.2283,
+      "step": 50865
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9174013063993422e-06,
+      "loss": 0.3473,
+      "step": 50866
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9170654710929765e-06,
+      "loss": 0.2243,
+      "step": 50867
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9167296357866103e-06,
+      "loss": 0.2854,
+      "step": 50868
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9163938004802445e-06,
+      "loss": 0.4286,
+      "step": 50869
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.916057965173879e-06,
+      "loss": 0.2597,
+      "step": 50870
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.915722129867513e-06,
+      "loss": 0.2034,
+      "step": 50871
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.915386294561147e-06,
+      "loss": 0.3349,
+      "step": 50872
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.915050459254782e-06,
+      "loss": 0.2459,
+      "step": 50873
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.914714623948416e-06,
+      "loss": 0.2892,
+      "step": 50874
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.91437878864205e-06,
+      "loss": 0.2698,
+      "step": 50875
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9140429533356845e-06,
+      "loss": 0.2629,
+      "step": 50876
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9137071180293187e-06,
+      "loss": 0.4701,
+      "step": 50877
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9133712827229525e-06,
+      "loss": 0.4952,
+      "step": 50878
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.913035447416587e-06,
+      "loss": 0.2324,
+      "step": 50879
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9126996121102214e-06,
+      "loss": 0.2039,
+      "step": 50880
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9123637768038556e-06,
+      "loss": 0.3231,
+      "step": 50881
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.91202794149749e-06,
+      "loss": 0.1401,
+      "step": 50882
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.911692106191124e-06,
+      "loss": 0.2986,
+      "step": 50883
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9113562708847583e-06,
+      "loss": 0.2429,
+      "step": 50884
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.911020435578393e-06,
+      "loss": 0.3082,
+      "step": 50885
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9106846002720267e-06,
+      "loss": 0.2301,
+      "step": 50886
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.910348764965661e-06,
+      "loss": 0.1487,
+      "step": 50887
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9100129296592956e-06,
+      "loss": 0.2055,
+      "step": 50888
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9096770943529294e-06,
+      "loss": 0.1882,
+      "step": 50889
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9093412590465636e-06,
+      "loss": 0.3249,
+      "step": 50890
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9090054237401983e-06,
+      "loss": 0.2213,
+      "step": 50891
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9086695884338325e-06,
+      "loss": 0.3117,
+      "step": 50892
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9083337531274663e-06,
+      "loss": 0.1662,
+      "step": 50893
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.907997917821101e-06,
+      "loss": 0.2408,
+      "step": 50894
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.907662082514735e-06,
+      "loss": 0.3422,
+      "step": 50895
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.907326247208369e-06,
+      "loss": 0.3209,
+      "step": 50896
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9069904119020036e-06,
+      "loss": 0.3346,
+      "step": 50897
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.906654576595638e-06,
+      "loss": 0.209,
+      "step": 50898
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9063187412892716e-06,
+      "loss": 0.3199,
+      "step": 50899
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9059829059829063e-06,
+      "loss": 0.3752,
+      "step": 50900
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9056470706765405e-06,
+      "loss": 0.2564,
+      "step": 50901
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9053112353701747e-06,
+      "loss": 0.2711,
+      "step": 50902
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9049754000638094e-06,
+      "loss": 0.2231,
+      "step": 50903
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.904639564757443e-06,
+      "loss": 0.3639,
+      "step": 50904
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9043037294510774e-06,
+      "loss": 0.227,
+      "step": 50905
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.903967894144712e-06,
+      "loss": 0.253,
+      "step": 50906
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.903632058838346e-06,
+      "loss": 0.2008,
+      "step": 50907
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.90329622353198e-06,
+      "loss": 0.14,
+      "step": 50908
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9029603882256147e-06,
+      "loss": 0.1997,
+      "step": 50909
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9026245529192485e-06,
+      "loss": 0.4347,
+      "step": 50910
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9022887176128827e-06,
+      "loss": 0.2937,
+      "step": 50911
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9019528823065174e-06,
+      "loss": 0.2648,
+      "step": 50912
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9016170470001516e-06,
+      "loss": 0.3667,
+      "step": 50913
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9012812116937854e-06,
+      "loss": 0.3485,
+      "step": 50914
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9009453763874196e-06,
+      "loss": 0.1529,
+      "step": 50915
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.9006095410810543e-06,
+      "loss": 0.1154,
+      "step": 50916
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.900273705774688e-06,
+      "loss": 0.3163,
+      "step": 50917
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8999378704683223e-06,
+      "loss": 0.3333,
+      "step": 50918
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.899602035161957e-06,
+      "loss": 0.344,
+      "step": 50919
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.899266199855591e-06,
+      "loss": 0.395,
+      "step": 50920
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.898930364549225e-06,
+      "loss": 0.2544,
+      "step": 50921
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8985945292428596e-06,
+      "loss": 0.131,
+      "step": 50922
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.898258693936494e-06,
+      "loss": 0.3804,
+      "step": 50923
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8979228586301277e-06,
+      "loss": 0.5472,
+      "step": 50924
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8975870233237623e-06,
+      "loss": 0.2277,
+      "step": 50925
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8972511880173965e-06,
+      "loss": 0.1759,
+      "step": 50926
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8969153527110303e-06,
+      "loss": 0.2116,
+      "step": 50927
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.896579517404665e-06,
+      "loss": 0.2874,
+      "step": 50928
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.896243682098299e-06,
+      "loss": 0.3732,
+      "step": 50929
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8959078467919334e-06,
+      "loss": 0.2666,
+      "step": 50930
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.895572011485568e-06,
+      "loss": 0.1665,
+      "step": 50931
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.895236176179202e-06,
+      "loss": 0.2117,
+      "step": 50932
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.894900340872836e-06,
+      "loss": 0.1954,
+      "step": 50933
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8945645055664707e-06,
+      "loss": 0.2752,
+      "step": 50934
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8942286702601045e-06,
+      "loss": 0.2848,
+      "step": 50935
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8938928349537388e-06,
+      "loss": 0.2661,
+      "step": 50936
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8935569996473734e-06,
+      "loss": 0.2965,
+      "step": 50937
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.893221164341007e-06,
+      "loss": 0.4197,
+      "step": 50938
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8928853290346414e-06,
+      "loss": 0.2657,
+      "step": 50939
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.892549493728276e-06,
+      "loss": 0.4719,
+      "step": 50940
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8922136584219103e-06,
+      "loss": 0.3245,
+      "step": 50941
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.891877823115544e-06,
+      "loss": 0.1845,
+      "step": 50942
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8915419878091788e-06,
+      "loss": 0.135,
+      "step": 50943
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.891206152502813e-06,
+      "loss": 0.1743,
+      "step": 50944
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8908703171964468e-06,
+      "loss": 0.2093,
+      "step": 50945
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8905344818900814e-06,
+      "loss": 0.2542,
+      "step": 50946
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8901986465837156e-06,
+      "loss": 0.1965,
+      "step": 50947
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.88986281127735e-06,
+      "loss": 0.1601,
+      "step": 50948
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.889526975970984e-06,
+      "loss": 0.2637,
+      "step": 50949
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8891911406646183e-06,
+      "loss": 0.4104,
+      "step": 50950
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8888553053582525e-06,
+      "loss": 0.3792,
+      "step": 50951
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.888519470051887e-06,
+      "loss": 0.3054,
+      "step": 50952
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.888183634745521e-06,
+      "loss": 0.2162,
+      "step": 50953
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8878477994391552e-06,
+      "loss": 0.2529,
+      "step": 50954
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.88751196413279e-06,
+      "loss": 0.2066,
+      "step": 50955
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8871761288264237e-06,
+      "loss": 0.2613,
+      "step": 50956
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.886840293520058e-06,
+      "loss": 0.3042,
+      "step": 50957
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8865044582136925e-06,
+      "loss": 0.2085,
+      "step": 50958
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8861686229073268e-06,
+      "loss": 0.2788,
+      "step": 50959
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8858327876009606e-06,
+      "loss": 0.2717,
+      "step": 50960
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8854969522945948e-06,
+      "loss": 0.3611,
+      "step": 50961
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8851611169882294e-06,
+      "loss": 0.2677,
+      "step": 50962
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8848252816818632e-06,
+      "loss": 0.194,
+      "step": 50963
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8844894463754975e-06,
+      "loss": 0.2182,
+      "step": 50964
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.884153611069132e-06,
+      "loss": 0.2439,
+      "step": 50965
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8838177757627663e-06,
+      "loss": 0.2768,
+      "step": 50966
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8834819404564e-06,
+      "loss": 0.2443,
+      "step": 50967
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8831461051500348e-06,
+      "loss": 0.1869,
+      "step": 50968
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.882810269843669e-06,
+      "loss": 0.5452,
+      "step": 50969
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.882474434537303e-06,
+      "loss": 0.3129,
+      "step": 50970
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8821385992309374e-06,
+      "loss": 0.266,
+      "step": 50971
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8818027639245717e-06,
+      "loss": 0.2026,
+      "step": 50972
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8814669286182055e-06,
+      "loss": 0.2578,
+      "step": 50973
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.88113109331184e-06,
+      "loss": 0.4105,
+      "step": 50974
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8807952580054743e-06,
+      "loss": 0.4715,
+      "step": 50975
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8804594226991086e-06,
+      "loss": 0.1752,
+      "step": 50976
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.880123587392743e-06,
+      "loss": 0.1526,
+      "step": 50977
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.879787752086377e-06,
+      "loss": 0.3522,
+      "step": 50978
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8794519167800112e-06,
+      "loss": 0.2714,
+      "step": 50979
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.879116081473646e-06,
+      "loss": 0.1964,
+      "step": 50980
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8787802461672797e-06,
+      "loss": 0.1662,
+      "step": 50981
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.878444410860914e-06,
+      "loss": 0.1325,
+      "step": 50982
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8781085755545486e-06,
+      "loss": 0.3686,
+      "step": 50983
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8777727402481824e-06,
+      "loss": 0.2224,
+      "step": 50984
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8774369049418166e-06,
+      "loss": 0.2488,
+      "step": 50985
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8771010696354512e-06,
+      "loss": 0.2114,
+      "step": 50986
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8767652343290854e-06,
+      "loss": 0.212,
+      "step": 50987
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8764293990227192e-06,
+      "loss": 0.2062,
+      "step": 50988
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.876093563716354e-06,
+      "loss": 0.231,
+      "step": 50989
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.875757728409988e-06,
+      "loss": 0.2361,
+      "step": 50990
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.875421893103622e-06,
+      "loss": 0.2431,
+      "step": 50991
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8750860577972566e-06,
+      "loss": 0.2678,
+      "step": 50992
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.874750222490891e-06,
+      "loss": 0.2073,
+      "step": 50993
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.874414387184525e-06,
+      "loss": 0.2216,
+      "step": 50994
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8740785518781592e-06,
+      "loss": 0.3762,
+      "step": 50995
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8737427165717935e-06,
+      "loss": 0.2601,
+      "step": 50996
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8734068812654277e-06,
+      "loss": 0.4272,
+      "step": 50997
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8730710459590623e-06,
+      "loss": 0.2385,
+      "step": 50998
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.872735210652696e-06,
+      "loss": 0.2528,
+      "step": 50999
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8723993753463304e-06,
+      "loss": 0.1698,
+      "step": 51000
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.872063540039965e-06,
+      "loss": 0.3652,
+      "step": 51001
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.871727704733599e-06,
+      "loss": 0.1788,
+      "step": 51002
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.871391869427233e-06,
+      "loss": 0.3268,
+      "step": 51003
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8710560341208673e-06,
+      "loss": 0.3331,
+      "step": 51004
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.870720198814502e-06,
+      "loss": 0.2139,
+      "step": 51005
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8703843635081357e-06,
+      "loss": 0.3293,
+      "step": 51006
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.87004852820177e-06,
+      "loss": 0.4579,
+      "step": 51007
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8697126928954046e-06,
+      "loss": 0.3598,
+      "step": 51008
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8693768575890384e-06,
+      "loss": 0.2857,
+      "step": 51009
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8690410222826726e-06,
+      "loss": 0.2878,
+      "step": 51010
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8687051869763072e-06,
+      "loss": 0.2418,
+      "step": 51011
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.868369351669941e-06,
+      "loss": 0.184,
+      "step": 51012
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8680335163635753e-06,
+      "loss": 0.3179,
+      "step": 51013
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.86769768105721e-06,
+      "loss": 0.2659,
+      "step": 51014
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.867361845750844e-06,
+      "loss": 0.2779,
+      "step": 51015
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.867026010444478e-06,
+      "loss": 0.4504,
+      "step": 51016
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8666901751381126e-06,
+      "loss": 0.2708,
+      "step": 51017
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.866354339831747e-06,
+      "loss": 0.2691,
+      "step": 51018
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8660185045253806e-06,
+      "loss": 0.1933,
+      "step": 51019
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8656826692190153e-06,
+      "loss": 0.2291,
+      "step": 51020
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8653468339126495e-06,
+      "loss": 0.2112,
+      "step": 51021
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8650109986062837e-06,
+      "loss": 0.2085,
+      "step": 51022
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.864675163299918e-06,
+      "loss": 0.3721,
+      "step": 51023
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.864339327993552e-06,
+      "loss": 0.2067,
+      "step": 51024
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8640034926871864e-06,
+      "loss": 0.1169,
+      "step": 51025
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.863667657380821e-06,
+      "loss": 0.2143,
+      "step": 51026
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.863331822074455e-06,
+      "loss": 0.1808,
+      "step": 51027
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.862995986768089e-06,
+      "loss": 0.3139,
+      "step": 51028
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8626601514617237e-06,
+      "loss": 0.2477,
+      "step": 51029
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8623243161553575e-06,
+      "loss": 0.232,
+      "step": 51030
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8619884808489917e-06,
+      "loss": 0.1925,
+      "step": 51031
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8616526455426264e-06,
+      "loss": 0.1857,
+      "step": 51032
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8613168102362606e-06,
+      "loss": 0.1135,
+      "step": 51033
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8609809749298944e-06,
+      "loss": 0.3315,
+      "step": 51034
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.860645139623529e-06,
+      "loss": 0.1883,
+      "step": 51035
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8603093043171633e-06,
+      "loss": 0.297,
+      "step": 51036
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.859973469010797e-06,
+      "loss": 0.3636,
+      "step": 51037
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8596376337044317e-06,
+      "loss": 0.3188,
+      "step": 51038
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.859301798398066e-06,
+      "loss": 0.3357,
+      "step": 51039
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8589659630916997e-06,
+      "loss": 0.3817,
+      "step": 51040
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8586301277853344e-06,
+      "loss": 0.1602,
+      "step": 51041
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8582942924789686e-06,
+      "loss": 0.2181,
+      "step": 51042
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.857958457172603e-06,
+      "loss": 0.368,
+      "step": 51043
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8576226218662375e-06,
+      "loss": 0.3683,
+      "step": 51044
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8572867865598713e-06,
+      "loss": 0.1369,
+      "step": 51045
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8569509512535055e-06,
+      "loss": 0.285,
+      "step": 51046
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.85661511594714e-06,
+      "loss": 0.1297,
+      "step": 51047
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.856279280640774e-06,
+      "loss": 0.276,
+      "step": 51048
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.855943445334408e-06,
+      "loss": 0.4937,
+      "step": 51049
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8556076100280424e-06,
+      "loss": 0.3178,
+      "step": 51050
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8552717747216766e-06,
+      "loss": 0.335,
+      "step": 51051
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.854935939415311e-06,
+      "loss": 0.208,
+      "step": 51052
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.854600104108945e-06,
+      "loss": 0.3807,
+      "step": 51053
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8542642688025797e-06,
+      "loss": 0.2652,
+      "step": 51054
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8539284334962135e-06,
+      "loss": 0.2988,
+      "step": 51055
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8535925981898477e-06,
+      "loss": 0.1983,
+      "step": 51056
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8532567628834824e-06,
+      "loss": 0.3039,
+      "step": 51057
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.852920927577116e-06,
+      "loss": 0.3681,
+      "step": 51058
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8525850922707504e-06,
+      "loss": 0.1299,
+      "step": 51059
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.852249256964385e-06,
+      "loss": 0.277,
+      "step": 51060
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8519134216580193e-06,
+      "loss": 0.2545,
+      "step": 51061
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.851577586351653e-06,
+      "loss": 0.2959,
+      "step": 51062
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8512417510452877e-06,
+      "loss": 0.3354,
+      "step": 51063
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.850905915738922e-06,
+      "loss": 0.2135,
+      "step": 51064
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8505700804325558e-06,
+      "loss": 0.3079,
+      "step": 51065
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8502342451261904e-06,
+      "loss": 0.3053,
+      "step": 51066
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8498984098198246e-06,
+      "loss": 0.2476,
+      "step": 51067
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8495625745134584e-06,
+      "loss": 0.1933,
+      "step": 51068
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.849226739207093e-06,
+      "loss": 0.4276,
+      "step": 51069
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8488909039007273e-06,
+      "loss": 0.3692,
+      "step": 51070
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8485550685943615e-06,
+      "loss": 0.2714,
+      "step": 51071
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.848219233287996e-06,
+      "loss": 0.265,
+      "step": 51072
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.84788339798163e-06,
+      "loss": 0.3013,
+      "step": 51073
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.847547562675264e-06,
+      "loss": 0.2659,
+      "step": 51074
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.847211727368899e-06,
+      "loss": 0.2831,
+      "step": 51075
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8468758920625326e-06,
+      "loss": 0.2695,
+      "step": 51076
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.846540056756167e-06,
+      "loss": 0.3543,
+      "step": 51077
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8462042214498015e-06,
+      "loss": 0.2852,
+      "step": 51078
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8458683861434353e-06,
+      "loss": 0.3018,
+      "step": 51079
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8455325508370695e-06,
+      "loss": 0.1599,
+      "step": 51080
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.845196715530704e-06,
+      "loss": 0.369,
+      "step": 51081
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8448608802243384e-06,
+      "loss": 0.27,
+      "step": 51082
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.844525044917972e-06,
+      "loss": 0.3735,
+      "step": 51083
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.844189209611607e-06,
+      "loss": 0.2166,
+      "step": 51084
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.843853374305241e-06,
+      "loss": 0.196,
+      "step": 51085
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.843517538998875e-06,
+      "loss": 0.3138,
+      "step": 51086
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8431817036925095e-06,
+      "loss": 0.162,
+      "step": 51087
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8428458683861438e-06,
+      "loss": 0.2464,
+      "step": 51088
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.842510033079778e-06,
+      "loss": 0.154,
+      "step": 51089
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.842174197773412e-06,
+      "loss": 0.1372,
+      "step": 51090
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8418383624670464e-06,
+      "loss": 0.2142,
+      "step": 51091
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8415025271606806e-06,
+      "loss": 0.3425,
+      "step": 51092
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8411666918543153e-06,
+      "loss": 0.1669,
+      "step": 51093
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.840830856547949e-06,
+      "loss": 0.3479,
+      "step": 51094
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8404950212415833e-06,
+      "loss": 0.2347,
+      "step": 51095
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8401591859352175e-06,
+      "loss": 0.239,
+      "step": 51096
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8398233506288518e-06,
+      "loss": 0.2641,
+      "step": 51097
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.839487515322486e-06,
+      "loss": 0.2512,
+      "step": 51098
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8391516800161202e-06,
+      "loss": 0.3162,
+      "step": 51099
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.838815844709755e-06,
+      "loss": 0.317,
+      "step": 51100
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8384800094033887e-06,
+      "loss": 0.3554,
+      "step": 51101
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.838144174097023e-06,
+      "loss": 0.1824,
+      "step": 51102
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8378083387906575e-06,
+      "loss": 0.3658,
+      "step": 51103
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8374725034842913e-06,
+      "loss": 0.3005,
+      "step": 51104
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8371366681779256e-06,
+      "loss": 0.2897,
+      "step": 51105
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.83680083287156e-06,
+      "loss": 0.168,
+      "step": 51106
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8364649975651944e-06,
+      "loss": 0.3218,
+      "step": 51107
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8361291622588282e-06,
+      "loss": 0.1351,
+      "step": 51108
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.835793326952463e-06,
+      "loss": 0.1216,
+      "step": 51109
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.835457491646097e-06,
+      "loss": 0.3407,
+      "step": 51110
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.835121656339731e-06,
+      "loss": 0.3556,
+      "step": 51111
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8347858210333656e-06,
+      "loss": 0.3065,
+      "step": 51112
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8344499857269998e-06,
+      "loss": 0.2311,
+      "step": 51113
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8341141504206336e-06,
+      "loss": 0.1627,
+      "step": 51114
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8337783151142682e-06,
+      "loss": 0.1529,
+      "step": 51115
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.8334424798079024e-06,
+      "loss": 0.2459,
+      "step": 51116
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8331066445015367e-06,
+      "loss": 0.1914,
+      "step": 51117
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8327708091951713e-06,
+      "loss": 0.3652,
+      "step": 51118
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.832434973888805e-06,
+      "loss": 0.3288,
+      "step": 51119
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8320991385824393e-06,
+      "loss": 0.1628,
+      "step": 51120
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.831763303276074e-06,
+      "loss": 0.3463,
+      "step": 51121
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.831427467969708e-06,
+      "loss": 0.2886,
+      "step": 51122
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.831091632663342e-06,
+      "loss": 0.1685,
+      "step": 51123
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8307557973569767e-06,
+      "loss": 0.2291,
+      "step": 51124
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8304199620506105e-06,
+      "loss": 0.1707,
+      "step": 51125
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8300841267442447e-06,
+      "loss": 0.2441,
+      "step": 51126
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8297482914378793e-06,
+      "loss": 0.3244,
+      "step": 51127
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8294124561315136e-06,
+      "loss": 0.5164,
+      "step": 51128
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8290766208251474e-06,
+      "loss": 0.3445,
+      "step": 51129
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.828740785518782e-06,
+      "loss": 0.2456,
+      "step": 51130
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8284049502124162e-06,
+      "loss": 0.1747,
+      "step": 51131
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.82806911490605e-06,
+      "loss": 0.3116,
+      "step": 51132
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8277332795996847e-06,
+      "loss": 0.4034,
+      "step": 51133
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.827397444293319e-06,
+      "loss": 0.1993,
+      "step": 51134
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.827061608986953e-06,
+      "loss": 0.2142,
+      "step": 51135
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8267257736805873e-06,
+      "loss": 0.4856,
+      "step": 51136
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8263899383742216e-06,
+      "loss": 0.1707,
+      "step": 51137
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.826054103067856e-06,
+      "loss": 0.2508,
+      "step": 51138
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8257182677614896e-06,
+      "loss": 0.3347,
+      "step": 51139
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8253824324551242e-06,
+      "loss": 0.2474,
+      "step": 51140
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8250465971487585e-06,
+      "loss": 0.2837,
+      "step": 51141
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8247107618423923e-06,
+      "loss": 0.3151,
+      "step": 51142
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.824374926536027e-06,
+      "loss": 0.2282,
+      "step": 51143
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.824039091229661e-06,
+      "loss": 0.3032,
+      "step": 51144
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8237032559232954e-06,
+      "loss": 0.2598,
+      "step": 51145
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.82336742061693e-06,
+      "loss": 0.1969,
+      "step": 51146
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.823031585310564e-06,
+      "loss": 0.2994,
+      "step": 51147
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.822695750004198e-06,
+      "loss": 0.2787,
+      "step": 51148
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8223599146978327e-06,
+      "loss": 0.2723,
+      "step": 51149
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8220240793914665e-06,
+      "loss": 0.2735,
+      "step": 51150
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8216882440851007e-06,
+      "loss": 0.3095,
+      "step": 51151
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8213524087787354e-06,
+      "loss": 0.1782,
+      "step": 51152
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.821016573472369e-06,
+      "loss": 0.2138,
+      "step": 51153
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8206807381660034e-06,
+      "loss": 0.2976,
+      "step": 51154
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.820344902859638e-06,
+      "loss": 0.2684,
+      "step": 51155
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8200090675532722e-06,
+      "loss": 0.3107,
+      "step": 51156
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.819673232246906e-06,
+      "loss": 0.2927,
+      "step": 51157
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8193373969405407e-06,
+      "loss": 0.4502,
+      "step": 51158
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.819001561634175e-06,
+      "loss": 0.1618,
+      "step": 51159
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8186657263278087e-06,
+      "loss": 0.3333,
+      "step": 51160
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8183298910214434e-06,
+      "loss": 0.1634,
+      "step": 51161
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8179940557150776e-06,
+      "loss": 0.1933,
+      "step": 51162
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.817658220408712e-06,
+      "loss": 0.3024,
+      "step": 51163
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.817322385102346e-06,
+      "loss": 0.2722,
+      "step": 51164
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8169865497959803e-06,
+      "loss": 0.3366,
+      "step": 51165
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8166507144896145e-06,
+      "loss": 0.3325,
+      "step": 51166
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.816314879183249e-06,
+      "loss": 0.2323,
+      "step": 51167
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.815979043876883e-06,
+      "loss": 0.2489,
+      "step": 51168
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.815643208570517e-06,
+      "loss": 0.2339,
+      "step": 51169
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.815307373264152e-06,
+      "loss": 0.2777,
+      "step": 51170
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8149715379577856e-06,
+      "loss": 0.2933,
+      "step": 51171
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.81463570265142e-06,
+      "loss": 0.2907,
+      "step": 51172
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8142998673450545e-06,
+      "loss": 0.4556,
+      "step": 51173
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8139640320386887e-06,
+      "loss": 0.151,
+      "step": 51174
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8136281967323225e-06,
+      "loss": 0.165,
+      "step": 51175
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.813292361425957e-06,
+      "loss": 0.3976,
+      "step": 51176
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8129565261195914e-06,
+      "loss": 0.2302,
+      "step": 51177
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.812620690813225e-06,
+      "loss": 0.3087,
+      "step": 51178
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.81228485550686e-06,
+      "loss": 0.2027,
+      "step": 51179
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.811949020200494e-06,
+      "loss": 0.2606,
+      "step": 51180
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.811613184894128e-06,
+      "loss": 0.4227,
+      "step": 51181
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8112773495877625e-06,
+      "loss": 0.2582,
+      "step": 51182
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8109415142813967e-06,
+      "loss": 0.2185,
+      "step": 51183
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.810605678975031e-06,
+      "loss": 0.2513,
+      "step": 51184
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8102698436686647e-06,
+      "loss": 0.254,
+      "step": 51185
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8099340083622994e-06,
+      "loss": 0.2155,
+      "step": 51186
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8095981730559336e-06,
+      "loss": 0.1907,
+      "step": 51187
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8092623377495674e-06,
+      "loss": 0.1851,
+      "step": 51188
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.808926502443202e-06,
+      "loss": 0.1756,
+      "step": 51189
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8085906671368363e-06,
+      "loss": 0.1613,
+      "step": 51190
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8082548318304705e-06,
+      "loss": 0.2034,
+      "step": 51191
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8079189965241047e-06,
+      "loss": 0.4513,
+      "step": 51192
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.807583161217739e-06,
+      "loss": 0.4083,
+      "step": 51193
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.807247325911373e-06,
+      "loss": 0.2789,
+      "step": 51194
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.806911490605008e-06,
+      "loss": 0.2005,
+      "step": 51195
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8065756552986416e-06,
+      "loss": 0.2292,
+      "step": 51196
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.806239819992276e-06,
+      "loss": 0.2731,
+      "step": 51197
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8059039846859105e-06,
+      "loss": 0.2465,
+      "step": 51198
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8055681493795443e-06,
+      "loss": 0.2984,
+      "step": 51199
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8052323140731785e-06,
+      "loss": 0.2292,
+      "step": 51200
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.804896478766813e-06,
+      "loss": 0.233,
+      "step": 51201
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8045606434604474e-06,
+      "loss": 0.1745,
+      "step": 51202
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.804224808154081e-06,
+      "loss": 0.2013,
+      "step": 51203
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.803888972847716e-06,
+      "loss": 0.3086,
+      "step": 51204
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.80355313754135e-06,
+      "loss": 0.1912,
+      "step": 51205
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.803217302234984e-06,
+      "loss": 0.2625,
+      "step": 51206
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8028814669286185e-06,
+      "loss": 0.3396,
+      "step": 51207
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8025456316222527e-06,
+      "loss": 0.4533,
+      "step": 51208
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8022097963158865e-06,
+      "loss": 0.1932,
+      "step": 51209
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.801873961009521e-06,
+      "loss": 0.3418,
+      "step": 51210
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8015381257031554e-06,
+      "loss": 0.316,
+      "step": 51211
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8012022903967896e-06,
+      "loss": 0.3089,
+      "step": 51212
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8008664550904243e-06,
+      "loss": 0.4573,
+      "step": 51213
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.800530619784058e-06,
+      "loss": 0.3327,
+      "step": 51214
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.8001947844776923e-06,
+      "loss": 0.2034,
+      "step": 51215
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.799858949171327e-06,
+      "loss": 0.1132,
+      "step": 51216
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7995231138649608e-06,
+      "loss": 0.3975,
+      "step": 51217
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.799187278558595e-06,
+      "loss": 0.2262,
+      "step": 51218
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7988514432522296e-06,
+      "loss": 0.3471,
+      "step": 51219
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7985156079458634e-06,
+      "loss": 0.1995,
+      "step": 51220
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7981797726394976e-06,
+      "loss": 0.353,
+      "step": 51221
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7978439373331323e-06,
+      "loss": 0.3008,
+      "step": 51222
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7975081020267665e-06,
+      "loss": 0.2247,
+      "step": 51223
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7971722667204003e-06,
+      "loss": 0.1971,
+      "step": 51224
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.796836431414035e-06,
+      "loss": 0.2269,
+      "step": 51225
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.796500596107669e-06,
+      "loss": 0.4252,
+      "step": 51226
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.796164760801303e-06,
+      "loss": 0.2335,
+      "step": 51227
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7958289254949376e-06,
+      "loss": 0.2553,
+      "step": 51228
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.795493090188572e-06,
+      "loss": 0.3284,
+      "step": 51229
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.795157254882206e-06,
+      "loss": 0.3891,
+      "step": 51230
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.79482141957584e-06,
+      "loss": 0.2546,
+      "step": 51231
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7944855842694745e-06,
+      "loss": 0.2265,
+      "step": 51232
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7941497489631088e-06,
+      "loss": 0.2761,
+      "step": 51233
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7938139136567426e-06,
+      "loss": 0.3143,
+      "step": 51234
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.793478078350377e-06,
+      "loss": 0.3209,
+      "step": 51235
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7931422430440114e-06,
+      "loss": 0.1598,
+      "step": 51236
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7928064077376457e-06,
+      "loss": 0.2043,
+      "step": 51237
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.79247057243128e-06,
+      "loss": 0.2172,
+      "step": 51238
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.792134737124914e-06,
+      "loss": 0.2938,
+      "step": 51239
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7917989018185483e-06,
+      "loss": 0.1419,
+      "step": 51240
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.791463066512183e-06,
+      "loss": 0.2715,
+      "step": 51241
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7911272312058168e-06,
+      "loss": 0.3842,
+      "step": 51242
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.790791395899451e-06,
+      "loss": 0.2463,
+      "step": 51243
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7904555605930856e-06,
+      "loss": 0.411,
+      "step": 51244
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7901197252867194e-06,
+      "loss": 0.1743,
+      "step": 51245
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7897838899803537e-06,
+      "loss": 0.2306,
+      "step": 51246
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7894480546739883e-06,
+      "loss": 0.3562,
+      "step": 51247
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7891122193676225e-06,
+      "loss": 0.4321,
+      "step": 51248
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7887763840612563e-06,
+      "loss": 0.4005,
+      "step": 51249
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.788440548754891e-06,
+      "loss": 0.1613,
+      "step": 51250
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.788104713448525e-06,
+      "loss": 0.1911,
+      "step": 51251
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.787768878142159e-06,
+      "loss": 0.1904,
+      "step": 51252
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7874330428357937e-06,
+      "loss": 0.3076,
+      "step": 51253
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.787097207529428e-06,
+      "loss": 0.2954,
+      "step": 51254
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7867613722230617e-06,
+      "loss": 0.1444,
+      "step": 51255
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7864255369166963e-06,
+      "loss": 0.2896,
+      "step": 51256
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7860897016103306e-06,
+      "loss": 0.3621,
+      "step": 51257
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7857538663039648e-06,
+      "loss": 0.2485,
+      "step": 51258
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7854180309975994e-06,
+      "loss": 0.2523,
+      "step": 51259
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7850821956912332e-06,
+      "loss": 0.2478,
+      "step": 51260
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7847463603848674e-06,
+      "loss": 0.34,
+      "step": 51261
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.784410525078502e-06,
+      "loss": 0.3104,
+      "step": 51262
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.784074689772136e-06,
+      "loss": 0.254,
+      "step": 51263
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.78373885446577e-06,
+      "loss": 0.2346,
+      "step": 51264
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7834030191594048e-06,
+      "loss": 0.1785,
+      "step": 51265
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7830671838530386e-06,
+      "loss": 0.1759,
+      "step": 51266
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.782731348546673e-06,
+      "loss": 0.33,
+      "step": 51267
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7823955132403074e-06,
+      "loss": 0.1964,
+      "step": 51268
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7820596779339417e-06,
+      "loss": 0.2503,
+      "step": 51269
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7817238426275755e-06,
+      "loss": 0.3435,
+      "step": 51270
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.78138800732121e-06,
+      "loss": 0.2392,
+      "step": 51271
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7810521720148443e-06,
+      "loss": 0.1596,
+      "step": 51272
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.780716336708478e-06,
+      "loss": 0.1946,
+      "step": 51273
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7803805014021128e-06,
+      "loss": 0.2371,
+      "step": 51274
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.780044666095747e-06,
+      "loss": 0.1469,
+      "step": 51275
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7797088307893812e-06,
+      "loss": 0.1518,
+      "step": 51276
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.779372995483015e-06,
+      "loss": 0.2859,
+      "step": 51277
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7790371601766497e-06,
+      "loss": 0.2837,
+      "step": 51278
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.778701324870284e-06,
+      "loss": 0.1853,
+      "step": 51279
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7783654895639177e-06,
+      "loss": 0.4506,
+      "step": 51280
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7780296542575523e-06,
+      "loss": 0.2431,
+      "step": 51281
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7776938189511866e-06,
+      "loss": 0.1396,
+      "step": 51282
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7773579836448204e-06,
+      "loss": 0.2254,
+      "step": 51283
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.777022148338455e-06,
+      "loss": 0.3543,
+      "step": 51284
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7766863130320892e-06,
+      "loss": 0.2176,
+      "step": 51285
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7763504777257235e-06,
+      "loss": 0.325,
+      "step": 51286
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.776014642419358e-06,
+      "loss": 0.2392,
+      "step": 51287
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.775678807112992e-06,
+      "loss": 0.2685,
+      "step": 51288
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.775342971806626e-06,
+      "loss": 0.3788,
+      "step": 51289
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7750071365002608e-06,
+      "loss": 0.3067,
+      "step": 51290
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7746713011938946e-06,
+      "loss": 0.1497,
+      "step": 51291
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.774335465887529e-06,
+      "loss": 0.1924,
+      "step": 51292
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7739996305811635e-06,
+      "loss": 0.2382,
+      "step": 51293
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7736637952747973e-06,
+      "loss": 0.2941,
+      "step": 51294
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7733279599684315e-06,
+      "loss": 0.3179,
+      "step": 51295
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.772992124662066e-06,
+      "loss": 0.2526,
+      "step": 51296
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7726562893557004e-06,
+      "loss": 0.3236,
+      "step": 51297
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.772320454049334e-06,
+      "loss": 0.3926,
+      "step": 51298
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.771984618742969e-06,
+      "loss": 0.1269,
+      "step": 51299
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.771648783436603e-06,
+      "loss": 0.3355,
+      "step": 51300
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.771312948130237e-06,
+      "loss": 0.3295,
+      "step": 51301
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7709771128238715e-06,
+      "loss": 0.6254,
+      "step": 51302
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7706412775175057e-06,
+      "loss": 0.2126,
+      "step": 51303
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.77030544221114e-06,
+      "loss": 0.2997,
+      "step": 51304
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.769969606904774e-06,
+      "loss": 0.3287,
+      "step": 51305
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7696337715984084e-06,
+      "loss": 0.2855,
+      "step": 51306
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7692979362920426e-06,
+      "loss": 0.3357,
+      "step": 51307
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7689621009856772e-06,
+      "loss": 0.2258,
+      "step": 51308
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.768626265679311e-06,
+      "loss": 0.1728,
+      "step": 51309
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7682904303729453e-06,
+      "loss": 0.2503,
+      "step": 51310
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.76795459506658e-06,
+      "loss": 0.1392,
+      "step": 51311
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7676187597602137e-06,
+      "loss": 0.2533,
+      "step": 51312
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.767282924453848e-06,
+      "loss": 0.2912,
+      "step": 51313
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2.7669470891474826e-06,
+      "loss": 0.2817,
+      "step": 51314
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.766611253841117e-06,
+      "loss": 0.3083,
+      "step": 51315
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7662754185347506e-06,
+      "loss": 0.2646,
+      "step": 51316
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7659395832283853e-06,
+      "loss": 0.4036,
+      "step": 51317
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7656037479220195e-06,
+      "loss": 0.5025,
+      "step": 51318
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7652679126156533e-06,
+      "loss": 0.1922,
+      "step": 51319
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7649320773092875e-06,
+      "loss": 0.3463,
+      "step": 51320
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.764596242002922e-06,
+      "loss": 0.2581,
+      "step": 51321
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.764260406696556e-06,
+      "loss": 0.1986,
+      "step": 51322
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.76392457139019e-06,
+      "loss": 0.1767,
+      "step": 51323
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.763588736083825e-06,
+      "loss": 0.2748,
+      "step": 51324
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.763252900777459e-06,
+      "loss": 0.2531,
+      "step": 51325
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.762917065471093e-06,
+      "loss": 0.3179,
+      "step": 51326
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7625812301647275e-06,
+      "loss": 0.3447,
+      "step": 51327
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7622453948583617e-06,
+      "loss": 0.3524,
+      "step": 51328
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7619095595519955e-06,
+      "loss": 0.1974,
+      "step": 51329
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.76157372424563e-06,
+      "loss": 0.2411,
+      "step": 51330
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7612378889392644e-06,
+      "loss": 0.2548,
+      "step": 51331
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7609020536328986e-06,
+      "loss": 0.2625,
+      "step": 51332
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.760566218326533e-06,
+      "loss": 0.306,
+      "step": 51333
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.760230383020167e-06,
+      "loss": 0.2555,
+      "step": 51334
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7598945477138013e-06,
+      "loss": 0.2484,
+      "step": 51335
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.759558712407436e-06,
+      "loss": 0.1354,
+      "step": 51336
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7592228771010697e-06,
+      "loss": 0.315,
+      "step": 51337
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.758887041794704e-06,
+      "loss": 0.3031,
+      "step": 51338
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7585512064883386e-06,
+      "loss": 0.2344,
+      "step": 51339
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7582153711819724e-06,
+      "loss": 0.1734,
+      "step": 51340
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7578795358756066e-06,
+      "loss": 0.2302,
+      "step": 51341
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7575437005692413e-06,
+      "loss": 0.1736,
+      "step": 51342
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7572078652628755e-06,
+      "loss": 0.1532,
+      "step": 51343
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7568720299565093e-06,
+      "loss": 0.1487,
+      "step": 51344
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.756536194650144e-06,
+      "loss": 0.2568,
+      "step": 51345
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.756200359343778e-06,
+      "loss": 0.3538,
+      "step": 51346
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.755864524037412e-06,
+      "loss": 0.2962,
+      "step": 51347
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7555286887310466e-06,
+      "loss": 0.2376,
+      "step": 51348
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.755192853424681e-06,
+      "loss": 0.2892,
+      "step": 51349
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7548570181183146e-06,
+      "loss": 0.3778,
+      "step": 51350
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7545211828119493e-06,
+      "loss": 0.3531,
+      "step": 51351
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7541853475055835e-06,
+      "loss": 0.1174,
+      "step": 51352
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7538495121992177e-06,
+      "loss": 0.1909,
+      "step": 51353
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7535136768928524e-06,
+      "loss": 0.2057,
+      "step": 51354
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.753177841586486e-06,
+      "loss": 0.1986,
+      "step": 51355
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7528420062801204e-06,
+      "loss": 0.3772,
+      "step": 51356
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.752506170973755e-06,
+      "loss": 0.174,
+      "step": 51357
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.752170335667389e-06,
+      "loss": 0.2841,
+      "step": 51358
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.751834500361023e-06,
+      "loss": 0.1433,
+      "step": 51359
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7514986650546577e-06,
+      "loss": 0.3159,
+      "step": 51360
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.751162829748292e-06,
+      "loss": 0.1709,
+      "step": 51361
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7508269944419258e-06,
+      "loss": 0.2408,
+      "step": 51362
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7504911591355604e-06,
+      "loss": 0.3036,
+      "step": 51363
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7501553238291946e-06,
+      "loss": 0.1828,
+      "step": 51364
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7498194885228284e-06,
+      "loss": 0.2929,
+      "step": 51365
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7494836532164626e-06,
+      "loss": 0.3299,
+      "step": 51366
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7491478179100973e-06,
+      "loss": 0.2841,
+      "step": 51367
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.748811982603731e-06,
+      "loss": 0.1517,
+      "step": 51368
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7484761472973653e-06,
+      "loss": 0.367,
+      "step": 51369
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.748140311991e-06,
+      "loss": 0.4296,
+      "step": 51370
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.747804476684634e-06,
+      "loss": 0.5345,
+      "step": 51371
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.747468641378268e-06,
+      "loss": 0.3626,
+      "step": 51372
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7471328060719026e-06,
+      "loss": 0.2042,
+      "step": 51373
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.746796970765537e-06,
+      "loss": 0.2292,
+      "step": 51374
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7464611354591707e-06,
+      "loss": 0.1539,
+      "step": 51375
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7461253001528053e-06,
+      "loss": 0.2659,
+      "step": 51376
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7457894648464395e-06,
+      "loss": 0.3172,
+      "step": 51377
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7454536295400738e-06,
+      "loss": 0.1476,
+      "step": 51378
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.745117794233708e-06,
+      "loss": 0.2414,
+      "step": 51379
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.744781958927342e-06,
+      "loss": 0.344,
+      "step": 51380
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7444461236209764e-06,
+      "loss": 0.2512,
+      "step": 51381
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.744110288314611e-06,
+      "loss": 0.2003,
+      "step": 51382
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.743774453008245e-06,
+      "loss": 0.2909,
+      "step": 51383
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.743438617701879e-06,
+      "loss": 0.1891,
+      "step": 51384
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7431027823955137e-06,
+      "loss": 0.3073,
+      "step": 51385
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7427669470891475e-06,
+      "loss": 0.2554,
+      "step": 51386
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7424311117827818e-06,
+      "loss": 0.2857,
+      "step": 51387
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7420952764764164e-06,
+      "loss": 0.211,
+      "step": 51388
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7417594411700506e-06,
+      "loss": 0.1896,
+      "step": 51389
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7414236058636844e-06,
+      "loss": 0.2828,
+      "step": 51390
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.741087770557319e-06,
+      "loss": 0.2859,
+      "step": 51391
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7407519352509533e-06,
+      "loss": 0.2311,
+      "step": 51392
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.740416099944587e-06,
+      "loss": 0.2667,
+      "step": 51393
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7400802646382218e-06,
+      "loss": 0.2797,
+      "step": 51394
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.739744429331856e-06,
+      "loss": 0.2592,
+      "step": 51395
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7394085940254898e-06,
+      "loss": 0.2406,
+      "step": 51396
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7390727587191244e-06,
+      "loss": 0.3632,
+      "step": 51397
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7387369234127587e-06,
+      "loss": 0.1836,
+      "step": 51398
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.738401088106393e-06,
+      "loss": 0.294,
+      "step": 51399
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7380652528000275e-06,
+      "loss": 0.2277,
+      "step": 51400
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7377294174936613e-06,
+      "loss": 0.2416,
+      "step": 51401
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7373935821872956e-06,
+      "loss": 0.2359,
+      "step": 51402
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.73705774688093e-06,
+      "loss": 0.2983,
+      "step": 51403
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.736721911574564e-06,
+      "loss": 0.1806,
+      "step": 51404
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7363860762681982e-06,
+      "loss": 0.3556,
+      "step": 51405
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.736050240961833e-06,
+      "loss": 0.2308,
+      "step": 51406
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7357144056554667e-06,
+      "loss": 0.3454,
+      "step": 51407
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.735378570349101e-06,
+      "loss": 0.5333,
+      "step": 51408
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7350427350427355e-06,
+      "loss": 0.2242,
+      "step": 51409
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7347068997363698e-06,
+      "loss": 0.2227,
+      "step": 51410
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7343710644300036e-06,
+      "loss": 0.2232,
+      "step": 51411
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.734035229123638e-06,
+      "loss": 0.3348,
+      "step": 51412
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7336993938172724e-06,
+      "loss": 0.267,
+      "step": 51413
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7333635585109062e-06,
+      "loss": 0.3346,
+      "step": 51414
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7330277232045405e-06,
+      "loss": 0.1995,
+      "step": 51415
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.732691887898175e-06,
+      "loss": 0.2792,
+      "step": 51416
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7323560525918093e-06,
+      "loss": 0.2429,
+      "step": 51417
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.732020217285443e-06,
+      "loss": 0.2921,
+      "step": 51418
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7316843819790778e-06,
+      "loss": 0.3094,
+      "step": 51419
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.731348546672712e-06,
+      "loss": 0.241,
+      "step": 51420
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.731012711366346e-06,
+      "loss": 0.4465,
+      "step": 51421
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7306768760599805e-06,
+      "loss": 0.2504,
+      "step": 51422
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7303410407536147e-06,
+      "loss": 0.2322,
+      "step": 51423
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7300052054472485e-06,
+      "loss": 0.2325,
+      "step": 51424
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.729669370140883e-06,
+      "loss": 0.1631,
+      "step": 51425
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7293335348345174e-06,
+      "loss": 0.2365,
+      "step": 51426
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7289976995281516e-06,
+      "loss": 0.293,
+      "step": 51427
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7286618642217862e-06,
+      "loss": 0.3043,
+      "step": 51428
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.72832602891542e-06,
+      "loss": 0.2534,
+      "step": 51429
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7279901936090542e-06,
+      "loss": 0.1876,
+      "step": 51430
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.727654358302689e-06,
+      "loss": 0.2569,
+      "step": 51431
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7273185229963227e-06,
+      "loss": 0.3569,
+      "step": 51432
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.726982687689957e-06,
+      "loss": 0.2619,
+      "step": 51433
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7266468523835916e-06,
+      "loss": 0.2095,
+      "step": 51434
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7263110170772254e-06,
+      "loss": 0.357,
+      "step": 51435
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7259751817708596e-06,
+      "loss": 0.1728,
+      "step": 51436
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7256393464644942e-06,
+      "loss": 0.4187,
+      "step": 51437
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7253035111581285e-06,
+      "loss": 0.2153,
+      "step": 51438
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7249676758517623e-06,
+      "loss": 0.3897,
+      "step": 51439
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.724631840545397e-06,
+      "loss": 0.1847,
+      "step": 51440
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.724296005239031e-06,
+      "loss": 0.1936,
+      "step": 51441
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.723960169932665e-06,
+      "loss": 0.286,
+      "step": 51442
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7236243346262996e-06,
+      "loss": 0.2557,
+      "step": 51443
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.723288499319934e-06,
+      "loss": 0.1118,
+      "step": 51444
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.722952664013568e-06,
+      "loss": 0.3002,
+      "step": 51445
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7226168287072023e-06,
+      "loss": 0.2141,
+      "step": 51446
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7222809934008365e-06,
+      "loss": 0.2546,
+      "step": 51447
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7219451580944707e-06,
+      "loss": 0.2172,
+      "step": 51448
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7216093227881053e-06,
+      "loss": 0.392,
+      "step": 51449
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.721273487481739e-06,
+      "loss": 0.2616,
+      "step": 51450
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7209376521753734e-06,
+      "loss": 0.2436,
+      "step": 51451
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.720601816869008e-06,
+      "loss": 0.2637,
+      "step": 51452
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.720265981562642e-06,
+      "loss": 0.1404,
+      "step": 51453
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.719930146256276e-06,
+      "loss": 0.1989,
+      "step": 51454
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7195943109499103e-06,
+      "loss": 0.273,
+      "step": 51455
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.719258475643545e-06,
+      "loss": 0.1734,
+      "step": 51456
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7189226403371787e-06,
+      "loss": 0.1915,
+      "step": 51457
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.718586805030813e-06,
+      "loss": 0.3894,
+      "step": 51458
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7182509697244476e-06,
+      "loss": 0.2061,
+      "step": 51459
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7179151344180814e-06,
+      "loss": 0.309,
+      "step": 51460
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7175792991117156e-06,
+      "loss": 0.3451,
+      "step": 51461
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7172434638053503e-06,
+      "loss": 0.2994,
+      "step": 51462
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.716907628498984e-06,
+      "loss": 0.1458,
+      "step": 51463
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7165717931926183e-06,
+      "loss": 0.2345,
+      "step": 51464
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.716235957886253e-06,
+      "loss": 0.3445,
+      "step": 51465
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.715900122579887e-06,
+      "loss": 0.217,
+      "step": 51466
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.715564287273521e-06,
+      "loss": 0.2182,
+      "step": 51467
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7152284519671556e-06,
+      "loss": 0.3657,
+      "step": 51468
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.71489261666079e-06,
+      "loss": 0.2456,
+      "step": 51469
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7145567813544236e-06,
+      "loss": 0.2981,
+      "step": 51470
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7142209460480583e-06,
+      "loss": 0.1095,
+      "step": 51471
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7138851107416925e-06,
+      "loss": 0.4228,
+      "step": 51472
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7135492754353267e-06,
+      "loss": 0.443,
+      "step": 51473
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.713213440128961e-06,
+      "loss": 0.2216,
+      "step": 51474
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.712877604822595e-06,
+      "loss": 0.3371,
+      "step": 51475
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7125417695162294e-06,
+      "loss": 0.3487,
+      "step": 51476
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.712205934209864e-06,
+      "loss": 0.4301,
+      "step": 51477
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.711870098903498e-06,
+      "loss": 0.3663,
+      "step": 51478
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.711534263597132e-06,
+      "loss": 0.2759,
+      "step": 51479
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7111984282907667e-06,
+      "loss": 0.3741,
+      "step": 51480
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7108625929844005e-06,
+      "loss": 0.3748,
+      "step": 51481
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7105267576780347e-06,
+      "loss": 0.2162,
+      "step": 51482
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7101909223716694e-06,
+      "loss": 0.2623,
+      "step": 51483
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7098550870653036e-06,
+      "loss": 0.1457,
+      "step": 51484
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7095192517589374e-06,
+      "loss": 0.2821,
+      "step": 51485
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.709183416452572e-06,
+      "loss": 0.2777,
+      "step": 51486
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7088475811462063e-06,
+      "loss": 0.1866,
+      "step": 51487
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.70851174583984e-06,
+      "loss": 0.2634,
+      "step": 51488
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7081759105334747e-06,
+      "loss": 0.2937,
+      "step": 51489
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.707840075227109e-06,
+      "loss": 0.3262,
+      "step": 51490
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7075042399207427e-06,
+      "loss": 0.2823,
+      "step": 51491
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7071684046143774e-06,
+      "loss": 0.3012,
+      "step": 51492
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7068325693080116e-06,
+      "loss": 0.267,
+      "step": 51493
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.706496734001646e-06,
+      "loss": 0.4392,
+      "step": 51494
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7061608986952805e-06,
+      "loss": 0.2965,
+      "step": 51495
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7058250633889143e-06,
+      "loss": 0.189,
+      "step": 51496
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7054892280825485e-06,
+      "loss": 0.2789,
+      "step": 51497
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.705153392776183e-06,
+      "loss": 0.1625,
+      "step": 51498
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.704817557469817e-06,
+      "loss": 0.265,
+      "step": 51499
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.704481722163451e-06,
+      "loss": 0.5813,
+      "step": 51500
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7041458868570854e-06,
+      "loss": 0.2098,
+      "step": 51501
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.70381005155072e-06,
+      "loss": 0.2504,
+      "step": 51502
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.703474216244354e-06,
+      "loss": 0.235,
+      "step": 51503
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.703138380937988e-06,
+      "loss": 0.2589,
+      "step": 51504
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7028025456316227e-06,
+      "loss": 0.2517,
+      "step": 51505
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7024667103252565e-06,
+      "loss": 0.3241,
+      "step": 51506
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7021308750188908e-06,
+      "loss": 0.2238,
+      "step": 51507
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7017950397125254e-06,
+      "loss": 0.3229,
+      "step": 51508
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.701459204406159e-06,
+      "loss": 0.2743,
+      "step": 51509
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7011233690997934e-06,
+      "loss": 0.2914,
+      "step": 51510
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.700787533793428e-06,
+      "loss": 0.3209,
+      "step": 51511
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.7004516984870623e-06,
+      "loss": 0.1936,
+      "step": 51512
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.700115863180696e-06,
+      "loss": 0.4142,
+      "step": 51513
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6997800278743307e-06,
+      "loss": 0.2622,
+      "step": 51514
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.699444192567965e-06,
+      "loss": 0.1784,
+      "step": 51515
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6991083572615988e-06,
+      "loss": 0.2558,
+      "step": 51516
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6987725219552334e-06,
+      "loss": 0.4335,
+      "step": 51517
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6984366866488676e-06,
+      "loss": 0.2582,
+      "step": 51518
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.698100851342502e-06,
+      "loss": 0.1947,
+      "step": 51519
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.697765016036136e-06,
+      "loss": 0.3838,
+      "step": 51520
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6974291807297703e-06,
+      "loss": 0.2433,
+      "step": 51521
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6970933454234045e-06,
+      "loss": 0.2476,
+      "step": 51522
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.696757510117039e-06,
+      "loss": 0.2011,
+      "step": 51523
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.696421674810673e-06,
+      "loss": 0.3161,
+      "step": 51524
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.696085839504307e-06,
+      "loss": 0.19,
+      "step": 51525
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.695750004197942e-06,
+      "loss": 0.2791,
+      "step": 51526
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6954141688915757e-06,
+      "loss": 0.3,
+      "step": 51527
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.69507833358521e-06,
+      "loss": 0.2958,
+      "step": 51528
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6947424982788445e-06,
+      "loss": 0.2134,
+      "step": 51529
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6944066629724788e-06,
+      "loss": 0.2933,
+      "step": 51530
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6940708276661126e-06,
+      "loss": 0.3068,
+      "step": 51531
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.693734992359747e-06,
+      "loss": 0.1082,
+      "step": 51532
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6933991570533814e-06,
+      "loss": 0.234,
+      "step": 51533
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6930633217470152e-06,
+      "loss": 0.2297,
+      "step": 51534
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.69272748644065e-06,
+      "loss": 0.2532,
+      "step": 51535
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.692391651134284e-06,
+      "loss": 0.3355,
+      "step": 51536
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.692055815827918e-06,
+      "loss": 0.2716,
+      "step": 51537
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6917199805215525e-06,
+      "loss": 0.2149,
+      "step": 51538
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6913841452151868e-06,
+      "loss": 0.3038,
+      "step": 51539
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.691048309908821e-06,
+      "loss": 0.1655,
+      "step": 51540
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6907124746024556e-06,
+      "loss": 0.2823,
+      "step": 51541
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6903766392960894e-06,
+      "loss": 0.3085,
+      "step": 51542
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6900408039897237e-06,
+      "loss": 0.2317,
+      "step": 51543
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6897049686833583e-06,
+      "loss": 0.3862,
+      "step": 51544
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.689369133376992e-06,
+      "loss": 0.2728,
+      "step": 51545
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6890332980706263e-06,
+      "loss": 0.3084,
+      "step": 51546
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6886974627642606e-06,
+      "loss": 0.3935,
+      "step": 51547
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6883616274578948e-06,
+      "loss": 0.487,
+      "step": 51548
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.688025792151529e-06,
+      "loss": 0.2259,
+      "step": 51549
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6876899568451632e-06,
+      "loss": 0.2598,
+      "step": 51550
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.687354121538798e-06,
+      "loss": 0.2465,
+      "step": 51551
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6870182862324317e-06,
+      "loss": 0.2979,
+      "step": 51552
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.686682450926066e-06,
+      "loss": 0.1958,
+      "step": 51553
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6863466156197005e-06,
+      "loss": 0.3245,
+      "step": 51554
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6860107803133343e-06,
+      "loss": 0.2653,
+      "step": 51555
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6856749450069686e-06,
+      "loss": 0.296,
+      "step": 51556
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6853391097006032e-06,
+      "loss": 0.2613,
+      "step": 51557
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6850032743942374e-06,
+      "loss": 0.3754,
+      "step": 51558
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6846674390878712e-06,
+      "loss": 0.2549,
+      "step": 51559
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.684331603781506e-06,
+      "loss": 0.2046,
+      "step": 51560
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.68399576847514e-06,
+      "loss": 0.4492,
+      "step": 51561
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.683659933168774e-06,
+      "loss": 0.1499,
+      "step": 51562
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6833240978624086e-06,
+      "loss": 0.217,
+      "step": 51563
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6829882625560428e-06,
+      "loss": 0.3066,
+      "step": 51564
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6826524272496766e-06,
+      "loss": 0.3075,
+      "step": 51565
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6823165919433112e-06,
+      "loss": 0.2689,
+      "step": 51566
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6819807566369455e-06,
+      "loss": 0.2686,
+      "step": 51567
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6816449213305797e-06,
+      "loss": 0.3046,
+      "step": 51568
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6813090860242143e-06,
+      "loss": 0.3331,
+      "step": 51569
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.680973250717848e-06,
+      "loss": 0.2397,
+      "step": 51570
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6806374154114824e-06,
+      "loss": 0.372,
+      "step": 51571
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.680301580105117e-06,
+      "loss": 0.4221,
+      "step": 51572
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.679965744798751e-06,
+      "loss": 0.3586,
+      "step": 51573
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.679629909492385e-06,
+      "loss": 0.3243,
+      "step": 51574
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6792940741860197e-06,
+      "loss": 0.3017,
+      "step": 51575
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6789582388796535e-06,
+      "loss": 0.1562,
+      "step": 51576
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6786224035732877e-06,
+      "loss": 0.286,
+      "step": 51577
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6782865682669223e-06,
+      "loss": 0.2696,
+      "step": 51578
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6779507329605566e-06,
+      "loss": 0.2194,
+      "step": 51579
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6776148976541904e-06,
+      "loss": 0.2857,
+      "step": 51580
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.677279062347825e-06,
+      "loss": 0.1831,
+      "step": 51581
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6769432270414592e-06,
+      "loss": 0.2213,
+      "step": 51582
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.676607391735093e-06,
+      "loss": 0.2929,
+      "step": 51583
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6762715564287277e-06,
+      "loss": 0.3311,
+      "step": 51584
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.675935721122362e-06,
+      "loss": 0.2934,
+      "step": 51585
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.675599885815996e-06,
+      "loss": 0.2628,
+      "step": 51586
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6752640505096304e-06,
+      "loss": 0.3135,
+      "step": 51587
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6749282152032646e-06,
+      "loss": 0.3002,
+      "step": 51588
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.674592379896899e-06,
+      "loss": 0.2479,
+      "step": 51589
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6742565445905335e-06,
+      "loss": 0.2218,
+      "step": 51590
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6739207092841673e-06,
+      "loss": 0.2256,
+      "step": 51591
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6735848739778015e-06,
+      "loss": 0.1853,
+      "step": 51592
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6732490386714353e-06,
+      "loss": 0.2134,
+      "step": 51593
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.67291320336507e-06,
+      "loss": 0.3804,
+      "step": 51594
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.672577368058704e-06,
+      "loss": 0.3246,
+      "step": 51595
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6722415327523384e-06,
+      "loss": 0.17,
+      "step": 51596
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.671905697445973e-06,
+      "loss": 0.3108,
+      "step": 51597
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.671569862139607e-06,
+      "loss": 0.2285,
+      "step": 51598
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.671234026833241e-06,
+      "loss": 0.1919,
+      "step": 51599
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6708981915268757e-06,
+      "loss": 0.3055,
+      "step": 51600
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6705623562205095e-06,
+      "loss": 0.2682,
+      "step": 51601
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6702265209141437e-06,
+      "loss": 0.1457,
+      "step": 51602
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6698906856077784e-06,
+      "loss": 0.3107,
+      "step": 51603
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.669554850301412e-06,
+      "loss": 0.1788,
+      "step": 51604
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6692190149950464e-06,
+      "loss": 0.2859,
+      "step": 51605
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.668883179688681e-06,
+      "loss": 0.2138,
+      "step": 51606
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6685473443823153e-06,
+      "loss": 0.4567,
+      "step": 51607
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.668211509075949e-06,
+      "loss": 0.2484,
+      "step": 51608
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6678756737695837e-06,
+      "loss": 0.2915,
+      "step": 51609
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.667539838463218e-06,
+      "loss": 0.1797,
+      "step": 51610
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6672040031568517e-06,
+      "loss": 0.3409,
+      "step": 51611
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6668681678504864e-06,
+      "loss": 0.2631,
+      "step": 51612
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6665323325441206e-06,
+      "loss": 0.2848,
+      "step": 51613
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.666196497237755e-06,
+      "loss": 0.2024,
+      "step": 51614
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.665860661931389e-06,
+      "loss": 0.2735,
+      "step": 51615
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6655248266250233e-06,
+      "loss": 0.1498,
+      "step": 51616
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6651889913186575e-06,
+      "loss": 0.3276,
+      "step": 51617
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.664853156012292e-06,
+      "loss": 0.1875,
+      "step": 51618
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.664517320705926e-06,
+      "loss": 0.2866,
+      "step": 51619
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.66418148539956e-06,
+      "loss": 0.2226,
+      "step": 51620
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.663845650093195e-06,
+      "loss": 0.348,
+      "step": 51621
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6635098147868286e-06,
+      "loss": 0.4076,
+      "step": 51622
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.663173979480463e-06,
+      "loss": 0.2704,
+      "step": 51623
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6628381441740975e-06,
+      "loss": 0.2227,
+      "step": 51624
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6625023088677317e-06,
+      "loss": 0.2682,
+      "step": 51625
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6621664735613655e-06,
+      "loss": 0.1913,
+      "step": 51626
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.661830638255e-06,
+      "loss": 0.2851,
+      "step": 51627
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6614948029486344e-06,
+      "loss": 0.181,
+      "step": 51628
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.661158967642268e-06,
+      "loss": 0.1606,
+      "step": 51629
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.660823132335903e-06,
+      "loss": 0.3837,
+      "step": 51630
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.660487297029537e-06,
+      "loss": 0.2865,
+      "step": 51631
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6601514617231713e-06,
+      "loss": 0.2557,
+      "step": 51632
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6598156264168055e-06,
+      "loss": 0.3347,
+      "step": 51633
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6594797911104397e-06,
+      "loss": 0.2762,
+      "step": 51634
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.659143955804074e-06,
+      "loss": 0.0879,
+      "step": 51635
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6588081204977078e-06,
+      "loss": 0.2173,
+      "step": 51636
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6584722851913424e-06,
+      "loss": 0.244,
+      "step": 51637
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6581364498849766e-06,
+      "loss": 0.2455,
+      "step": 51638
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6578006145786104e-06,
+      "loss": 0.2318,
+      "step": 51639
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.657464779272245e-06,
+      "loss": 0.2215,
+      "step": 51640
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6571289439658793e-06,
+      "loss": 0.2803,
+      "step": 51641
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6567931086595135e-06,
+      "loss": 0.2127,
+      "step": 51642
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.656457273353148e-06,
+      "loss": 0.1676,
+      "step": 51643
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.656121438046782e-06,
+      "loss": 0.3499,
+      "step": 51644
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.655785602740416e-06,
+      "loss": 0.1806,
+      "step": 51645
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.655449767434051e-06,
+      "loss": 0.2257,
+      "step": 51646
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6551139321276846e-06,
+      "loss": 0.3408,
+      "step": 51647
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.654778096821319e-06,
+      "loss": 0.2975,
+      "step": 51648
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6544422615149535e-06,
+      "loss": 0.2675,
+      "step": 51649
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6541064262085873e-06,
+      "loss": 0.2076,
+      "step": 51650
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6537705909022215e-06,
+      "loss": 0.308,
+      "step": 51651
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.653434755595856e-06,
+      "loss": 0.1447,
+      "step": 51652
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6530989202894904e-06,
+      "loss": 0.2163,
+      "step": 51653
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.652763084983124e-06,
+      "loss": 0.2081,
+      "step": 51654
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.652427249676759e-06,
+      "loss": 0.4074,
+      "step": 51655
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.652091414370393e-06,
+      "loss": 0.4849,
+      "step": 51656
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.651755579064027e-06,
+      "loss": 0.2503,
+      "step": 51657
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6514197437576615e-06,
+      "loss": 0.3973,
+      "step": 51658
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6510839084512957e-06,
+      "loss": 0.1965,
+      "step": 51659
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.65074807314493e-06,
+      "loss": 0.2186,
+      "step": 51660
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.650412237838564e-06,
+      "loss": 0.2625,
+      "step": 51661
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6500764025321984e-06,
+      "loss": 0.2682,
+      "step": 51662
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6497405672258326e-06,
+      "loss": 0.3154,
+      "step": 51663
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6494047319194673e-06,
+      "loss": 0.3365,
+      "step": 51664
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.649068896613101e-06,
+      "loss": 0.2242,
+      "step": 51665
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6487330613067353e-06,
+      "loss": 0.1653,
+      "step": 51666
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.64839722600037e-06,
+      "loss": 0.2554,
+      "step": 51667
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6480613906940038e-06,
+      "loss": 0.3538,
+      "step": 51668
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.647725555387638e-06,
+      "loss": 0.3411,
+      "step": 51669
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6473897200812726e-06,
+      "loss": 0.2724,
+      "step": 51670
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.647053884774907e-06,
+      "loss": 0.1725,
+      "step": 51671
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6467180494685407e-06,
+      "loss": 0.1698,
+      "step": 51672
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6463822141621753e-06,
+      "loss": 0.2313,
+      "step": 51673
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6460463788558095e-06,
+      "loss": 0.4556,
+      "step": 51674
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6457105435494433e-06,
+      "loss": 0.2527,
+      "step": 51675
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.645374708243078e-06,
+      "loss": 0.1218,
+      "step": 51676
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.645038872936712e-06,
+      "loss": 0.2847,
+      "step": 51677
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.644703037630346e-06,
+      "loss": 0.3717,
+      "step": 51678
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6443672023239806e-06,
+      "loss": 0.2606,
+      "step": 51679
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.644031367017615e-06,
+      "loss": 0.2596,
+      "step": 51680
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.643695531711249e-06,
+      "loss": 0.2706,
+      "step": 51681
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.643359696404883e-06,
+      "loss": 0.2022,
+      "step": 51682
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6430238610985175e-06,
+      "loss": 0.3047,
+      "step": 51683
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6426880257921518e-06,
+      "loss": 0.2296,
+      "step": 51684
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6423521904857856e-06,
+      "loss": 0.1983,
+      "step": 51685
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6420163551794202e-06,
+      "loss": 0.2065,
+      "step": 51686
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6416805198730544e-06,
+      "loss": 0.3724,
+      "step": 51687
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6413446845666887e-06,
+      "loss": 0.2188,
+      "step": 51688
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.641008849260323e-06,
+      "loss": 0.1256,
+      "step": 51689
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.640673013953957e-06,
+      "loss": 0.3568,
+      "step": 51690
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6403371786475913e-06,
+      "loss": 0.2441,
+      "step": 51691
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.640001343341226e-06,
+      "loss": 0.1413,
+      "step": 51692
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6396655080348598e-06,
+      "loss": 0.154,
+      "step": 51693
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.639329672728494e-06,
+      "loss": 0.257,
+      "step": 51694
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6389938374221287e-06,
+      "loss": 0.1671,
+      "step": 51695
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6386580021157625e-06,
+      "loss": 0.1834,
+      "step": 51696
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6383221668093967e-06,
+      "loss": 0.105,
+      "step": 51697
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6379863315030313e-06,
+      "loss": 0.3054,
+      "step": 51698
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6376504961966655e-06,
+      "loss": 0.2579,
+      "step": 51699
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6373146608902993e-06,
+      "loss": 0.2677,
+      "step": 51700
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.636978825583934e-06,
+      "loss": 0.2758,
+      "step": 51701
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6366429902775682e-06,
+      "loss": 0.3024,
+      "step": 51702
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.636307154971202e-06,
+      "loss": 0.432,
+      "step": 51703
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6359713196648367e-06,
+      "loss": 0.3373,
+      "step": 51704
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.635635484358471e-06,
+      "loss": 0.2562,
+      "step": 51705
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6352996490521047e-06,
+      "loss": 0.4338,
+      "step": 51706
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6349638137457393e-06,
+      "loss": 0.3642,
+      "step": 51707
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6346279784393736e-06,
+      "loss": 0.4273,
+      "step": 51708
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6342921431330078e-06,
+      "loss": 0.4105,
+      "step": 51709
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6339563078266424e-06,
+      "loss": 0.262,
+      "step": 51710
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2.6336204725202762e-06,
+      "loss": 0.2779,
+      "step": 51711
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6332846372139105e-06,
+      "loss": 0.3953,
+      "step": 51712
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.632948801907545e-06,
+      "loss": 0.2579,
+      "step": 51713
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.632612966601179e-06,
+      "loss": 0.1443,
+      "step": 51714
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.632277131294813e-06,
+      "loss": 0.2524,
+      "step": 51715
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6319412959884478e-06,
+      "loss": 0.3158,
+      "step": 51716
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6316054606820816e-06,
+      "loss": 0.3403,
+      "step": 51717
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.631269625375716e-06,
+      "loss": 0.2604,
+      "step": 51718
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6309337900693505e-06,
+      "loss": 0.2669,
+      "step": 51719
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6305979547629847e-06,
+      "loss": 0.2953,
+      "step": 51720
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6302621194566185e-06,
+      "loss": 0.289,
+      "step": 51721
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.629926284150253e-06,
+      "loss": 0.2256,
+      "step": 51722
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6295904488438873e-06,
+      "loss": 0.4417,
+      "step": 51723
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.629254613537521e-06,
+      "loss": 0.322,
+      "step": 51724
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.628918778231156e-06,
+      "loss": 0.2135,
+      "step": 51725
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.62858294292479e-06,
+      "loss": 0.2495,
+      "step": 51726
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6282471076184242e-06,
+      "loss": 0.2376,
+      "step": 51727
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.627911272312058e-06,
+      "loss": 0.1901,
+      "step": 51728
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6275754370056927e-06,
+      "loss": 0.1776,
+      "step": 51729
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.627239601699327e-06,
+      "loss": 0.3983,
+      "step": 51730
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6269037663929607e-06,
+      "loss": 0.2042,
+      "step": 51731
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6265679310865954e-06,
+      "loss": 0.3921,
+      "step": 51732
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6262320957802296e-06,
+      "loss": 0.1305,
+      "step": 51733
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6258962604738634e-06,
+      "loss": 0.282,
+      "step": 51734
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.625560425167498e-06,
+      "loss": 0.1363,
+      "step": 51735
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6252245898611323e-06,
+      "loss": 0.2725,
+      "step": 51736
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6248887545547665e-06,
+      "loss": 0.3429,
+      "step": 51737
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.624552919248401e-06,
+      "loss": 0.3018,
+      "step": 51738
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.624217083942035e-06,
+      "loss": 0.4753,
+      "step": 51739
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.623881248635669e-06,
+      "loss": 0.2903,
+      "step": 51740
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.623545413329304e-06,
+      "loss": 0.2354,
+      "step": 51741
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6232095780229376e-06,
+      "loss": 0.1688,
+      "step": 51742
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.622873742716572e-06,
+      "loss": 0.2438,
+      "step": 51743
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6225379074102065e-06,
+      "loss": 0.215,
+      "step": 51744
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6222020721038403e-06,
+      "loss": 0.271,
+      "step": 51745
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6218662367974745e-06,
+      "loss": 0.1935,
+      "step": 51746
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.621530401491109e-06,
+      "loss": 0.274,
+      "step": 51747
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6211945661847434e-06,
+      "loss": 0.155,
+      "step": 51748
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.620858730878377e-06,
+      "loss": 0.2118,
+      "step": 51749
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.620522895572012e-06,
+      "loss": 0.4023,
+      "step": 51750
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.620187060265646e-06,
+      "loss": 0.2161,
+      "step": 51751
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.61985122495928e-06,
+      "loss": 0.2267,
+      "step": 51752
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6195153896529145e-06,
+      "loss": 0.1741,
+      "step": 51753
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6191795543465487e-06,
+      "loss": 0.23,
+      "step": 51754
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.618843719040183e-06,
+      "loss": 0.2246,
+      "step": 51755
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.618507883733817e-06,
+      "loss": 0.2101,
+      "step": 51756
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6181720484274514e-06,
+      "loss": 0.2073,
+      "step": 51757
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6178362131210856e-06,
+      "loss": 0.1762,
+      "step": 51758
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6175003778147203e-06,
+      "loss": 0.1185,
+      "step": 51759
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.617164542508354e-06,
+      "loss": 0.1891,
+      "step": 51760
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6168287072019883e-06,
+      "loss": 0.2556,
+      "step": 51761
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.616492871895623e-06,
+      "loss": 0.2374,
+      "step": 51762
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6161570365892567e-06,
+      "loss": 0.195,
+      "step": 51763
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.615821201282891e-06,
+      "loss": 0.3349,
+      "step": 51764
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6154853659765256e-06,
+      "loss": 0.2934,
+      "step": 51765
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.61514953067016e-06,
+      "loss": 0.3437,
+      "step": 51766
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6148136953637936e-06,
+      "loss": 0.204,
+      "step": 51767
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6144778600574283e-06,
+      "loss": 0.2779,
+      "step": 51768
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6141420247510625e-06,
+      "loss": 0.3844,
+      "step": 51769
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6138061894446963e-06,
+      "loss": 0.2882,
+      "step": 51770
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6134703541383305e-06,
+      "loss": 0.3354,
+      "step": 51771
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.613134518831965e-06,
+      "loss": 0.1697,
+      "step": 51772
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6127986835255994e-06,
+      "loss": 0.2649,
+      "step": 51773
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.612462848219233e-06,
+      "loss": 0.3025,
+      "step": 51774
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.612127012912868e-06,
+      "loss": 0.1719,
+      "step": 51775
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.611791177606502e-06,
+      "loss": 0.2079,
+      "step": 51776
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.611455342300136e-06,
+      "loss": 0.3026,
+      "step": 51777
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6111195069937705e-06,
+      "loss": 0.3618,
+      "step": 51778
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6107836716874047e-06,
+      "loss": 0.2318,
+      "step": 51779
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6104478363810385e-06,
+      "loss": 0.3053,
+      "step": 51780
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.610112001074673e-06,
+      "loss": 0.329,
+      "step": 51781
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6097761657683074e-06,
+      "loss": 0.1984,
+      "step": 51782
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6094403304619416e-06,
+      "loss": 0.2971,
+      "step": 51783
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6091044951555763e-06,
+      "loss": 0.1976,
+      "step": 51784
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.60876865984921e-06,
+      "loss": 0.1934,
+      "step": 51785
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6084328245428443e-06,
+      "loss": 0.2148,
+      "step": 51786
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.608096989236479e-06,
+      "loss": 0.111,
+      "step": 51787
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6077611539301127e-06,
+      "loss": 0.2222,
+      "step": 51788
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.607425318623747e-06,
+      "loss": 0.279,
+      "step": 51789
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6070894833173816e-06,
+      "loss": 0.4907,
+      "step": 51790
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6067536480110154e-06,
+      "loss": 0.3577,
+      "step": 51791
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6064178127046496e-06,
+      "loss": 0.3498,
+      "step": 51792
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6060819773982843e-06,
+      "loss": 0.2603,
+      "step": 51793
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6057461420919185e-06,
+      "loss": 0.3147,
+      "step": 51794
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6054103067855523e-06,
+      "loss": 0.2008,
+      "step": 51795
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.605074471479187e-06,
+      "loss": 0.3686,
+      "step": 51796
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.604738636172821e-06,
+      "loss": 0.1462,
+      "step": 51797
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.604402800866455e-06,
+      "loss": 0.3273,
+      "step": 51798
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6040669655600896e-06,
+      "loss": 0.2059,
+      "step": 51799
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.603731130253724e-06,
+      "loss": 0.2902,
+      "step": 51800
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.603395294947358e-06,
+      "loss": 0.2788,
+      "step": 51801
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6030594596409923e-06,
+      "loss": 0.3498,
+      "step": 51802
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6027236243346265e-06,
+      "loss": 0.2485,
+      "step": 51803
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6023877890282607e-06,
+      "loss": 0.4432,
+      "step": 51804
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6020519537218954e-06,
+      "loss": 0.2804,
+      "step": 51805
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.601716118415529e-06,
+      "loss": 0.2369,
+      "step": 51806
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6013802831091634e-06,
+      "loss": 0.1703,
+      "step": 51807
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.601044447802798e-06,
+      "loss": 0.1354,
+      "step": 51808
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.600708612496432e-06,
+      "loss": 0.2661,
+      "step": 51809
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.600372777190066e-06,
+      "loss": 0.3486,
+      "step": 51810
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.6000369418837007e-06,
+      "loss": 0.2648,
+      "step": 51811
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.599701106577335e-06,
+      "loss": 0.3204,
+      "step": 51812
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5993652712709688e-06,
+      "loss": 0.224,
+      "step": 51813
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5990294359646034e-06,
+      "loss": 0.2234,
+      "step": 51814
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5986936006582376e-06,
+      "loss": 0.251,
+      "step": 51815
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5983577653518714e-06,
+      "loss": 0.3289,
+      "step": 51816
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5980219300455057e-06,
+      "loss": 0.193,
+      "step": 51817
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5976860947391403e-06,
+      "loss": 0.2228,
+      "step": 51818
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.597350259432774e-06,
+      "loss": 0.2474,
+      "step": 51819
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5970144241264083e-06,
+      "loss": 0.1712,
+      "step": 51820
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.596678588820043e-06,
+      "loss": 0.2497,
+      "step": 51821
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.596342753513677e-06,
+      "loss": 0.1577,
+      "step": 51822
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.596006918207311e-06,
+      "loss": 0.196,
+      "step": 51823
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5956710829009457e-06,
+      "loss": 0.2093,
+      "step": 51824
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.59533524759458e-06,
+      "loss": 0.1779,
+      "step": 51825
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5949994122882137e-06,
+      "loss": 0.1364,
+      "step": 51826
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5946635769818483e-06,
+      "loss": 0.254,
+      "step": 51827
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5943277416754825e-06,
+      "loss": 0.3249,
+      "step": 51828
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5939919063691168e-06,
+      "loss": 0.133,
+      "step": 51829
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.593656071062751e-06,
+      "loss": 0.2822,
+      "step": 51830
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5933202357563852e-06,
+      "loss": 0.1139,
+      "step": 51831
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5929844004500194e-06,
+      "loss": 0.2672,
+      "step": 51832
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.592648565143654e-06,
+      "loss": 0.3354,
+      "step": 51833
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.592312729837288e-06,
+      "loss": 0.3256,
+      "step": 51834
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.591976894530922e-06,
+      "loss": 0.2496,
+      "step": 51835
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5916410592245568e-06,
+      "loss": 0.2669,
+      "step": 51836
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5913052239181906e-06,
+      "loss": 0.1623,
+      "step": 51837
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5909693886118248e-06,
+      "loss": 0.3316,
+      "step": 51838
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5906335533054594e-06,
+      "loss": 0.2557,
+      "step": 51839
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5902977179990937e-06,
+      "loss": 0.2877,
+      "step": 51840
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5899618826927275e-06,
+      "loss": 0.3425,
+      "step": 51841
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.589626047386362e-06,
+      "loss": 0.1055,
+      "step": 51842
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5892902120799963e-06,
+      "loss": 0.1945,
+      "step": 51843
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.58895437677363e-06,
+      "loss": 0.2671,
+      "step": 51844
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5886185414672648e-06,
+      "loss": 0.1707,
+      "step": 51845
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.588282706160899e-06,
+      "loss": 0.231,
+      "step": 51846
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.587946870854533e-06,
+      "loss": 0.314,
+      "step": 51847
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5876110355481674e-06,
+      "loss": 0.2431,
+      "step": 51848
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5872752002418017e-06,
+      "loss": 0.2039,
+      "step": 51849
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.586939364935436e-06,
+      "loss": 0.2199,
+      "step": 51850
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5866035296290705e-06,
+      "loss": 0.2147,
+      "step": 51851
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5862676943227043e-06,
+      "loss": 0.2282,
+      "step": 51852
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5859318590163386e-06,
+      "loss": 0.1109,
+      "step": 51853
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5855960237099732e-06,
+      "loss": 0.2418,
+      "step": 51854
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.585260188403607e-06,
+      "loss": 0.1886,
+      "step": 51855
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5849243530972412e-06,
+      "loss": 0.3319,
+      "step": 51856
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.584588517790876e-06,
+      "loss": 0.2205,
+      "step": 51857
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5842526824845097e-06,
+      "loss": 0.2309,
+      "step": 51858
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.583916847178144e-06,
+      "loss": 0.2239,
+      "step": 51859
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5835810118717786e-06,
+      "loss": 0.2703,
+      "step": 51860
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5832451765654128e-06,
+      "loss": 0.2798,
+      "step": 51861
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5829093412590466e-06,
+      "loss": 0.2099,
+      "step": 51862
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.582573505952681e-06,
+      "loss": 0.4205,
+      "step": 51863
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5822376706463155e-06,
+      "loss": 0.1059,
+      "step": 51864
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5819018353399493e-06,
+      "loss": 0.2982,
+      "step": 51865
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5815660000335835e-06,
+      "loss": 0.2277,
+      "step": 51866
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.581230164727218e-06,
+      "loss": 0.2793,
+      "step": 51867
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5808943294208523e-06,
+      "loss": 0.1884,
+      "step": 51868
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.580558494114486e-06,
+      "loss": 0.2457,
+      "step": 51869
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.580222658808121e-06,
+      "loss": 0.3017,
+      "step": 51870
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.579886823501755e-06,
+      "loss": 0.2596,
+      "step": 51871
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.579550988195389e-06,
+      "loss": 0.119,
+      "step": 51872
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5792151528890235e-06,
+      "loss": 0.3989,
+      "step": 51873
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5788793175826577e-06,
+      "loss": 0.3212,
+      "step": 51874
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5785434822762915e-06,
+      "loss": 0.1725,
+      "step": 51875
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.578207646969926e-06,
+      "loss": 0.4196,
+      "step": 51876
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5778718116635604e-06,
+      "loss": 0.517,
+      "step": 51877
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5775359763571946e-06,
+      "loss": 0.1955,
+      "step": 51878
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5772001410508292e-06,
+      "loss": 0.3331,
+      "step": 51879
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.576864305744463e-06,
+      "loss": 0.1694,
+      "step": 51880
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5765284704380973e-06,
+      "loss": 0.2733,
+      "step": 51881
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.576192635131732e-06,
+      "loss": 0.1213,
+      "step": 51882
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5758567998253657e-06,
+      "loss": 0.2235,
+      "step": 51883
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.575520964519e-06,
+      "loss": 0.2835,
+      "step": 51884
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5751851292126346e-06,
+      "loss": 0.3416,
+      "step": 51885
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5748492939062684e-06,
+      "loss": 0.3324,
+      "step": 51886
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5745134585999026e-06,
+      "loss": 0.2143,
+      "step": 51887
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5741776232935372e-06,
+      "loss": 0.0986,
+      "step": 51888
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5738417879871715e-06,
+      "loss": 0.3023,
+      "step": 51889
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5735059526808053e-06,
+      "loss": 0.2938,
+      "step": 51890
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.57317011737444e-06,
+      "loss": 0.305,
+      "step": 51891
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.572834282068074e-06,
+      "loss": 0.2094,
+      "step": 51892
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.572498446761708e-06,
+      "loss": 0.2025,
+      "step": 51893
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5721626114553426e-06,
+      "loss": 0.1552,
+      "step": 51894
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.571826776148977e-06,
+      "loss": 0.1441,
+      "step": 51895
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.571490940842611e-06,
+      "loss": 0.2704,
+      "step": 51896
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5711551055362457e-06,
+      "loss": 0.4098,
+      "step": 51897
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5708192702298795e-06,
+      "loss": 0.2089,
+      "step": 51898
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5704834349235137e-06,
+      "loss": 0.3526,
+      "step": 51899
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5701475996171484e-06,
+      "loss": 0.2434,
+      "step": 51900
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.569811764310782e-06,
+      "loss": 0.3987,
+      "step": 51901
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5694759290044164e-06,
+      "loss": 0.2311,
+      "step": 51902
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.569140093698051e-06,
+      "loss": 0.24,
+      "step": 51903
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.568804258391685e-06,
+      "loss": 0.2523,
+      "step": 51904
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.568468423085319e-06,
+      "loss": 0.4039,
+      "step": 51905
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5681325877789533e-06,
+      "loss": 0.3305,
+      "step": 51906
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.567796752472588e-06,
+      "loss": 0.3544,
+      "step": 51907
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5674609171662217e-06,
+      "loss": 0.3885,
+      "step": 51908
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.567125081859856e-06,
+      "loss": 0.1812,
+      "step": 51909
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.5667892465534906e-06,
+      "loss": 0.3659,
+      "step": 51910
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5664534112471244e-06,
+      "loss": 0.6364,
+      "step": 51911
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5661175759407586e-06,
+      "loss": 0.2464,
+      "step": 51912
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5657817406343933e-06,
+      "loss": 0.169,
+      "step": 51913
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5654459053280275e-06,
+      "loss": 0.1787,
+      "step": 51914
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5651100700216613e-06,
+      "loss": 0.3095,
+      "step": 51915
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.564774234715296e-06,
+      "loss": 0.2311,
+      "step": 51916
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.56443839940893e-06,
+      "loss": 0.2301,
+      "step": 51917
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.564102564102564e-06,
+      "loss": 0.1974,
+      "step": 51918
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5637667287961986e-06,
+      "loss": 0.3463,
+      "step": 51919
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.563430893489833e-06,
+      "loss": 0.1665,
+      "step": 51920
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5630950581834666e-06,
+      "loss": 0.3213,
+      "step": 51921
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5627592228771013e-06,
+      "loss": 0.3142,
+      "step": 51922
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5624233875707355e-06,
+      "loss": 0.2728,
+      "step": 51923
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5620875522643697e-06,
+      "loss": 0.2126,
+      "step": 51924
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5617517169580044e-06,
+      "loss": 0.183,
+      "step": 51925
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.561415881651638e-06,
+      "loss": 0.3175,
+      "step": 51926
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5610800463452724e-06,
+      "loss": 0.378,
+      "step": 51927
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.560744211038907e-06,
+      "loss": 0.2265,
+      "step": 51928
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.560408375732541e-06,
+      "loss": 0.3149,
+      "step": 51929
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.560072540426175e-06,
+      "loss": 0.2023,
+      "step": 51930
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5597367051198097e-06,
+      "loss": 0.25,
+      "step": 51931
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5594008698134435e-06,
+      "loss": 0.1912,
+      "step": 51932
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5590650345070777e-06,
+      "loss": 0.2576,
+      "step": 51933
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5587291992007124e-06,
+      "loss": 0.2133,
+      "step": 51934
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5583933638943466e-06,
+      "loss": 0.1292,
+      "step": 51935
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5580575285879804e-06,
+      "loss": 0.2818,
+      "step": 51936
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.557721693281615e-06,
+      "loss": 0.2998,
+      "step": 51937
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5573858579752493e-06,
+      "loss": 0.2588,
+      "step": 51938
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.557050022668883e-06,
+      "loss": 0.3131,
+      "step": 51939
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5567141873625177e-06,
+      "loss": 0.4167,
+      "step": 51940
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.556378352056152e-06,
+      "loss": 0.3638,
+      "step": 51941
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.556042516749786e-06,
+      "loss": 0.1561,
+      "step": 51942
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5557066814434204e-06,
+      "loss": 0.3282,
+      "step": 51943
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5553708461370546e-06,
+      "loss": 0.1909,
+      "step": 51944
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.555035010830689e-06,
+      "loss": 0.4106,
+      "step": 51945
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5546991755243235e-06,
+      "loss": 0.2041,
+      "step": 51946
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5543633402179573e-06,
+      "loss": 0.2833,
+      "step": 51947
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5540275049115915e-06,
+      "loss": 0.2638,
+      "step": 51948
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.553691669605226e-06,
+      "loss": 0.3657,
+      "step": 51949
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.55335583429886e-06,
+      "loss": 0.3882,
+      "step": 51950
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.553019998992494e-06,
+      "loss": 0.1789,
+      "step": 51951
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5526841636861284e-06,
+      "loss": 0.1395,
+      "step": 51952
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.552348328379763e-06,
+      "loss": 0.2616,
+      "step": 51953
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.552012493073397e-06,
+      "loss": 0.2393,
+      "step": 51954
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.551676657767031e-06,
+      "loss": 0.2681,
+      "step": 51955
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5513408224606657e-06,
+      "loss": 0.2338,
+      "step": 51956
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5510049871542995e-06,
+      "loss": 0.2473,
+      "step": 51957
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5506691518479338e-06,
+      "loss": 0.2737,
+      "step": 51958
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5503333165415684e-06,
+      "loss": 0.2645,
+      "step": 51959
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5499974812352022e-06,
+      "loss": 0.3412,
+      "step": 51960
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5496616459288364e-06,
+      "loss": 0.3275,
+      "step": 51961
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.549325810622471e-06,
+      "loss": 0.4533,
+      "step": 51962
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5489899753161053e-06,
+      "loss": 0.2822,
+      "step": 51963
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.548654140009739e-06,
+      "loss": 0.1764,
+      "step": 51964
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5483183047033738e-06,
+      "loss": 0.2944,
+      "step": 51965
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.547982469397008e-06,
+      "loss": 0.1949,
+      "step": 51966
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5476466340906418e-06,
+      "loss": 0.2102,
+      "step": 51967
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5473107987842764e-06,
+      "loss": 0.2332,
+      "step": 51968
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5469749634779107e-06,
+      "loss": 0.2411,
+      "step": 51969
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.546639128171545e-06,
+      "loss": 0.2609,
+      "step": 51970
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.546303292865179e-06,
+      "loss": 0.1738,
+      "step": 51971
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5459674575588133e-06,
+      "loss": 0.2626,
+      "step": 51972
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5456316222524475e-06,
+      "loss": 0.2661,
+      "step": 51973
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.545295786946082e-06,
+      "loss": 0.39,
+      "step": 51974
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.544959951639716e-06,
+      "loss": 0.1672,
+      "step": 51975
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5446241163333502e-06,
+      "loss": 0.1357,
+      "step": 51976
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.544288281026985e-06,
+      "loss": 0.3279,
+      "step": 51977
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5439524457206187e-06,
+      "loss": 0.2848,
+      "step": 51978
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.543616610414253e-06,
+      "loss": 0.2305,
+      "step": 51979
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5432807751078875e-06,
+      "loss": 0.3087,
+      "step": 51980
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5429449398015218e-06,
+      "loss": 0.2877,
+      "step": 51981
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5426091044951556e-06,
+      "loss": 0.3651,
+      "step": 51982
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.54227326918879e-06,
+      "loss": 0.2318,
+      "step": 51983
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5419374338824244e-06,
+      "loss": 0.2992,
+      "step": 51984
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5416015985760582e-06,
+      "loss": 0.153,
+      "step": 51985
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.541265763269693e-06,
+      "loss": 0.2516,
+      "step": 51986
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.540929927963327e-06,
+      "loss": 0.292,
+      "step": 51987
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.540594092656961e-06,
+      "loss": 0.3221,
+      "step": 51988
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5402582573505956e-06,
+      "loss": 0.2069,
+      "step": 51989
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5399224220442298e-06,
+      "loss": 0.2331,
+      "step": 51990
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.539586586737864e-06,
+      "loss": 0.2351,
+      "step": 51991
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5392507514314986e-06,
+      "loss": 0.4375,
+      "step": 51992
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5389149161251324e-06,
+      "loss": 0.3901,
+      "step": 51993
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5385790808187667e-06,
+      "loss": 0.2614,
+      "step": 51994
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5382432455124013e-06,
+      "loss": 0.3864,
+      "step": 51995
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.537907410206035e-06,
+      "loss": 0.238,
+      "step": 51996
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5375715748996693e-06,
+      "loss": 0.1744,
+      "step": 51997
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5372357395933036e-06,
+      "loss": 0.263,
+      "step": 51998
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.536899904286938e-06,
+      "loss": 0.2608,
+      "step": 51999
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.536564068980572e-06,
+      "loss": 0.1548,
+      "step": 52000
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5362282336742062e-06,
+      "loss": 0.2614,
+      "step": 52001
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.535892398367841e-06,
+      "loss": 0.329,
+      "step": 52002
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5355565630614747e-06,
+      "loss": 0.2042,
+      "step": 52003
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.535220727755109e-06,
+      "loss": 0.1228,
+      "step": 52004
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5348848924487436e-06,
+      "loss": 0.3202,
+      "step": 52005
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5345490571423774e-06,
+      "loss": 0.3621,
+      "step": 52006
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5342132218360116e-06,
+      "loss": 0.2354,
+      "step": 52007
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5338773865296462e-06,
+      "loss": 0.2047,
+      "step": 52008
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5335415512232805e-06,
+      "loss": 0.3129,
+      "step": 52009
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5332057159169143e-06,
+      "loss": 0.2375,
+      "step": 52010
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.532869880610549e-06,
+      "loss": 0.3482,
+      "step": 52011
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.532534045304183e-06,
+      "loss": 0.2292,
+      "step": 52012
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.532198209997817e-06,
+      "loss": 0.1875,
+      "step": 52013
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5318623746914516e-06,
+      "loss": 0.3515,
+      "step": 52014
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.531526539385086e-06,
+      "loss": 0.3207,
+      "step": 52015
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5311907040787196e-06,
+      "loss": 0.2057,
+      "step": 52016
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5308548687723542e-06,
+      "loss": 0.1854,
+      "step": 52017
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5305190334659885e-06,
+      "loss": 0.2186,
+      "step": 52018
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5301831981596227e-06,
+      "loss": 0.2194,
+      "step": 52019
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5298473628532573e-06,
+      "loss": 0.341,
+      "step": 52020
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.529511527546891e-06,
+      "loss": 0.3368,
+      "step": 52021
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5291756922405254e-06,
+      "loss": 0.1996,
+      "step": 52022
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.52883985693416e-06,
+      "loss": 0.401,
+      "step": 52023
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.528504021627794e-06,
+      "loss": 0.2616,
+      "step": 52024
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.528168186321428e-06,
+      "loss": 0.3343,
+      "step": 52025
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5278323510150627e-06,
+      "loss": 0.2527,
+      "step": 52026
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.527496515708697e-06,
+      "loss": 0.1864,
+      "step": 52027
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5271606804023307e-06,
+      "loss": 0.2508,
+      "step": 52028
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5268248450959654e-06,
+      "loss": 0.1599,
+      "step": 52029
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5264890097895996e-06,
+      "loss": 0.2477,
+      "step": 52030
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5261531744832334e-06,
+      "loss": 0.257,
+      "step": 52031
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.525817339176868e-06,
+      "loss": 0.417,
+      "step": 52032
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5254815038705023e-06,
+      "loss": 0.2557,
+      "step": 52033
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.525145668564136e-06,
+      "loss": 0.3543,
+      "step": 52034
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5248098332577707e-06,
+      "loss": 0.2404,
+      "step": 52035
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.524473997951405e-06,
+      "loss": 0.3048,
+      "step": 52036
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.524138162645039e-06,
+      "loss": 0.2387,
+      "step": 52037
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.523802327338674e-06,
+      "loss": 0.1818,
+      "step": 52038
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5234664920323076e-06,
+      "loss": 0.1927,
+      "step": 52039
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.523130656725942e-06,
+      "loss": 0.1463,
+      "step": 52040
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5227948214195765e-06,
+      "loss": 0.2163,
+      "step": 52041
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5224589861132103e-06,
+      "loss": 0.1576,
+      "step": 52042
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5221231508068445e-06,
+      "loss": 0.1721,
+      "step": 52043
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5217873155004787e-06,
+      "loss": 0.2863,
+      "step": 52044
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.521451480194113e-06,
+      "loss": 0.3188,
+      "step": 52045
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.521115644887747e-06,
+      "loss": 0.2008,
+      "step": 52046
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5207798095813814e-06,
+      "loss": 0.2277,
+      "step": 52047
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.520443974275016e-06,
+      "loss": 0.3774,
+      "step": 52048
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.52010813896865e-06,
+      "loss": 0.2332,
+      "step": 52049
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.519772303662284e-06,
+      "loss": 0.1912,
+      "step": 52050
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5194364683559187e-06,
+      "loss": 0.3386,
+      "step": 52051
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5191006330495525e-06,
+      "loss": 0.1822,
+      "step": 52052
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5187647977431867e-06,
+      "loss": 0.3415,
+      "step": 52053
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5184289624368214e-06,
+      "loss": 0.2802,
+      "step": 52054
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5180931271304556e-06,
+      "loss": 0.2291,
+      "step": 52055
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5177572918240894e-06,
+      "loss": 0.2065,
+      "step": 52056
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.517421456517724e-06,
+      "loss": 0.2992,
+      "step": 52057
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5170856212113583e-06,
+      "loss": 0.1431,
+      "step": 52058
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.516749785904992e-06,
+      "loss": 0.3151,
+      "step": 52059
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5164139505986267e-06,
+      "loss": 0.1487,
+      "step": 52060
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.516078115292261e-06,
+      "loss": 0.137,
+      "step": 52061
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5157422799858947e-06,
+      "loss": 0.1572,
+      "step": 52062
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5154064446795294e-06,
+      "loss": 0.2909,
+      "step": 52063
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5150706093731636e-06,
+      "loss": 0.2212,
+      "step": 52064
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.514734774066798e-06,
+      "loss": 0.2622,
+      "step": 52065
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5143989387604325e-06,
+      "loss": 0.2132,
+      "step": 52066
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5140631034540663e-06,
+      "loss": 0.1777,
+      "step": 52067
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5137272681477005e-06,
+      "loss": 0.2744,
+      "step": 52068
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.513391432841335e-06,
+      "loss": 0.2127,
+      "step": 52069
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.513055597534969e-06,
+      "loss": 0.266,
+      "step": 52070
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.512719762228603e-06,
+      "loss": 0.2596,
+      "step": 52071
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.512383926922238e-06,
+      "loss": 0.1406,
+      "step": 52072
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5120480916158716e-06,
+      "loss": 0.252,
+      "step": 52073
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.511712256309506e-06,
+      "loss": 0.1338,
+      "step": 52074
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5113764210031405e-06,
+      "loss": 0.2025,
+      "step": 52075
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5110405856967747e-06,
+      "loss": 0.4123,
+      "step": 52076
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5107047503904085e-06,
+      "loss": 0.16,
+      "step": 52077
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.510368915084043e-06,
+      "loss": 0.1872,
+      "step": 52078
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5100330797776774e-06,
+      "loss": 0.2132,
+      "step": 52079
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.509697244471311e-06,
+      "loss": 0.201,
+      "step": 52080
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.509361409164946e-06,
+      "loss": 0.1899,
+      "step": 52081
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.50902557385858e-06,
+      "loss": 0.3077,
+      "step": 52082
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5086897385522143e-06,
+      "loss": 0.3397,
+      "step": 52083
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5083539032458485e-06,
+      "loss": 0.2058,
+      "step": 52084
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5080180679394827e-06,
+      "loss": 0.1161,
+      "step": 52085
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.507682232633117e-06,
+      "loss": 0.3707,
+      "step": 52086
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5073463973267508e-06,
+      "loss": 0.3005,
+      "step": 52087
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5070105620203854e-06,
+      "loss": 0.2485,
+      "step": 52088
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5066747267140196e-06,
+      "loss": 0.2662,
+      "step": 52089
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5063388914076534e-06,
+      "loss": 0.2239,
+      "step": 52090
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.506003056101288e-06,
+      "loss": 0.254,
+      "step": 52091
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5056672207949223e-06,
+      "loss": 0.2136,
+      "step": 52092
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5053313854885565e-06,
+      "loss": 0.2129,
+      "step": 52093
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.504995550182191e-06,
+      "loss": 0.3979,
+      "step": 52094
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.504659714875825e-06,
+      "loss": 0.2072,
+      "step": 52095
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.504323879569459e-06,
+      "loss": 0.2541,
+      "step": 52096
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.503988044263094e-06,
+      "loss": 0.2478,
+      "step": 52097
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5036522089567276e-06,
+      "loss": 0.1492,
+      "step": 52098
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.503316373650362e-06,
+      "loss": 0.2023,
+      "step": 52099
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5029805383439965e-06,
+      "loss": 0.27,
+      "step": 52100
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5026447030376303e-06,
+      "loss": 0.2876,
+      "step": 52101
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5023088677312645e-06,
+      "loss": 0.318,
+      "step": 52102
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.501973032424899e-06,
+      "loss": 0.2582,
+      "step": 52103
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5016371971185334e-06,
+      "loss": 0.3075,
+      "step": 52104
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.5013013618121672e-06,
+      "loss": 0.3686,
+      "step": 52105
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.500965526505802e-06,
+      "loss": 0.1763,
+      "step": 52106
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.500629691199436e-06,
+      "loss": 0.2049,
+      "step": 52107
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.50029385589307e-06,
+      "loss": 0.2663,
+      "step": 52108
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4999580205867045e-06,
+      "loss": 0.3549,
+      "step": 52109
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4996221852803388e-06,
+      "loss": 0.3048,
+      "step": 52110
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.499286349973973e-06,
+      "loss": 0.3098,
+      "step": 52111
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.498950514667607e-06,
+      "loss": 0.2101,
+      "step": 52112
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4986146793612414e-06,
+      "loss": 0.3307,
+      "step": 52113
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4982788440548757e-06,
+      "loss": 0.2163,
+      "step": 52114
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.49794300874851e-06,
+      "loss": 0.2971,
+      "step": 52115
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.497607173442144e-06,
+      "loss": 0.3087,
+      "step": 52116
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4972713381357787e-06,
+      "loss": 0.3428,
+      "step": 52117
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4969355028294126e-06,
+      "loss": 0.3257,
+      "step": 52118
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4965996675230468e-06,
+      "loss": 0.2208,
+      "step": 52119
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4962638322166814e-06,
+      "loss": 0.5026,
+      "step": 52120
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4959279969103152e-06,
+      "loss": 0.2731,
+      "step": 52121
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.49559216160395e-06,
+      "loss": 0.4621,
+      "step": 52122
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4952563262975837e-06,
+      "loss": 0.3722,
+      "step": 52123
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.494920490991218e-06,
+      "loss": 0.1342,
+      "step": 52124
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4945846556848525e-06,
+      "loss": 0.2964,
+      "step": 52125
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4942488203784863e-06,
+      "loss": 0.1996,
+      "step": 52126
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.493912985072121e-06,
+      "loss": 0.2603,
+      "step": 52127
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4935771497657552e-06,
+      "loss": 0.2799,
+      "step": 52128
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.493241314459389e-06,
+      "loss": 0.2988,
+      "step": 52129
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4929054791530237e-06,
+      "loss": 0.3663,
+      "step": 52130
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.492569643846658e-06,
+      "loss": 0.1295,
+      "step": 52131
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.492233808540292e-06,
+      "loss": 0.1636,
+      "step": 52132
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4918979732339263e-06,
+      "loss": 0.2316,
+      "step": 52133
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4915621379275606e-06,
+      "loss": 0.1641,
+      "step": 52134
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4912263026211948e-06,
+      "loss": 0.3894,
+      "step": 52135
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.490890467314829e-06,
+      "loss": 0.2952,
+      "step": 52136
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4905546320084632e-06,
+      "loss": 0.2968,
+      "step": 52137
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4902187967020975e-06,
+      "loss": 0.2286,
+      "step": 52138
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4898829613957317e-06,
+      "loss": 0.3375,
+      "step": 52139
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.489547126089366e-06,
+      "loss": 0.2094,
+      "step": 52140
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.489211290783e-06,
+      "loss": 0.3436,
+      "step": 52141
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4888754554766343e-06,
+      "loss": 0.5167,
+      "step": 52142
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4885396201702686e-06,
+      "loss": 0.2443,
+      "step": 52143
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.488203784863903e-06,
+      "loss": 0.4128,
+      "step": 52144
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4878679495575374e-06,
+      "loss": 0.1957,
+      "step": 52145
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4875321142511712e-06,
+      "loss": 0.2886,
+      "step": 52146
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4871962789448055e-06,
+      "loss": 0.3862,
+      "step": 52147
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.48686044363844e-06,
+      "loss": 0.2576,
+      "step": 52148
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.486524608332074e-06,
+      "loss": 0.3881,
+      "step": 52149
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4861887730257086e-06,
+      "loss": 0.239,
+      "step": 52150
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4858529377193428e-06,
+      "loss": 0.2619,
+      "step": 52151
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4855171024129766e-06,
+      "loss": 0.2715,
+      "step": 52152
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4851812671066112e-06,
+      "loss": 0.2659,
+      "step": 52153
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4848454318002455e-06,
+      "loss": 0.3345,
+      "step": 52154
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4845095964938797e-06,
+      "loss": 0.4022,
+      "step": 52155
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.484173761187514e-06,
+      "loss": 0.2687,
+      "step": 52156
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.483837925881148e-06,
+      "loss": 0.2208,
+      "step": 52157
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4835020905747824e-06,
+      "loss": 0.4075,
+      "step": 52158
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4831662552684166e-06,
+      "loss": 0.1767,
+      "step": 52159
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.482830419962051e-06,
+      "loss": 0.365,
+      "step": 52160
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.482494584655685e-06,
+      "loss": 0.2382,
+      "step": 52161
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4821587493493192e-06,
+      "loss": 0.2352,
+      "step": 52162
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4818229140429535e-06,
+      "loss": 0.1925,
+      "step": 52163
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4814870787365877e-06,
+      "loss": 0.3074,
+      "step": 52164
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.481151243430222e-06,
+      "loss": 0.1727,
+      "step": 52165
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.480815408123856e-06,
+      "loss": 0.2113,
+      "step": 52166
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4804795728174904e-06,
+      "loss": 0.2351,
+      "step": 52167
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.480143737511125e-06,
+      "loss": 0.2786,
+      "step": 52168
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.479807902204759e-06,
+      "loss": 0.1426,
+      "step": 52169
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.479472066898393e-06,
+      "loss": 0.3443,
+      "step": 52170
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4791362315920277e-06,
+      "loss": 0.3078,
+      "step": 52171
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4788003962856615e-06,
+      "loss": 0.2315,
+      "step": 52172
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.478464560979296e-06,
+      "loss": 0.2625,
+      "step": 52173
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4781287256729304e-06,
+      "loss": 0.4986,
+      "step": 52174
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.477792890366564e-06,
+      "loss": 0.3882,
+      "step": 52175
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.477457055060199e-06,
+      "loss": 0.2679,
+      "step": 52176
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.477121219753833e-06,
+      "loss": 0.2008,
+      "step": 52177
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4767853844474673e-06,
+      "loss": 0.1996,
+      "step": 52178
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4764495491411015e-06,
+      "loss": 0.2358,
+      "step": 52179
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4761137138347357e-06,
+      "loss": 0.1207,
+      "step": 52180
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.47577787852837e-06,
+      "loss": 0.2246,
+      "step": 52181
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.475442043222004e-06,
+      "loss": 0.1523,
+      "step": 52182
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4751062079156384e-06,
+      "loss": 0.2795,
+      "step": 52183
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4747703726092726e-06,
+      "loss": 0.2309,
+      "step": 52184
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.474434537302907e-06,
+      "loss": 0.224,
+      "step": 52185
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.474098701996541e-06,
+      "loss": 0.3192,
+      "step": 52186
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4737628666901753e-06,
+      "loss": 0.3234,
+      "step": 52187
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4734270313838095e-06,
+      "loss": 0.1336,
+      "step": 52188
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4730911960774437e-06,
+      "loss": 0.1646,
+      "step": 52189
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.472755360771078e-06,
+      "loss": 0.3562,
+      "step": 52190
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.472419525464712e-06,
+      "loss": 0.4362,
+      "step": 52191
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4720836901583464e-06,
+      "loss": 0.3623,
+      "step": 52192
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4717478548519806e-06,
+      "loss": 0.4489,
+      "step": 52193
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4714120195456153e-06,
+      "loss": 0.2374,
+      "step": 52194
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.471076184239249e-06,
+      "loss": 0.2836,
+      "step": 52195
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4707403489328837e-06,
+      "loss": 0.3188,
+      "step": 52196
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.470404513626518e-06,
+      "loss": 0.2085,
+      "step": 52197
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4700686783201517e-06,
+      "loss": 0.1719,
+      "step": 52198
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4697328430137864e-06,
+      "loss": 0.1746,
+      "step": 52199
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4693970077074206e-06,
+      "loss": 0.3759,
+      "step": 52200
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.469061172401055e-06,
+      "loss": 0.2366,
+      "step": 52201
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.468725337094689e-06,
+      "loss": 0.2339,
+      "step": 52202
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4683895017883233e-06,
+      "loss": 0.4193,
+      "step": 52203
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4680536664819575e-06,
+      "loss": 0.2442,
+      "step": 52204
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4677178311755917e-06,
+      "loss": 0.2898,
+      "step": 52205
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.467381995869226e-06,
+      "loss": 0.2767,
+      "step": 52206
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.46704616056286e-06,
+      "loss": 0.2917,
+      "step": 52207
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4667103252564944e-06,
+      "loss": 0.3339,
+      "step": 52208
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4663744899501286e-06,
+      "loss": 0.2913,
+      "step": 52209
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.466038654643763e-06,
+      "loss": 0.2938,
+      "step": 52210
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.465702819337397e-06,
+      "loss": 0.2739,
+      "step": 52211
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4653669840310313e-06,
+      "loss": 0.2387,
+      "step": 52212
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4650311487246655e-06,
+      "loss": 0.2751,
+      "step": 52213
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4646953134182997e-06,
+      "loss": 0.3061,
+      "step": 52214
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.464359478111934e-06,
+      "loss": 0.2069,
+      "step": 52215
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.464023642805568e-06,
+      "loss": 0.2447,
+      "step": 52216
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.463687807499203e-06,
+      "loss": 0.2231,
+      "step": 52217
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4633519721928366e-06,
+      "loss": 0.3003,
+      "step": 52218
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.463016136886471e-06,
+      "loss": 0.2335,
+      "step": 52219
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4626803015801055e-06,
+      "loss": 0.2713,
+      "step": 52220
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4623444662737393e-06,
+      "loss": 0.2889,
+      "step": 52221
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.462008630967374e-06,
+      "loss": 0.5871,
+      "step": 52222
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.461672795661008e-06,
+      "loss": 0.3112,
+      "step": 52223
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4613369603546424e-06,
+      "loss": 0.1091,
+      "step": 52224
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4610011250482766e-06,
+      "loss": 0.3954,
+      "step": 52225
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.460665289741911e-06,
+      "loss": 0.3779,
+      "step": 52226
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.460329454435545e-06,
+      "loss": 0.217,
+      "step": 52227
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4599936191291793e-06,
+      "loss": 0.2741,
+      "step": 52228
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4596577838228135e-06,
+      "loss": 0.2724,
+      "step": 52229
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4593219485164477e-06,
+      "loss": 0.2906,
+      "step": 52230
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.458986113210082e-06,
+      "loss": 0.2445,
+      "step": 52231
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.458650277903716e-06,
+      "loss": 0.292,
+      "step": 52232
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4583144425973504e-06,
+      "loss": 0.4119,
+      "step": 52233
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4579786072909846e-06,
+      "loss": 0.4261,
+      "step": 52234
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.457642771984619e-06,
+      "loss": 0.4084,
+      "step": 52235
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.457306936678253e-06,
+      "loss": 0.3459,
+      "step": 52236
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4569711013718873e-06,
+      "loss": 0.28,
+      "step": 52237
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4566352660655215e-06,
+      "loss": 0.457,
+      "step": 52238
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4562994307591558e-06,
+      "loss": 0.3002,
+      "step": 52239
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4559635954527904e-06,
+      "loss": 0.2748,
+      "step": 52240
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.455627760146424e-06,
+      "loss": 0.3155,
+      "step": 52241
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4552919248400584e-06,
+      "loss": 0.2882,
+      "step": 52242
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.454956089533693e-06,
+      "loss": 0.3303,
+      "step": 52243
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.454620254227327e-06,
+      "loss": 0.4907,
+      "step": 52244
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4542844189209615e-06,
+      "loss": 0.2493,
+      "step": 52245
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4539485836145957e-06,
+      "loss": 0.2796,
+      "step": 52246
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.45361274830823e-06,
+      "loss": 0.1488,
+      "step": 52247
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.453276913001864e-06,
+      "loss": 0.2724,
+      "step": 52248
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4529410776954984e-06,
+      "loss": 0.2704,
+      "step": 52249
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4526052423891326e-06,
+      "loss": 0.2359,
+      "step": 52250
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.452269407082767e-06,
+      "loss": 0.2589,
+      "step": 52251
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.451933571776401e-06,
+      "loss": 0.2867,
+      "step": 52252
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4515977364700353e-06,
+      "loss": 0.354,
+      "step": 52253
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4512619011636695e-06,
+      "loss": 0.2189,
+      "step": 52254
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4509260658573038e-06,
+      "loss": 0.1499,
+      "step": 52255
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.450590230550938e-06,
+      "loss": 0.1932,
+      "step": 52256
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.450254395244572e-06,
+      "loss": 0.3164,
+      "step": 52257
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4499185599382064e-06,
+      "loss": 0.2826,
+      "step": 52258
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4495827246318407e-06,
+      "loss": 0.1731,
+      "step": 52259
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.449246889325475e-06,
+      "loss": 0.3799,
+      "step": 52260
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.448911054019109e-06,
+      "loss": 0.2738,
+      "step": 52261
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4485752187127433e-06,
+      "loss": 0.2825,
+      "step": 52262
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.448239383406378e-06,
+      "loss": 0.3009,
+      "step": 52263
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4479035481000118e-06,
+      "loss": 0.2321,
+      "step": 52264
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.447567712793646e-06,
+      "loss": 0.2908,
+      "step": 52265
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4472318774872806e-06,
+      "loss": 0.1427,
+      "step": 52266
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4468960421809144e-06,
+      "loss": 0.1788,
+      "step": 52267
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.446560206874549e-06,
+      "loss": 0.3578,
+      "step": 52268
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4462243715681833e-06,
+      "loss": 0.2179,
+      "step": 52269
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.445888536261817e-06,
+      "loss": 0.2228,
+      "step": 52270
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4455527009554518e-06,
+      "loss": 0.2086,
+      "step": 52271
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.445216865649086e-06,
+      "loss": 0.3973,
+      "step": 52272
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4448810303427202e-06,
+      "loss": 0.2006,
+      "step": 52273
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4445451950363544e-06,
+      "loss": 0.4015,
+      "step": 52274
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4442093597299887e-06,
+      "loss": 0.2804,
+      "step": 52275
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.443873524423623e-06,
+      "loss": 0.148,
+      "step": 52276
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.443537689117257e-06,
+      "loss": 0.396,
+      "step": 52277
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4432018538108913e-06,
+      "loss": 0.161,
+      "step": 52278
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4428660185045256e-06,
+      "loss": 0.1928,
+      "step": 52279
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4425301831981598e-06,
+      "loss": 0.1556,
+      "step": 52280
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.442194347891794e-06,
+      "loss": 0.2532,
+      "step": 52281
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4418585125854282e-06,
+      "loss": 0.4025,
+      "step": 52282
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4415226772790625e-06,
+      "loss": 0.4229,
+      "step": 52283
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4411868419726967e-06,
+      "loss": 0.249,
+      "step": 52284
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.440851006666331e-06,
+      "loss": 0.1837,
+      "step": 52285
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4405151713599655e-06,
+      "loss": 0.216,
+      "step": 52286
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4401793360535993e-06,
+      "loss": 0.4049,
+      "step": 52287
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4398435007472336e-06,
+      "loss": 0.3009,
+      "step": 52288
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4395076654408682e-06,
+      "loss": 0.1844,
+      "step": 52289
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.439171830134502e-06,
+      "loss": 0.2423,
+      "step": 52290
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4388359948281367e-06,
+      "loss": 0.1696,
+      "step": 52291
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.438500159521771e-06,
+      "loss": 0.2989,
+      "step": 52292
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4381643242154047e-06,
+      "loss": 0.1498,
+      "step": 52293
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4378284889090393e-06,
+      "loss": 0.3196,
+      "step": 52294
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4374926536026736e-06,
+      "loss": 0.2791,
+      "step": 52295
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4371568182963078e-06,
+      "loss": 0.2682,
+      "step": 52296
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.436820982989942e-06,
+      "loss": 0.1717,
+      "step": 52297
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4364851476835762e-06,
+      "loss": 0.4503,
+      "step": 52298
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4361493123772105e-06,
+      "loss": 0.4033,
+      "step": 52299
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4358134770708447e-06,
+      "loss": 0.3507,
+      "step": 52300
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.435477641764479e-06,
+      "loss": 0.4686,
+      "step": 52301
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.435141806458113e-06,
+      "loss": 0.2264,
+      "step": 52302
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4348059711517474e-06,
+      "loss": 0.2531,
+      "step": 52303
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4344701358453816e-06,
+      "loss": 0.1591,
+      "step": 52304
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.434134300539016e-06,
+      "loss": 0.2492,
+      "step": 52305
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.43379846523265e-06,
+      "loss": 0.2068,
+      "step": 52306
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2.4334626299262842e-06,
+      "loss": 0.3937,
+      "step": 52307
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4331267946199185e-06,
+      "loss": 0.1938,
+      "step": 52308
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.432790959313553e-06,
+      "loss": 0.2047,
+      "step": 52309
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.432455124007187e-06,
+      "loss": 0.244,
+      "step": 52310
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.432119288700821e-06,
+      "loss": 0.2589,
+      "step": 52311
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.431783453394456e-06,
+      "loss": 0.2701,
+      "step": 52312
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4314476180880896e-06,
+      "loss": 0.1256,
+      "step": 52313
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4311117827817242e-06,
+      "loss": 0.3362,
+      "step": 52314
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4307759474753585e-06,
+      "loss": 0.2771,
+      "step": 52315
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4304401121689923e-06,
+      "loss": 0.1578,
+      "step": 52316
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.430104276862627e-06,
+      "loss": 0.3498,
+      "step": 52317
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.429768441556261e-06,
+      "loss": 0.1175,
+      "step": 52318
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4294326062498954e-06,
+      "loss": 0.1768,
+      "step": 52319
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4290967709435296e-06,
+      "loss": 0.3509,
+      "step": 52320
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.428760935637164e-06,
+      "loss": 0.3571,
+      "step": 52321
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.428425100330798e-06,
+      "loss": 0.2695,
+      "step": 52322
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4280892650244323e-06,
+      "loss": 0.2062,
+      "step": 52323
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4277534297180665e-06,
+      "loss": 0.1697,
+      "step": 52324
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4274175944117007e-06,
+      "loss": 0.2857,
+      "step": 52325
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.427081759105335e-06,
+      "loss": 0.2433,
+      "step": 52326
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.426745923798969e-06,
+      "loss": 0.2374,
+      "step": 52327
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4264100884926034e-06,
+      "loss": 0.4541,
+      "step": 52328
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4260742531862376e-06,
+      "loss": 0.3779,
+      "step": 52329
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.425738417879872e-06,
+      "loss": 0.1939,
+      "step": 52330
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.425402582573506e-06,
+      "loss": 0.3088,
+      "step": 52331
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4250667472671403e-06,
+      "loss": 0.1241,
+      "step": 52332
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4247309119607745e-06,
+      "loss": 0.2059,
+      "step": 52333
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4243950766544087e-06,
+      "loss": 0.444,
+      "step": 52334
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4240592413480434e-06,
+      "loss": 0.3532,
+      "step": 52335
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.423723406041677e-06,
+      "loss": 0.2554,
+      "step": 52336
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.423387570735312e-06,
+      "loss": 0.36,
+      "step": 52337
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.423051735428946e-06,
+      "loss": 0.254,
+      "step": 52338
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.42271590012258e-06,
+      "loss": 0.3118,
+      "step": 52339
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4223800648162145e-06,
+      "loss": 0.1835,
+      "step": 52340
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4220442295098487e-06,
+      "loss": 0.2892,
+      "step": 52341
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.421708394203483e-06,
+      "loss": 0.3127,
+      "step": 52342
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.421372558897117e-06,
+      "loss": 0.4049,
+      "step": 52343
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4210367235907514e-06,
+      "loss": 0.1591,
+      "step": 52344
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4207008882843856e-06,
+      "loss": 0.2018,
+      "step": 52345
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.42036505297802e-06,
+      "loss": 0.3236,
+      "step": 52346
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.420029217671654e-06,
+      "loss": 0.2552,
+      "step": 52347
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4196933823652883e-06,
+      "loss": 0.141,
+      "step": 52348
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4193575470589225e-06,
+      "loss": 0.3694,
+      "step": 52349
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4190217117525567e-06,
+      "loss": 0.2636,
+      "step": 52350
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.418685876446191e-06,
+      "loss": 0.3542,
+      "step": 52351
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.418350041139825e-06,
+      "loss": 0.3275,
+      "step": 52352
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4180142058334594e-06,
+      "loss": 0.3181,
+      "step": 52353
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4176783705270936e-06,
+      "loss": 0.2365,
+      "step": 52354
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.417342535220728e-06,
+      "loss": 0.1729,
+      "step": 52355
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.417006699914362e-06,
+      "loss": 0.2576,
+      "step": 52356
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4166708646079963e-06,
+      "loss": 0.2534,
+      "step": 52357
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.416335029301631e-06,
+      "loss": 0.2524,
+      "step": 52358
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4159991939952647e-06,
+      "loss": 0.1735,
+      "step": 52359
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4156633586888994e-06,
+      "loss": 0.1714,
+      "step": 52360
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4153275233825336e-06,
+      "loss": 0.4195,
+      "step": 52361
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4149916880761674e-06,
+      "loss": 0.2947,
+      "step": 52362
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.414655852769802e-06,
+      "loss": 0.3165,
+      "step": 52363
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4143200174634363e-06,
+      "loss": 0.3816,
+      "step": 52364
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4139841821570705e-06,
+      "loss": 0.3685,
+      "step": 52365
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4136483468507047e-06,
+      "loss": 0.1724,
+      "step": 52366
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.413312511544339e-06,
+      "loss": 0.3567,
+      "step": 52367
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.412976676237973e-06,
+      "loss": 0.3342,
+      "step": 52368
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4126408409316074e-06,
+      "loss": 0.172,
+      "step": 52369
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4123050056252416e-06,
+      "loss": 0.2311,
+      "step": 52370
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.411969170318876e-06,
+      "loss": 0.2206,
+      "step": 52371
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.41163333501251e-06,
+      "loss": 0.3132,
+      "step": 52372
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4112974997061443e-06,
+      "loss": 0.3086,
+      "step": 52373
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4109616643997785e-06,
+      "loss": 0.1967,
+      "step": 52374
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4106258290934127e-06,
+      "loss": 0.324,
+      "step": 52375
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.410289993787047e-06,
+      "loss": 0.3151,
+      "step": 52376
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.409954158480681e-06,
+      "loss": 0.1634,
+      "step": 52377
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4096183231743154e-06,
+      "loss": 0.4024,
+      "step": 52378
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4092824878679496e-06,
+      "loss": 0.2128,
+      "step": 52379
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.408946652561584e-06,
+      "loss": 0.3016,
+      "step": 52380
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4086108172552185e-06,
+      "loss": 0.468,
+      "step": 52381
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4082749819488523e-06,
+      "loss": 0.2573,
+      "step": 52382
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4079391466424865e-06,
+      "loss": 0.3587,
+      "step": 52383
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.407603311336121e-06,
+      "loss": 0.324,
+      "step": 52384
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.407267476029755e-06,
+      "loss": 0.3467,
+      "step": 52385
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4069316407233896e-06,
+      "loss": 0.2776,
+      "step": 52386
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.406595805417024e-06,
+      "loss": 0.204,
+      "step": 52387
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.406259970110658e-06,
+      "loss": 0.2171,
+      "step": 52388
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4059241348042923e-06,
+      "loss": 0.1671,
+      "step": 52389
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4055882994979265e-06,
+      "loss": 0.1868,
+      "step": 52390
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4052524641915607e-06,
+      "loss": 0.2687,
+      "step": 52391
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.404916628885195e-06,
+      "loss": 0.2644,
+      "step": 52392
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.404580793578829e-06,
+      "loss": 0.2558,
+      "step": 52393
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4042449582724634e-06,
+      "loss": 0.2655,
+      "step": 52394
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4039091229660976e-06,
+      "loss": 0.1979,
+      "step": 52395
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.403573287659732e-06,
+      "loss": 0.2134,
+      "step": 52396
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.403237452353366e-06,
+      "loss": 0.2623,
+      "step": 52397
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4029016170470003e-06,
+      "loss": 0.2361,
+      "step": 52398
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4025657817406345e-06,
+      "loss": 0.2408,
+      "step": 52399
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4022299464342688e-06,
+      "loss": 0.214,
+      "step": 52400
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.401894111127903e-06,
+      "loss": 0.3554,
+      "step": 52401
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.401558275821537e-06,
+      "loss": 0.3267,
+      "step": 52402
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.4012224405151714e-06,
+      "loss": 0.1223,
+      "step": 52403
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.400886605208806e-06,
+      "loss": 0.2163,
+      "step": 52404
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.40055076990244e-06,
+      "loss": 0.295,
+      "step": 52405
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.400214934596074e-06,
+      "loss": 0.2934,
+      "step": 52406
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3998790992897088e-06,
+      "loss": 0.1897,
+      "step": 52407
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3995432639833426e-06,
+      "loss": 0.3734,
+      "step": 52408
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.399207428676977e-06,
+      "loss": 0.2853,
+      "step": 52409
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3988715933706114e-06,
+      "loss": 0.291,
+      "step": 52410
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3985357580642452e-06,
+      "loss": 0.2072,
+      "step": 52411
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.39819992275788e-06,
+      "loss": 0.3384,
+      "step": 52412
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.397864087451514e-06,
+      "loss": 0.2217,
+      "step": 52413
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3975282521451483e-06,
+      "loss": 0.3695,
+      "step": 52414
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3971924168387825e-06,
+      "loss": 0.2748,
+      "step": 52415
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3968565815324168e-06,
+      "loss": 0.4365,
+      "step": 52416
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.396520746226051e-06,
+      "loss": 0.1215,
+      "step": 52417
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3961849109196852e-06,
+      "loss": 0.2014,
+      "step": 52418
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3958490756133194e-06,
+      "loss": 0.3138,
+      "step": 52419
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3955132403069537e-06,
+      "loss": 0.1308,
+      "step": 52420
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.395177405000588e-06,
+      "loss": 0.4445,
+      "step": 52421
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.394841569694222e-06,
+      "loss": 0.2179,
+      "step": 52422
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3945057343878563e-06,
+      "loss": 0.3286,
+      "step": 52423
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3941698990814906e-06,
+      "loss": 0.3119,
+      "step": 52424
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3938340637751248e-06,
+      "loss": 0.1825,
+      "step": 52425
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.393498228468759e-06,
+      "loss": 0.2504,
+      "step": 52426
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3931623931623937e-06,
+      "loss": 0.1438,
+      "step": 52427
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3928265578560275e-06,
+      "loss": 0.1923,
+      "step": 52428
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3924907225496617e-06,
+      "loss": 0.1913,
+      "step": 52429
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3921548872432963e-06,
+      "loss": 0.3718,
+      "step": 52430
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.39181905193693e-06,
+      "loss": 0.1595,
+      "step": 52431
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3914832166305648e-06,
+      "loss": 0.2307,
+      "step": 52432
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.391147381324199e-06,
+      "loss": 0.2857,
+      "step": 52433
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.390811546017833e-06,
+      "loss": 0.2451,
+      "step": 52434
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3904757107114674e-06,
+      "loss": 0.3319,
+      "step": 52435
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3901398754051012e-06,
+      "loss": 0.2492,
+      "step": 52436
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.389804040098736e-06,
+      "loss": 0.1881,
+      "step": 52437
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.38946820479237e-06,
+      "loss": 0.4589,
+      "step": 52438
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3891323694860043e-06,
+      "loss": 0.2384,
+      "step": 52439
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3887965341796386e-06,
+      "loss": 0.2628,
+      "step": 52440
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.388460698873273e-06,
+      "loss": 0.3629,
+      "step": 52441
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.388124863566907e-06,
+      "loss": 0.2532,
+      "step": 52442
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3877890282605412e-06,
+      "loss": 0.1847,
+      "step": 52443
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3874531929541755e-06,
+      "loss": 0.1508,
+      "step": 52444
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3871173576478097e-06,
+      "loss": 0.2128,
+      "step": 52445
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.386781522341444e-06,
+      "loss": 0.2138,
+      "step": 52446
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.386445687035078e-06,
+      "loss": 0.3237,
+      "step": 52447
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3861098517287124e-06,
+      "loss": 0.2908,
+      "step": 52448
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3857740164223466e-06,
+      "loss": 0.2615,
+      "step": 52449
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3854381811159812e-06,
+      "loss": 0.3024,
+      "step": 52450
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.385102345809615e-06,
+      "loss": 0.3023,
+      "step": 52451
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3847665105032493e-06,
+      "loss": 0.2396,
+      "step": 52452
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.384430675196884e-06,
+      "loss": 0.2825,
+      "step": 52453
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3840948398905177e-06,
+      "loss": 0.2689,
+      "step": 52454
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3837590045841523e-06,
+      "loss": 0.1876,
+      "step": 52455
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3834231692777866e-06,
+      "loss": 0.3728,
+      "step": 52456
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3830873339714204e-06,
+      "loss": 0.2208,
+      "step": 52457
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.382751498665055e-06,
+      "loss": 0.2579,
+      "step": 52458
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.382415663358689e-06,
+      "loss": 0.2577,
+      "step": 52459
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3820798280523235e-06,
+      "loss": 0.3851,
+      "step": 52460
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3817439927459577e-06,
+      "loss": 0.1863,
+      "step": 52461
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3814081574395915e-06,
+      "loss": 0.1122,
+      "step": 52462
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.381072322133226e-06,
+      "loss": 0.411,
+      "step": 52463
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3807364868268604e-06,
+      "loss": 0.3075,
+      "step": 52464
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3804006515204946e-06,
+      "loss": 0.3286,
+      "step": 52465
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.380064816214129e-06,
+      "loss": 0.3048,
+      "step": 52466
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.379728980907763e-06,
+      "loss": 0.2495,
+      "step": 52467
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3793931456013973e-06,
+      "loss": 0.3985,
+      "step": 52468
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3790573102950315e-06,
+      "loss": 0.237,
+      "step": 52469
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3787214749886657e-06,
+      "loss": 0.3967,
+      "step": 52470
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3783856396823e-06,
+      "loss": 0.321,
+      "step": 52471
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.378049804375934e-06,
+      "loss": 0.2931,
+      "step": 52472
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3777139690695684e-06,
+      "loss": 0.3384,
+      "step": 52473
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3773781337632026e-06,
+      "loss": 0.1655,
+      "step": 52474
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.377042298456837e-06,
+      "loss": 0.2442,
+      "step": 52475
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3767064631504715e-06,
+      "loss": 0.3399,
+      "step": 52476
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3763706278441053e-06,
+      "loss": 0.4541,
+      "step": 52477
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.37603479253774e-06,
+      "loss": 0.3488,
+      "step": 52478
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.375698957231374e-06,
+      "loss": 0.3484,
+      "step": 52479
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.375363121925008e-06,
+      "loss": 0.2633,
+      "step": 52480
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3750272866186426e-06,
+      "loss": 0.1553,
+      "step": 52481
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3746914513122764e-06,
+      "loss": 0.1745,
+      "step": 52482
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.374355616005911e-06,
+      "loss": 0.2803,
+      "step": 52483
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3740197806995453e-06,
+      "loss": 0.2862,
+      "step": 52484
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.373683945393179e-06,
+      "loss": 0.21,
+      "step": 52485
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3733481100868137e-06,
+      "loss": 0.2405,
+      "step": 52486
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.373012274780448e-06,
+      "loss": 0.225,
+      "step": 52487
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.372676439474082e-06,
+      "loss": 0.2819,
+      "step": 52488
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3723406041677164e-06,
+      "loss": 0.178,
+      "step": 52489
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3720047688613506e-06,
+      "loss": 0.34,
+      "step": 52490
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.371668933554985e-06,
+      "loss": 0.2461,
+      "step": 52491
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.371333098248619e-06,
+      "loss": 0.2544,
+      "step": 52492
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3709972629422533e-06,
+      "loss": 0.2102,
+      "step": 52493
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3706614276358875e-06,
+      "loss": 0.2382,
+      "step": 52494
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3703255923295217e-06,
+      "loss": 0.3822,
+      "step": 52495
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.369989757023156e-06,
+      "loss": 0.3181,
+      "step": 52496
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.36965392171679e-06,
+      "loss": 0.3843,
+      "step": 52497
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3693180864104244e-06,
+      "loss": 0.2147,
+      "step": 52498
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.368982251104059e-06,
+      "loss": 0.1139,
+      "step": 52499
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.368646415797693e-06,
+      "loss": 0.3282,
+      "step": 52500
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3683105804913275e-06,
+      "loss": 0.38,
+      "step": 52501
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3679747451849617e-06,
+      "loss": 0.2368,
+      "step": 52502
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.3676389098785955e-06,
+      "loss": 0.2629,
+      "step": 52503
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.36730307457223e-06,
+      "loss": 0.2492,
+      "step": 52504
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.366967239265864e-06,
+      "loss": 0.2252,
+      "step": 52505
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3666314039594986e-06,
+      "loss": 0.2403,
+      "step": 52506
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.366295568653133e-06,
+      "loss": 0.2382,
+      "step": 52507
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3659597333467666e-06,
+      "loss": 0.2103,
+      "step": 52508
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3656238980404013e-06,
+      "loss": 0.4389,
+      "step": 52509
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3652880627340355e-06,
+      "loss": 0.129,
+      "step": 52510
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3649522274276697e-06,
+      "loss": 0.3343,
+      "step": 52511
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.364616392121304e-06,
+      "loss": 0.2603,
+      "step": 52512
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.364280556814938e-06,
+      "loss": 0.3284,
+      "step": 52513
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3639447215085724e-06,
+      "loss": 0.272,
+      "step": 52514
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3636088862022066e-06,
+      "loss": 0.1951,
+      "step": 52515
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.363273050895841e-06,
+      "loss": 0.2224,
+      "step": 52516
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.362937215589475e-06,
+      "loss": 0.2166,
+      "step": 52517
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3626013802831093e-06,
+      "loss": 0.2332,
+      "step": 52518
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3622655449767435e-06,
+      "loss": 0.3652,
+      "step": 52519
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3619297096703777e-06,
+      "loss": 0.2488,
+      "step": 52520
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.361593874364012e-06,
+      "loss": 0.2372,
+      "step": 52521
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3612580390576466e-06,
+      "loss": 0.3077,
+      "step": 52522
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3609222037512804e-06,
+      "loss": 0.2539,
+      "step": 52523
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3605863684449146e-06,
+      "loss": 0.4052,
+      "step": 52524
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3602505331385493e-06,
+      "loss": 0.2179,
+      "step": 52525
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.359914697832183e-06,
+      "loss": 0.3217,
+      "step": 52526
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3595788625258177e-06,
+      "loss": 0.3391,
+      "step": 52527
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3592430272194515e-06,
+      "loss": 0.2118,
+      "step": 52528
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.358907191913086e-06,
+      "loss": 0.209,
+      "step": 52529
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3585713566067204e-06,
+      "loss": 0.3544,
+      "step": 52530
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.358235521300354e-06,
+      "loss": 0.144,
+      "step": 52531
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.357899685993989e-06,
+      "loss": 0.262,
+      "step": 52532
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.357563850687623e-06,
+      "loss": 0.3866,
+      "step": 52533
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3572280153812573e-06,
+      "loss": 0.2098,
+      "step": 52534
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3568921800748915e-06,
+      "loss": 0.1697,
+      "step": 52535
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3565563447685258e-06,
+      "loss": 0.2225,
+      "step": 52536
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.35622050946216e-06,
+      "loss": 0.2615,
+      "step": 52537
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.355884674155794e-06,
+      "loss": 0.1946,
+      "step": 52538
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3555488388494284e-06,
+      "loss": 0.2287,
+      "step": 52539
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3552130035430626e-06,
+      "loss": 0.365,
+      "step": 52540
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.354877168236697e-06,
+      "loss": 0.2209,
+      "step": 52541
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.354541332930331e-06,
+      "loss": 0.1455,
+      "step": 52542
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3542054976239653e-06,
+      "loss": 0.3854,
+      "step": 52543
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3538696623175995e-06,
+      "loss": 0.2135,
+      "step": 52544
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.353533827011234e-06,
+      "loss": 0.3814,
+      "step": 52545
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.353197991704868e-06,
+      "loss": 0.2329,
+      "step": 52546
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3528621563985022e-06,
+      "loss": 0.3533,
+      "step": 52547
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.352526321092137e-06,
+      "loss": 0.3081,
+      "step": 52548
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3521904857857707e-06,
+      "loss": 0.3355,
+      "step": 52549
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3518546504794053e-06,
+      "loss": 0.4623,
+      "step": 52550
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.351518815173039e-06,
+      "loss": 0.2245,
+      "step": 52551
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3511829798666733e-06,
+      "loss": 0.2758,
+      "step": 52552
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.350847144560308e-06,
+      "loss": 0.1599,
+      "step": 52553
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3505113092539418e-06,
+      "loss": 0.1854,
+      "step": 52554
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3501754739475764e-06,
+      "loss": 0.2176,
+      "step": 52555
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3498396386412107e-06,
+      "loss": 0.4295,
+      "step": 52556
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.349503803334845e-06,
+      "loss": 0.2274,
+      "step": 52557
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.349167968028479e-06,
+      "loss": 0.2049,
+      "step": 52558
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3488321327221133e-06,
+      "loss": 0.3075,
+      "step": 52559
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3484962974157475e-06,
+      "loss": 0.2389,
+      "step": 52560
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3481604621093818e-06,
+      "loss": 0.2959,
+      "step": 52561
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.347824626803016e-06,
+      "loss": 0.1923,
+      "step": 52562
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3474887914966502e-06,
+      "loss": 0.363,
+      "step": 52563
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3471529561902844e-06,
+      "loss": 0.2288,
+      "step": 52564
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3468171208839187e-06,
+      "loss": 0.1791,
+      "step": 52565
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.346481285577553e-06,
+      "loss": 0.2761,
+      "step": 52566
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.346145450271187e-06,
+      "loss": 0.3503,
+      "step": 52567
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3458096149648218e-06,
+      "loss": 0.2423,
+      "step": 52568
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3454737796584556e-06,
+      "loss": 0.2304,
+      "step": 52569
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3451379443520898e-06,
+      "loss": 0.1637,
+      "step": 52570
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3448021090457244e-06,
+      "loss": 0.3745,
+      "step": 52571
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3444662737393582e-06,
+      "loss": 0.1858,
+      "step": 52572
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.344130438432993e-06,
+      "loss": 0.2025,
+      "step": 52573
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3437946031266267e-06,
+      "loss": 0.2305,
+      "step": 52574
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.343458767820261e-06,
+      "loss": 0.234,
+      "step": 52575
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3431229325138956e-06,
+      "loss": 0.3968,
+      "step": 52576
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3427870972075294e-06,
+      "loss": 0.2052,
+      "step": 52577
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.342451261901164e-06,
+      "loss": 0.2857,
+      "step": 52578
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3421154265947982e-06,
+      "loss": 0.2851,
+      "step": 52579
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3417795912884324e-06,
+      "loss": 0.3411,
+      "step": 52580
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3414437559820667e-06,
+      "loss": 0.2731,
+      "step": 52581
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.341107920675701e-06,
+      "loss": 0.141,
+      "step": 52582
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.340772085369335e-06,
+      "loss": 0.1907,
+      "step": 52583
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3404362500629693e-06,
+      "loss": 0.1964,
+      "step": 52584
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3401004147566036e-06,
+      "loss": 0.3333,
+      "step": 52585
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.339764579450238e-06,
+      "loss": 0.1891,
+      "step": 52586
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.339428744143872e-06,
+      "loss": 0.3025,
+      "step": 52587
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3390929088375062e-06,
+      "loss": 0.3041,
+      "step": 52588
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3387570735311405e-06,
+      "loss": 0.1878,
+      "step": 52589
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3384212382247747e-06,
+      "loss": 0.1456,
+      "step": 52590
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3380854029184093e-06,
+      "loss": 0.2816,
+      "step": 52591
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.337749567612043e-06,
+      "loss": 0.3001,
+      "step": 52592
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3374137323056774e-06,
+      "loss": 0.3556,
+      "step": 52593
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3370778969993116e-06,
+      "loss": 0.4821,
+      "step": 52594
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.336742061692946e-06,
+      "loss": 0.2216,
+      "step": 52595
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3364062263865805e-06,
+      "loss": 0.2723,
+      "step": 52596
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3360703910802143e-06,
+      "loss": 0.2177,
+      "step": 52597
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3357345557738485e-06,
+      "loss": 0.2721,
+      "step": 52598
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.335398720467483e-06,
+      "loss": 0.1402,
+      "step": 52599
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.335062885161117e-06,
+      "loss": 0.2509,
+      "step": 52600
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3347270498547516e-06,
+      "loss": 0.2716,
+      "step": 52601
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.334391214548386e-06,
+      "loss": 0.2464,
+      "step": 52602
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3340553792420196e-06,
+      "loss": 0.3646,
+      "step": 52603
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3337195439356542e-06,
+      "loss": 0.2423,
+      "step": 52604
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3333837086292885e-06,
+      "loss": 0.2021,
+      "step": 52605
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3330478733229227e-06,
+      "loss": 0.2384,
+      "step": 52606
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.332712038016557e-06,
+      "loss": 0.179,
+      "step": 52607
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.332376202710191e-06,
+      "loss": 0.2549,
+      "step": 52608
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3320403674038254e-06,
+      "loss": 0.3076,
+      "step": 52609
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3317045320974596e-06,
+      "loss": 0.4148,
+      "step": 52610
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.331368696791094e-06,
+      "loss": 0.2666,
+      "step": 52611
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.331032861484728e-06,
+      "loss": 0.2661,
+      "step": 52612
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3306970261783623e-06,
+      "loss": 0.1723,
+      "step": 52613
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3303611908719965e-06,
+      "loss": 0.3181,
+      "step": 52614
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3300253555656307e-06,
+      "loss": 0.3527,
+      "step": 52615
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.329689520259265e-06,
+      "loss": 0.2091,
+      "step": 52616
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.329353684952899e-06,
+      "loss": 0.3474,
+      "step": 52617
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3290178496465334e-06,
+      "loss": 0.2412,
+      "step": 52618
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.328682014340168e-06,
+      "loss": 0.2372,
+      "step": 52619
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.328346179033802e-06,
+      "loss": 0.2096,
+      "step": 52620
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.328010343727436e-06,
+      "loss": 0.1944,
+      "step": 52621
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3276745084210707e-06,
+      "loss": 0.3354,
+      "step": 52622
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3273386731147045e-06,
+      "loss": 0.289,
+      "step": 52623
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.327002837808339e-06,
+      "loss": 0.1791,
+      "step": 52624
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3266670025019734e-06,
+      "loss": 0.2291,
+      "step": 52625
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.326331167195607e-06,
+      "loss": 0.2429,
+      "step": 52626
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.325995331889242e-06,
+      "loss": 0.3064,
+      "step": 52627
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.325659496582876e-06,
+      "loss": 0.4403,
+      "step": 52628
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3253236612765103e-06,
+      "loss": 0.1278,
+      "step": 52629
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3249878259701445e-06,
+      "loss": 0.2148,
+      "step": 52630
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3246519906637787e-06,
+      "loss": 0.4592,
+      "step": 52631
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.324316155357413e-06,
+      "loss": 0.2637,
+      "step": 52632
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.323980320051047e-06,
+      "loss": 0.275,
+      "step": 52633
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3236444847446814e-06,
+      "loss": 0.1826,
+      "step": 52634
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3233086494383156e-06,
+      "loss": 0.2662,
+      "step": 52635
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.32297281413195e-06,
+      "loss": 0.2114,
+      "step": 52636
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.322636978825584e-06,
+      "loss": 0.2829,
+      "step": 52637
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3223011435192183e-06,
+      "loss": 0.208,
+      "step": 52638
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3219653082128525e-06,
+      "loss": 0.209,
+      "step": 52639
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3216294729064867e-06,
+      "loss": 0.1382,
+      "step": 52640
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.321293637600121e-06,
+      "loss": 0.2614,
+      "step": 52641
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3209578022937556e-06,
+      "loss": 0.3143,
+      "step": 52642
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3206219669873894e-06,
+      "loss": 0.2319,
+      "step": 52643
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3202861316810236e-06,
+      "loss": 0.217,
+      "step": 52644
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3199502963746583e-06,
+      "loss": 0.319,
+      "step": 52645
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.319614461068292e-06,
+      "loss": 0.1959,
+      "step": 52646
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3192786257619267e-06,
+      "loss": 0.2716,
+      "step": 52647
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.318942790455561e-06,
+      "loss": 0.6089,
+      "step": 52648
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3186069551491947e-06,
+      "loss": 0.1695,
+      "step": 52649
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3182711198428294e-06,
+      "loss": 0.3909,
+      "step": 52650
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3179352845364636e-06,
+      "loss": 0.2663,
+      "step": 52651
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.317599449230098e-06,
+      "loss": 0.3318,
+      "step": 52652
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.317263613923732e-06,
+      "loss": 0.3393,
+      "step": 52653
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3169277786173663e-06,
+      "loss": 0.2931,
+      "step": 52654
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3165919433110005e-06,
+      "loss": 0.3609,
+      "step": 52655
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3162561080046347e-06,
+      "loss": 0.217,
+      "step": 52656
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.315920272698269e-06,
+      "loss": 0.1814,
+      "step": 52657
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.315584437391903e-06,
+      "loss": 0.2618,
+      "step": 52658
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3152486020855374e-06,
+      "loss": 0.362,
+      "step": 52659
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3149127667791716e-06,
+      "loss": 0.1948,
+      "step": 52660
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.314576931472806e-06,
+      "loss": 0.2289,
+      "step": 52661
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.31424109616644e-06,
+      "loss": 0.2406,
+      "step": 52662
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3139052608600743e-06,
+      "loss": 0.4153,
+      "step": 52663
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3135694255537085e-06,
+      "loss": 0.3,
+      "step": 52664
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3132335902473427e-06,
+      "loss": 0.1911,
+      "step": 52665
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.312897754940977e-06,
+      "loss": 0.1361,
+      "step": 52666
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.312561919634611e-06,
+      "loss": 0.2583,
+      "step": 52667
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.312226084328246e-06,
+      "loss": 0.2363,
+      "step": 52668
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3118902490218796e-06,
+      "loss": 0.3622,
+      "step": 52669
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3115544137155143e-06,
+      "loss": 0.1969,
+      "step": 52670
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3112185784091485e-06,
+      "loss": 0.2969,
+      "step": 52671
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3108827431027823e-06,
+      "loss": 0.1968,
+      "step": 52672
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.310546907796417e-06,
+      "loss": 0.3174,
+      "step": 52673
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.310211072490051e-06,
+      "loss": 0.3384,
+      "step": 52674
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3098752371836854e-06,
+      "loss": 0.2762,
+      "step": 52675
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3095394018773196e-06,
+      "loss": 0.2674,
+      "step": 52676
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.309203566570954e-06,
+      "loss": 0.2762,
+      "step": 52677
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.308867731264588e-06,
+      "loss": 0.3057,
+      "step": 52678
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3085318959582223e-06,
+      "loss": 0.276,
+      "step": 52679
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3081960606518565e-06,
+      "loss": 0.2077,
+      "step": 52680
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3078602253454908e-06,
+      "loss": 0.1732,
+      "step": 52681
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.307524390039125e-06,
+      "loss": 0.3581,
+      "step": 52682
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.307188554732759e-06,
+      "loss": 0.2113,
+      "step": 52683
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3068527194263934e-06,
+      "loss": 0.2487,
+      "step": 52684
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3065168841200276e-06,
+      "loss": 0.1558,
+      "step": 52685
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.306181048813662e-06,
+      "loss": 0.1816,
+      "step": 52686
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.305845213507296e-06,
+      "loss": 0.4075,
+      "step": 52687
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3055093782009303e-06,
+      "loss": 0.2443,
+      "step": 52688
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3051735428945645e-06,
+      "loss": 0.1659,
+      "step": 52689
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3048377075881988e-06,
+      "loss": 0.2023,
+      "step": 52690
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3045018722818334e-06,
+      "loss": 0.366,
+      "step": 52691
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3041660369754672e-06,
+      "loss": 0.3335,
+      "step": 52692
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.303830201669102e-06,
+      "loss": 0.1091,
+      "step": 52693
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.303494366362736e-06,
+      "loss": 0.3316,
+      "step": 52694
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.30315853105637e-06,
+      "loss": 0.2308,
+      "step": 52695
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3028226957500045e-06,
+      "loss": 0.4215,
+      "step": 52696
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3024868604436388e-06,
+      "loss": 0.352,
+      "step": 52697
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.302151025137273e-06,
+      "loss": 0.2644,
+      "step": 52698
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.301815189830907e-06,
+      "loss": 0.2522,
+      "step": 52699
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3014793545245414e-06,
+      "loss": 0.1515,
+      "step": 52700
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3011435192181757e-06,
+      "loss": 0.1703,
+      "step": 52701
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.30080768391181e-06,
+      "loss": 0.2885,
+      "step": 52702
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.300471848605444e-06,
+      "loss": 0.1553,
+      "step": 52703
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2.3001360132990783e-06,
+      "loss": 0.2974,
+      "step": 52704
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2998001779927125e-06,
+      "loss": 0.27,
+      "step": 52705
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2994643426863468e-06,
+      "loss": 0.3391,
+      "step": 52706
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.299128507379981e-06,
+      "loss": 0.1976,
+      "step": 52707
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2987926720736152e-06,
+      "loss": 0.4342,
+      "step": 52708
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2984568367672494e-06,
+      "loss": 0.2779,
+      "step": 52709
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2981210014608837e-06,
+      "loss": 0.3403,
+      "step": 52710
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.297785166154518e-06,
+      "loss": 0.2165,
+      "step": 52711
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.297449330848152e-06,
+      "loss": 0.2334,
+      "step": 52712
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2971134955417863e-06,
+      "loss": 0.4042,
+      "step": 52713
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.296777660235421e-06,
+      "loss": 0.4076,
+      "step": 52714
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2964418249290548e-06,
+      "loss": 0.2474,
+      "step": 52715
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.296105989622689e-06,
+      "loss": 0.335,
+      "step": 52716
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2957701543163237e-06,
+      "loss": 0.078,
+      "step": 52717
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2954343190099575e-06,
+      "loss": 0.0914,
+      "step": 52718
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.295098483703592e-06,
+      "loss": 0.2445,
+      "step": 52719
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2947626483972263e-06,
+      "loss": 0.2226,
+      "step": 52720
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2944268130908606e-06,
+      "loss": 0.2368,
+      "step": 52721
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2940909777844948e-06,
+      "loss": 0.3963,
+      "step": 52722
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.293755142478129e-06,
+      "loss": 0.3224,
+      "step": 52723
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2934193071717632e-06,
+      "loss": 0.3169,
+      "step": 52724
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2930834718653975e-06,
+      "loss": 0.1987,
+      "step": 52725
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2927476365590317e-06,
+      "loss": 0.2977,
+      "step": 52726
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.292411801252666e-06,
+      "loss": 0.1908,
+      "step": 52727
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2920759659463e-06,
+      "loss": 0.268,
+      "step": 52728
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2917401306399343e-06,
+      "loss": 0.2696,
+      "step": 52729
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2914042953335686e-06,
+      "loss": 0.2123,
+      "step": 52730
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.291068460027203e-06,
+      "loss": 0.1361,
+      "step": 52731
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.290732624720837e-06,
+      "loss": 0.5598,
+      "step": 52732
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2903967894144712e-06,
+      "loss": 0.4365,
+      "step": 52733
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2900609541081055e-06,
+      "loss": 0.2018,
+      "step": 52734
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2897251188017397e-06,
+      "loss": 0.1466,
+      "step": 52735
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.289389283495374e-06,
+      "loss": 0.2106,
+      "step": 52736
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2890534481890086e-06,
+      "loss": 0.4257,
+      "step": 52737
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2887176128826424e-06,
+      "loss": 0.2356,
+      "step": 52738
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2883817775762766e-06,
+      "loss": 0.2385,
+      "step": 52739
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2880459422699112e-06,
+      "loss": 0.2996,
+      "step": 52740
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.287710106963545e-06,
+      "loss": 0.3139,
+      "step": 52741
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2873742716571797e-06,
+      "loss": 0.1603,
+      "step": 52742
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.287038436350814e-06,
+      "loss": 0.3143,
+      "step": 52743
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2867026010444477e-06,
+      "loss": 0.1917,
+      "step": 52744
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2863667657380824e-06,
+      "loss": 0.5433,
+      "step": 52745
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2860309304317166e-06,
+      "loss": 0.143,
+      "step": 52746
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.285695095125351e-06,
+      "loss": 0.1855,
+      "step": 52747
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.285359259818985e-06,
+      "loss": 0.2807,
+      "step": 52748
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2850234245126192e-06,
+      "loss": 0.21,
+      "step": 52749
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2846875892062535e-06,
+      "loss": 0.2323,
+      "step": 52750
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2843517538998877e-06,
+      "loss": 0.4772,
+      "step": 52751
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.284015918593522e-06,
+      "loss": 0.2887,
+      "step": 52752
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.283680083287156e-06,
+      "loss": 0.2411,
+      "step": 52753
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2833442479807904e-06,
+      "loss": 0.186,
+      "step": 52754
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2830084126744246e-06,
+      "loss": 0.1747,
+      "step": 52755
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.282672577368059e-06,
+      "loss": 0.2919,
+      "step": 52756
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.282336742061693e-06,
+      "loss": 0.3021,
+      "step": 52757
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2820009067553273e-06,
+      "loss": 0.3714,
+      "step": 52758
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2816650714489615e-06,
+      "loss": 0.2186,
+      "step": 52759
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.281329236142596e-06,
+      "loss": 0.1456,
+      "step": 52760
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.28099340083623e-06,
+      "loss": 0.1709,
+      "step": 52761
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.280657565529864e-06,
+      "loss": 0.3941,
+      "step": 52762
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.280321730223499e-06,
+      "loss": 0.1695,
+      "step": 52763
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2799858949171326e-06,
+      "loss": 0.3018,
+      "step": 52764
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2796500596107673e-06,
+      "loss": 0.326,
+      "step": 52765
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2793142243044015e-06,
+      "loss": 0.294,
+      "step": 52766
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2789783889980353e-06,
+      "loss": 0.1448,
+      "step": 52767
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.27864255369167e-06,
+      "loss": 0.2924,
+      "step": 52768
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.278306718385304e-06,
+      "loss": 0.4295,
+      "step": 52769
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2779708830789384e-06,
+      "loss": 0.2128,
+      "step": 52770
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2776350477725726e-06,
+      "loss": 0.1971,
+      "step": 52771
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.277299212466207e-06,
+      "loss": 0.3008,
+      "step": 52772
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.276963377159841e-06,
+      "loss": 0.3075,
+      "step": 52773
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2766275418534753e-06,
+      "loss": 0.3777,
+      "step": 52774
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2762917065471095e-06,
+      "loss": 0.2505,
+      "step": 52775
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2759558712407437e-06,
+      "loss": 0.1805,
+      "step": 52776
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.275620035934378e-06,
+      "loss": 0.3536,
+      "step": 52777
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.275284200628012e-06,
+      "loss": 0.3118,
+      "step": 52778
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2749483653216464e-06,
+      "loss": 0.225,
+      "step": 52779
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2746125300152806e-06,
+      "loss": 0.1389,
+      "step": 52780
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.274276694708915e-06,
+      "loss": 0.2765,
+      "step": 52781
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.273940859402549e-06,
+      "loss": 0.2612,
+      "step": 52782
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2736050240961837e-06,
+      "loss": 0.338,
+      "step": 52783
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2732691887898175e-06,
+      "loss": 0.4286,
+      "step": 52784
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2729333534834517e-06,
+      "loss": 0.2014,
+      "step": 52785
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2725975181770864e-06,
+      "loss": 0.1407,
+      "step": 52786
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.27226168287072e-06,
+      "loss": 0.3199,
+      "step": 52787
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.271925847564355e-06,
+      "loss": 0.2174,
+      "step": 52788
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.271590012257989e-06,
+      "loss": 0.2994,
+      "step": 52789
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.271254176951623e-06,
+      "loss": 0.1836,
+      "step": 52790
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2709183416452575e-06,
+      "loss": 0.2022,
+      "step": 52791
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2705825063388917e-06,
+      "loss": 0.3882,
+      "step": 52792
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.270246671032526e-06,
+      "loss": 0.1397,
+      "step": 52793
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.26991083572616e-06,
+      "loss": 0.1986,
+      "step": 52794
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2695750004197944e-06,
+      "loss": 0.2379,
+      "step": 52795
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2692391651134286e-06,
+      "loss": 0.2098,
+      "step": 52796
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.268903329807063e-06,
+      "loss": 0.1992,
+      "step": 52797
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.268567494500697e-06,
+      "loss": 0.193,
+      "step": 52798
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2682316591943313e-06,
+      "loss": 0.224,
+      "step": 52799
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2678958238879655e-06,
+      "loss": 0.3311,
+      "step": 52800
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2675599885815997e-06,
+      "loss": 0.1767,
+      "step": 52801
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.267224153275234e-06,
+      "loss": 0.3273,
+      "step": 52802
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.266888317968868e-06,
+      "loss": 0.3395,
+      "step": 52803
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2665524826625024e-06,
+      "loss": 0.3926,
+      "step": 52804
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2662166473561366e-06,
+      "loss": 0.4475,
+      "step": 52805
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.265880812049771e-06,
+      "loss": 0.1961,
+      "step": 52806
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.265544976743405e-06,
+      "loss": 0.204,
+      "step": 52807
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2652091414370393e-06,
+      "loss": 0.5215,
+      "step": 52808
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.264873306130674e-06,
+      "loss": 0.2808,
+      "step": 52809
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2645374708243077e-06,
+      "loss": 0.289,
+      "step": 52810
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2642016355179424e-06,
+      "loss": 0.2397,
+      "step": 52811
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2638658002115766e-06,
+      "loss": 0.2463,
+      "step": 52812
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2635299649052104e-06,
+      "loss": 0.3269,
+      "step": 52813
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.263194129598845e-06,
+      "loss": 0.2123,
+      "step": 52814
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2628582942924793e-06,
+      "loss": 0.2439,
+      "step": 52815
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2625224589861135e-06,
+      "loss": 0.2831,
+      "step": 52816
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2621866236797477e-06,
+      "loss": 0.0981,
+      "step": 52817
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.261850788373382e-06,
+      "loss": 0.2368,
+      "step": 52818
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.261514953067016e-06,
+      "loss": 0.22,
+      "step": 52819
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2611791177606504e-06,
+      "loss": 0.3153,
+      "step": 52820
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2608432824542846e-06,
+      "loss": 0.2313,
+      "step": 52821
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.260507447147919e-06,
+      "loss": 0.2457,
+      "step": 52822
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.260171611841553e-06,
+      "loss": 0.1885,
+      "step": 52823
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2598357765351873e-06,
+      "loss": 0.1594,
+      "step": 52824
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2594999412288215e-06,
+      "loss": 0.1928,
+      "step": 52825
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2591641059224558e-06,
+      "loss": 0.2524,
+      "step": 52826
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.25882827061609e-06,
+      "loss": 0.2387,
+      "step": 52827
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.258492435309724e-06,
+      "loss": 0.3859,
+      "step": 52828
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2581566000033584e-06,
+      "loss": 0.3106,
+      "step": 52829
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2578207646969927e-06,
+      "loss": 0.2372,
+      "step": 52830
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.257484929390627e-06,
+      "loss": 0.2708,
+      "step": 52831
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2571490940842615e-06,
+      "loss": 0.2414,
+      "step": 52832
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2568132587778953e-06,
+      "loss": 0.1731,
+      "step": 52833
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.25647742347153e-06,
+      "loss": 0.1318,
+      "step": 52834
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.256141588165164e-06,
+      "loss": 0.2267,
+      "step": 52835
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.255805752858798e-06,
+      "loss": 0.3973,
+      "step": 52836
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2554699175524326e-06,
+      "loss": 0.1924,
+      "step": 52837
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.255134082246067e-06,
+      "loss": 0.1219,
+      "step": 52838
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.254798246939701e-06,
+      "loss": 0.278,
+      "step": 52839
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2544624116333353e-06,
+      "loss": 0.2501,
+      "step": 52840
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2541265763269695e-06,
+      "loss": 0.2234,
+      "step": 52841
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2537907410206038e-06,
+      "loss": 0.3397,
+      "step": 52842
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.253454905714238e-06,
+      "loss": 0.1221,
+      "step": 52843
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.253119070407872e-06,
+      "loss": 0.3427,
+      "step": 52844
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2527832351015064e-06,
+      "loss": 0.2841,
+      "step": 52845
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2524473997951407e-06,
+      "loss": 0.2399,
+      "step": 52846
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.252111564488775e-06,
+      "loss": 0.2542,
+      "step": 52847
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.251775729182409e-06,
+      "loss": 0.2273,
+      "step": 52848
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2514398938760433e-06,
+      "loss": 0.2841,
+      "step": 52849
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2511040585696776e-06,
+      "loss": 0.2914,
+      "step": 52850
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2507682232633118e-06,
+      "loss": 0.2169,
+      "step": 52851
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.250432387956946e-06,
+      "loss": 0.1622,
+      "step": 52852
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2500965526505802e-06,
+      "loss": 0.27,
+      "step": 52853
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2497607173442144e-06,
+      "loss": 0.2478,
+      "step": 52854
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.249424882037849e-06,
+      "loss": 0.4003,
+      "step": 52855
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.249089046731483e-06,
+      "loss": 0.4466,
+      "step": 52856
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.248753211425117e-06,
+      "loss": 0.1666,
+      "step": 52857
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2484173761187518e-06,
+      "loss": 0.2388,
+      "step": 52858
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2480815408123856e-06,
+      "loss": 0.2672,
+      "step": 52859
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2477457055060202e-06,
+      "loss": 0.1807,
+      "step": 52860
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2474098701996544e-06,
+      "loss": 0.2789,
+      "step": 52861
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2470740348932887e-06,
+      "loss": 0.2455,
+      "step": 52862
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.246738199586923e-06,
+      "loss": 0.3959,
+      "step": 52863
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.246402364280557e-06,
+      "loss": 0.3873,
+      "step": 52864
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2460665289741913e-06,
+      "loss": 0.2682,
+      "step": 52865
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2457306936678256e-06,
+      "loss": 0.1916,
+      "step": 52866
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2453948583614598e-06,
+      "loss": 0.2668,
+      "step": 52867
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.245059023055094e-06,
+      "loss": 0.2746,
+      "step": 52868
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2447231877487282e-06,
+      "loss": 0.2622,
+      "step": 52869
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2443873524423625e-06,
+      "loss": 0.4459,
+      "step": 52870
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2440515171359967e-06,
+      "loss": 0.2341,
+      "step": 52871
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.243715681829631e-06,
+      "loss": 0.3551,
+      "step": 52872
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.243379846523265e-06,
+      "loss": 0.3776,
+      "step": 52873
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2430440112168993e-06,
+      "loss": 0.2792,
+      "step": 52874
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2427081759105336e-06,
+      "loss": 0.2082,
+      "step": 52875
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.242372340604168e-06,
+      "loss": 0.1541,
+      "step": 52876
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.242036505297802e-06,
+      "loss": 0.2197,
+      "step": 52877
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2417006699914367e-06,
+      "loss": 0.3149,
+      "step": 52878
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2413648346850705e-06,
+      "loss": 0.2916,
+      "step": 52879
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2410289993787047e-06,
+      "loss": 0.2143,
+      "step": 52880
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2406931640723393e-06,
+      "loss": 0.3012,
+      "step": 52881
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.240357328765973e-06,
+      "loss": 0.3414,
+      "step": 52882
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2400214934596078e-06,
+      "loss": 0.4168,
+      "step": 52883
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.239685658153242e-06,
+      "loss": 0.1803,
+      "step": 52884
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.239349822846876e-06,
+      "loss": 0.201,
+      "step": 52885
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2390139875405105e-06,
+      "loss": 0.2106,
+      "step": 52886
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2386781522341447e-06,
+      "loss": 0.2393,
+      "step": 52887
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.238342316927779e-06,
+      "loss": 0.3848,
+      "step": 52888
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.238006481621413e-06,
+      "loss": 0.2841,
+      "step": 52889
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2376706463150474e-06,
+      "loss": 0.2307,
+      "step": 52890
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2373348110086816e-06,
+      "loss": 0.1984,
+      "step": 52891
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.236998975702316e-06,
+      "loss": 0.2506,
+      "step": 52892
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.23666314039595e-06,
+      "loss": 0.2548,
+      "step": 52893
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2363273050895842e-06,
+      "loss": 0.1797,
+      "step": 52894
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2359914697832185e-06,
+      "loss": 0.4042,
+      "step": 52895
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2356556344768527e-06,
+      "loss": 0.285,
+      "step": 52896
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.235319799170487e-06,
+      "loss": 0.1636,
+      "step": 52897
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.234983963864121e-06,
+      "loss": 0.2692,
+      "step": 52898
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2346481285577554e-06,
+      "loss": 0.3756,
+      "step": 52899
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2343122932513896e-06,
+      "loss": 0.3033,
+      "step": 52900
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.2339764579450242e-06,
+      "loss": 0.2788,
+      "step": 52901
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.233640622638658e-06,
+      "loss": 0.1171,
+      "step": 52902
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2333047873322923e-06,
+      "loss": 0.2537,
+      "step": 52903
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.232968952025927e-06,
+      "loss": 0.2618,
+      "step": 52904
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2326331167195607e-06,
+      "loss": 0.3011,
+      "step": 52905
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2322972814131954e-06,
+      "loss": 0.372,
+      "step": 52906
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2319614461068296e-06,
+      "loss": 0.2699,
+      "step": 52907
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2316256108004634e-06,
+      "loss": 0.404,
+      "step": 52908
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.231289775494098e-06,
+      "loss": 0.1753,
+      "step": 52909
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.230953940187732e-06,
+      "loss": 0.3209,
+      "step": 52910
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2306181048813665e-06,
+      "loss": 0.4425,
+      "step": 52911
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2302822695750007e-06,
+      "loss": 0.3602,
+      "step": 52912
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.229946434268635e-06,
+      "loss": 0.3775,
+      "step": 52913
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.229610598962269e-06,
+      "loss": 0.3132,
+      "step": 52914
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2292747636559034e-06,
+      "loss": 0.2216,
+      "step": 52915
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2289389283495376e-06,
+      "loss": 0.2596,
+      "step": 52916
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.228603093043172e-06,
+      "loss": 0.4317,
+      "step": 52917
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.228267257736806e-06,
+      "loss": 0.2575,
+      "step": 52918
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2279314224304403e-06,
+      "loss": 0.2193,
+      "step": 52919
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2275955871240745e-06,
+      "loss": 0.4031,
+      "step": 52920
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2272597518177087e-06,
+      "loss": 0.1527,
+      "step": 52921
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.226923916511343e-06,
+      "loss": 0.3072,
+      "step": 52922
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.226588081204977e-06,
+      "loss": 0.1873,
+      "step": 52923
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.226252245898612e-06,
+      "loss": 0.3335,
+      "step": 52924
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2259164105922456e-06,
+      "loss": 0.289,
+      "step": 52925
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.22558057528588e-06,
+      "loss": 0.1632,
+      "step": 52926
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2252447399795145e-06,
+      "loss": 0.279,
+      "step": 52927
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2249089046731483e-06,
+      "loss": 0.1411,
+      "step": 52928
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.224573069366783e-06,
+      "loss": 0.3048,
+      "step": 52929
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.224237234060417e-06,
+      "loss": 0.179,
+      "step": 52930
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.223901398754051e-06,
+      "loss": 0.1196,
+      "step": 52931
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2235655634476856e-06,
+      "loss": 0.3495,
+      "step": 52932
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2232297281413194e-06,
+      "loss": 0.218,
+      "step": 52933
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.222893892834954e-06,
+      "loss": 0.282,
+      "step": 52934
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2225580575285883e-06,
+      "loss": 0.1427,
+      "step": 52935
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.222222222222222e-06,
+      "loss": 0.3539,
+      "step": 52936
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2218863869158567e-06,
+      "loss": 0.2795,
+      "step": 52937
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.221550551609491e-06,
+      "loss": 0.1826,
+      "step": 52938
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.221214716303125e-06,
+      "loss": 0.3396,
+      "step": 52939
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2208788809967594e-06,
+      "loss": 0.2991,
+      "step": 52940
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2205430456903936e-06,
+      "loss": 0.3631,
+      "step": 52941
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.220207210384028e-06,
+      "loss": 0.2235,
+      "step": 52942
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.219871375077662e-06,
+      "loss": 0.2395,
+      "step": 52943
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2195355397712963e-06,
+      "loss": 0.1281,
+      "step": 52944
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2191997044649305e-06,
+      "loss": 0.195,
+      "step": 52945
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2188638691585647e-06,
+      "loss": 0.3014,
+      "step": 52946
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.218528033852199e-06,
+      "loss": 0.3273,
+      "step": 52947
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.218192198545833e-06,
+      "loss": 0.3502,
+      "step": 52948
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2178563632394674e-06,
+      "loss": 0.2874,
+      "step": 52949
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.217520527933102e-06,
+      "loss": 0.2195,
+      "step": 52950
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.217184692626736e-06,
+      "loss": 0.2814,
+      "step": 52951
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2168488573203705e-06,
+      "loss": 0.2165,
+      "step": 52952
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2165130220140047e-06,
+      "loss": 0.3097,
+      "step": 52953
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2161771867076385e-06,
+      "loss": 0.1663,
+      "step": 52954
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.215841351401273e-06,
+      "loss": 0.3312,
+      "step": 52955
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.215505516094907e-06,
+      "loss": 0.3109,
+      "step": 52956
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2151696807885416e-06,
+      "loss": 0.229,
+      "step": 52957
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.214833845482176e-06,
+      "loss": 0.1323,
+      "step": 52958
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2144980101758096e-06,
+      "loss": 0.2798,
+      "step": 52959
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2141621748694443e-06,
+      "loss": 0.2703,
+      "step": 52960
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2138263395630785e-06,
+      "loss": 0.1981,
+      "step": 52961
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2134905042567127e-06,
+      "loss": 0.3048,
+      "step": 52962
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.213154668950347e-06,
+      "loss": 0.2835,
+      "step": 52963
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.212818833643981e-06,
+      "loss": 0.4085,
+      "step": 52964
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2124829983376154e-06,
+      "loss": 0.5305,
+      "step": 52965
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2121471630312496e-06,
+      "loss": 0.2107,
+      "step": 52966
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.211811327724884e-06,
+      "loss": 0.267,
+      "step": 52967
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.211475492418518e-06,
+      "loss": 0.2277,
+      "step": 52968
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2111396571121523e-06,
+      "loss": 0.2091,
+      "step": 52969
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2108038218057865e-06,
+      "loss": 0.2183,
+      "step": 52970
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2104679864994208e-06,
+      "loss": 0.2817,
+      "step": 52971
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.210132151193055e-06,
+      "loss": 0.2787,
+      "step": 52972
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2097963158866896e-06,
+      "loss": 0.2035,
+      "step": 52973
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2094604805803234e-06,
+      "loss": 0.1748,
+      "step": 52974
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.209124645273958e-06,
+      "loss": 0.3118,
+      "step": 52975
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2087888099675923e-06,
+      "loss": 0.3258,
+      "step": 52976
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.208452974661226e-06,
+      "loss": 0.2173,
+      "step": 52977
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2081171393548607e-06,
+      "loss": 0.3689,
+      "step": 52978
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2077813040484945e-06,
+      "loss": 0.2323,
+      "step": 52979
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.207445468742129e-06,
+      "loss": 0.2797,
+      "step": 52980
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2071096334357634e-06,
+      "loss": 0.3704,
+      "step": 52981
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2067737981293972e-06,
+      "loss": 0.2152,
+      "step": 52982
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.206437962823032e-06,
+      "loss": 0.1664,
+      "step": 52983
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.206102127516666e-06,
+      "loss": 0.2068,
+      "step": 52984
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2057662922103003e-06,
+      "loss": 0.3682,
+      "step": 52985
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2054304569039345e-06,
+      "loss": 0.2015,
+      "step": 52986
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2050946215975688e-06,
+      "loss": 0.2118,
+      "step": 52987
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.204758786291203e-06,
+      "loss": 0.2169,
+      "step": 52988
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.204422950984837e-06,
+      "loss": 0.308,
+      "step": 52989
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2040871156784714e-06,
+      "loss": 0.1583,
+      "step": 52990
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2037512803721057e-06,
+      "loss": 0.2458,
+      "step": 52991
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.20341544506574e-06,
+      "loss": 0.3161,
+      "step": 52992
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.203079609759374e-06,
+      "loss": 0.3119,
+      "step": 52993
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2027437744530083e-06,
+      "loss": 0.1639,
+      "step": 52994
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2024079391466426e-06,
+      "loss": 0.2606,
+      "step": 52995
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.202072103840277e-06,
+      "loss": 0.357,
+      "step": 52996
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.201736268533911e-06,
+      "loss": 0.1781,
+      "step": 52997
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2014004332275452e-06,
+      "loss": 0.3157,
+      "step": 52998
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.20106459792118e-06,
+      "loss": 0.2224,
+      "step": 52999
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2007287626148137e-06,
+      "loss": 0.3086,
+      "step": 53000
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.2003929273084483e-06,
+      "loss": 0.1752,
+      "step": 53001
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.200057092002082e-06,
+      "loss": 0.1478,
+      "step": 53002
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1997212566957168e-06,
+      "loss": 0.2616,
+      "step": 53003
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.199385421389351e-06,
+      "loss": 0.4717,
+      "step": 53004
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.199049586082985e-06,
+      "loss": 0.1917,
+      "step": 53005
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1987137507766194e-06,
+      "loss": 0.2555,
+      "step": 53006
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1983779154702537e-06,
+      "loss": 0.1907,
+      "step": 53007
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.198042080163888e-06,
+      "loss": 0.1723,
+      "step": 53008
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.197706244857522e-06,
+      "loss": 0.2232,
+      "step": 53009
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1973704095511563e-06,
+      "loss": 0.2394,
+      "step": 53010
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1970345742447906e-06,
+      "loss": 0.2259,
+      "step": 53011
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1966987389384248e-06,
+      "loss": 0.2171,
+      "step": 53012
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.196362903632059e-06,
+      "loss": 0.1293,
+      "step": 53013
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1960270683256932e-06,
+      "loss": 0.1691,
+      "step": 53014
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1956912330193275e-06,
+      "loss": 0.2717,
+      "step": 53015
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1953553977129617e-06,
+      "loss": 0.142,
+      "step": 53016
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.195019562406596e-06,
+      "loss": 0.2055,
+      "step": 53017
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.19468372710023e-06,
+      "loss": 0.3117,
+      "step": 53018
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1943478917938648e-06,
+      "loss": 0.2689,
+      "step": 53019
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1940120564874986e-06,
+      "loss": 0.2406,
+      "step": 53020
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.193676221181133e-06,
+      "loss": 0.2129,
+      "step": 53021
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1933403858747674e-06,
+      "loss": 0.3217,
+      "step": 53022
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1930045505684012e-06,
+      "loss": 0.4358,
+      "step": 53023
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.192668715262036e-06,
+      "loss": 0.3506,
+      "step": 53024
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1923328799556697e-06,
+      "loss": 0.1816,
+      "step": 53025
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1919970446493043e-06,
+      "loss": 0.2876,
+      "step": 53026
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1916612093429386e-06,
+      "loss": 0.2319,
+      "step": 53027
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1913253740365724e-06,
+      "loss": 0.1788,
+      "step": 53028
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.190989538730207e-06,
+      "loss": 0.2133,
+      "step": 53029
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1906537034238412e-06,
+      "loss": 0.1825,
+      "step": 53030
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1903178681174755e-06,
+      "loss": 0.3343,
+      "step": 53031
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1899820328111097e-06,
+      "loss": 0.3518,
+      "step": 53032
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.189646197504744e-06,
+      "loss": 0.3059,
+      "step": 53033
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.189310362198378e-06,
+      "loss": 0.1724,
+      "step": 53034
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1889745268920124e-06,
+      "loss": 0.4744,
+      "step": 53035
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1886386915856466e-06,
+      "loss": 0.0971,
+      "step": 53036
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.188302856279281e-06,
+      "loss": 0.2503,
+      "step": 53037
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.187967020972915e-06,
+      "loss": 0.3544,
+      "step": 53038
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1876311856665493e-06,
+      "loss": 0.2296,
+      "step": 53039
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1872953503601835e-06,
+      "loss": 0.2955,
+      "step": 53040
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1869595150538177e-06,
+      "loss": 0.2921,
+      "step": 53041
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1866236797474523e-06,
+      "loss": 0.1765,
+      "step": 53042
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.186287844441086e-06,
+      "loss": 0.2298,
+      "step": 53043
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1859520091347204e-06,
+      "loss": 0.3374,
+      "step": 53044
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.185616173828355e-06,
+      "loss": 0.2661,
+      "step": 53045
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.185280338521989e-06,
+      "loss": 0.3134,
+      "step": 53046
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1849445032156235e-06,
+      "loss": 0.3972,
+      "step": 53047
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1846086679092573e-06,
+      "loss": 0.2048,
+      "step": 53048
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1842728326028915e-06,
+      "loss": 0.2614,
+      "step": 53049
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.183936997296526e-06,
+      "loss": 0.1694,
+      "step": 53050
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.18360116199016e-06,
+      "loss": 0.4236,
+      "step": 53051
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1832653266837946e-06,
+      "loss": 0.2656,
+      "step": 53052
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.182929491377429e-06,
+      "loss": 0.3318,
+      "step": 53053
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.182593656071063e-06,
+      "loss": 0.3858,
+      "step": 53054
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1822578207646973e-06,
+      "loss": 0.2205,
+      "step": 53055
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1819219854583315e-06,
+      "loss": 0.2525,
+      "step": 53056
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1815861501519657e-06,
+      "loss": 0.2283,
+      "step": 53057
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1812503148456e-06,
+      "loss": 0.3158,
+      "step": 53058
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.180914479539234e-06,
+      "loss": 0.357,
+      "step": 53059
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1805786442328684e-06,
+      "loss": 0.2366,
+      "step": 53060
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1802428089265026e-06,
+      "loss": 0.3177,
+      "step": 53061
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.179906973620137e-06,
+      "loss": 0.1862,
+      "step": 53062
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.179571138313771e-06,
+      "loss": 0.2226,
+      "step": 53063
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1792353030074053e-06,
+      "loss": 0.1767,
+      "step": 53064
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.17889946770104e-06,
+      "loss": 0.3793,
+      "step": 53065
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1785636323946737e-06,
+      "loss": 0.2226,
+      "step": 53066
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.178227797088308e-06,
+      "loss": 0.2207,
+      "step": 53067
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.177891961781942e-06,
+      "loss": 0.235,
+      "step": 53068
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1775561264755764e-06,
+      "loss": 0.1399,
+      "step": 53069
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.177220291169211e-06,
+      "loss": 0.1985,
+      "step": 53070
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.176884455862845e-06,
+      "loss": 0.2147,
+      "step": 53071
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.176548620556479e-06,
+      "loss": 0.3103,
+      "step": 53072
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1762127852501137e-06,
+      "loss": 0.3741,
+      "step": 53073
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1758769499437475e-06,
+      "loss": 0.3543,
+      "step": 53074
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.175541114637382e-06,
+      "loss": 0.273,
+      "step": 53075
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1752052793310164e-06,
+      "loss": 0.3028,
+      "step": 53076
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.17486944402465e-06,
+      "loss": 0.2493,
+      "step": 53077
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.174533608718285e-06,
+      "loss": 0.3303,
+      "step": 53078
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.174197773411919e-06,
+      "loss": 0.1972,
+      "step": 53079
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1738619381055533e-06,
+      "loss": 0.2659,
+      "step": 53080
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1735261027991875e-06,
+      "loss": 0.2634,
+      "step": 53081
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1731902674928217e-06,
+      "loss": 0.3361,
+      "step": 53082
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.172854432186456e-06,
+      "loss": 0.2205,
+      "step": 53083
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.17251859688009e-06,
+      "loss": 0.2044,
+      "step": 53084
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1721827615737244e-06,
+      "loss": 0.3561,
+      "step": 53085
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1718469262673586e-06,
+      "loss": 0.1882,
+      "step": 53086
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.171511090960993e-06,
+      "loss": 0.2156,
+      "step": 53087
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.171175255654627e-06,
+      "loss": 0.1407,
+      "step": 53088
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1708394203482613e-06,
+      "loss": 0.1602,
+      "step": 53089
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1705035850418955e-06,
+      "loss": 0.2195,
+      "step": 53090
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1701677497355297e-06,
+      "loss": 0.1506,
+      "step": 53091
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.169831914429164e-06,
+      "loss": 0.3376,
+      "step": 53092
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1694960791227986e-06,
+      "loss": 0.3579,
+      "step": 53093
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1691602438164324e-06,
+      "loss": 0.4061,
+      "step": 53094
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1688244085100666e-06,
+      "loss": 0.2666,
+      "step": 53095
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1684885732037013e-06,
+      "loss": 0.1811,
+      "step": 53096
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.168152737897335e-06,
+      "loss": 0.1886,
+      "step": 53097
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1678169025909697e-06,
+      "loss": 0.1937,
+      "step": 53098
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.167481067284604e-06,
+      "loss": 0.1736,
+      "step": 53099
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1671452319782378e-06,
+      "loss": 0.1295,
+      "step": 53100
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.1668093966718724e-06,
+      "loss": 0.3165,
+      "step": 53101
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1664735613655066e-06,
+      "loss": 0.2841,
+      "step": 53102
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.166137726059141e-06,
+      "loss": 0.3277,
+      "step": 53103
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.165801890752775e-06,
+      "loss": 0.3734,
+      "step": 53104
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1654660554464093e-06,
+      "loss": 0.2988,
+      "step": 53105
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1651302201400435e-06,
+      "loss": 0.1623,
+      "step": 53106
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1647943848336777e-06,
+      "loss": 0.143,
+      "step": 53107
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.164458549527312e-06,
+      "loss": 0.2489,
+      "step": 53108
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.164122714220946e-06,
+      "loss": 0.2986,
+      "step": 53109
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1637868789145804e-06,
+      "loss": 0.2417,
+      "step": 53110
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1634510436082146e-06,
+      "loss": 0.1957,
+      "step": 53111
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.163115208301849e-06,
+      "loss": 0.3108,
+      "step": 53112
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.162779372995483e-06,
+      "loss": 0.1746,
+      "step": 53113
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1624435376891173e-06,
+      "loss": 0.147,
+      "step": 53114
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1621077023827515e-06,
+      "loss": 0.2617,
+      "step": 53115
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.161771867076386e-06,
+      "loss": 0.2462,
+      "step": 53116
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.16143603177002e-06,
+      "loss": 0.3271,
+      "step": 53117
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.161100196463654e-06,
+      "loss": 0.2526,
+      "step": 53118
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.160764361157289e-06,
+      "loss": 0.2336,
+      "step": 53119
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1604285258509227e-06,
+      "loss": 0.3183,
+      "step": 53120
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1600926905445573e-06,
+      "loss": 0.4024,
+      "step": 53121
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1597568552381915e-06,
+      "loss": 0.144,
+      "step": 53122
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1594210199318253e-06,
+      "loss": 0.1599,
+      "step": 53123
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.15908518462546e-06,
+      "loss": 0.5313,
+      "step": 53124
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.158749349319094e-06,
+      "loss": 0.2103,
+      "step": 53125
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1584135140127284e-06,
+      "loss": 0.1546,
+      "step": 53126
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1580776787063626e-06,
+      "loss": 0.2956,
+      "step": 53127
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.157741843399997e-06,
+      "loss": 0.2216,
+      "step": 53128
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.157406008093631e-06,
+      "loss": 0.1559,
+      "step": 53129
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1570701727872653e-06,
+      "loss": 0.3618,
+      "step": 53130
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1567343374808995e-06,
+      "loss": 0.2572,
+      "step": 53131
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1563985021745338e-06,
+      "loss": 0.3254,
+      "step": 53132
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.156062666868168e-06,
+      "loss": 0.341,
+      "step": 53133
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1557268315618022e-06,
+      "loss": 0.3677,
+      "step": 53134
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1553909962554364e-06,
+      "loss": 0.2814,
+      "step": 53135
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1550551609490707e-06,
+      "loss": 0.3559,
+      "step": 53136
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.154719325642705e-06,
+      "loss": 0.28,
+      "step": 53137
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.154383490336339e-06,
+      "loss": 0.2096,
+      "step": 53138
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1540476550299733e-06,
+      "loss": 0.2711,
+      "step": 53139
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1537118197236076e-06,
+      "loss": 0.3177,
+      "step": 53140
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1533759844172418e-06,
+      "loss": 0.4218,
+      "step": 53141
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1530401491108764e-06,
+      "loss": 0.0962,
+      "step": 53142
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1527043138045102e-06,
+      "loss": 0.2641,
+      "step": 53143
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.152368478498145e-06,
+      "loss": 0.2902,
+      "step": 53144
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.152032643191779e-06,
+      "loss": 0.2426,
+      "step": 53145
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.151696807885413e-06,
+      "loss": 0.2506,
+      "step": 53146
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1513609725790475e-06,
+      "loss": 0.1223,
+      "step": 53147
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1510251372726818e-06,
+      "loss": 0.1901,
+      "step": 53148
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.150689301966316e-06,
+      "loss": 0.2446,
+      "step": 53149
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1503534666599502e-06,
+      "loss": 0.1775,
+      "step": 53150
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1500176313535844e-06,
+      "loss": 0.2397,
+      "step": 53151
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1496817960472187e-06,
+      "loss": 0.2897,
+      "step": 53152
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.149345960740853e-06,
+      "loss": 0.2148,
+      "step": 53153
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.149010125434487e-06,
+      "loss": 0.1634,
+      "step": 53154
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1486742901281213e-06,
+      "loss": 0.2631,
+      "step": 53155
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1483384548217556e-06,
+      "loss": 0.3304,
+      "step": 53156
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1480026195153898e-06,
+      "loss": 0.1402,
+      "step": 53157
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.147666784209024e-06,
+      "loss": 0.1774,
+      "step": 53158
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1473309489026582e-06,
+      "loss": 0.2525,
+      "step": 53159
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1469951135962925e-06,
+      "loss": 0.2377,
+      "step": 53160
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1466592782899267e-06,
+      "loss": 0.2849,
+      "step": 53161
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.146323442983561e-06,
+      "loss": 0.1362,
+      "step": 53162
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.145987607677195e-06,
+      "loss": 0.1696,
+      "step": 53163
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1456517723708294e-06,
+      "loss": 0.2084,
+      "step": 53164
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.145315937064464e-06,
+      "loss": 0.1739,
+      "step": 53165
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.144980101758098e-06,
+      "loss": 0.3754,
+      "step": 53166
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1446442664517324e-06,
+      "loss": 0.1975,
+      "step": 53167
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1443084311453667e-06,
+      "loss": 0.1968,
+      "step": 53168
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1439725958390005e-06,
+      "loss": 0.3496,
+      "step": 53169
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.143636760532635e-06,
+      "loss": 0.5357,
+      "step": 53170
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1433009252262693e-06,
+      "loss": 0.1873,
+      "step": 53171
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1429650899199036e-06,
+      "loss": 0.392,
+      "step": 53172
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.142629254613538e-06,
+      "loss": 0.2544,
+      "step": 53173
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.142293419307172e-06,
+      "loss": 0.2659,
+      "step": 53174
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1419575840008062e-06,
+      "loss": 0.2924,
+      "step": 53175
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1416217486944405e-06,
+      "loss": 0.3336,
+      "step": 53176
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1412859133880747e-06,
+      "loss": 0.1562,
+      "step": 53177
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.140950078081709e-06,
+      "loss": 0.2942,
+      "step": 53178
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.140614242775343e-06,
+      "loss": 0.4728,
+      "step": 53179
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1402784074689774e-06,
+      "loss": 0.2314,
+      "step": 53180
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1399425721626116e-06,
+      "loss": 0.1911,
+      "step": 53181
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.139606736856246e-06,
+      "loss": 0.2846,
+      "step": 53182
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.13927090154988e-06,
+      "loss": 0.2722,
+      "step": 53183
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1389350662435143e-06,
+      "loss": 0.192,
+      "step": 53184
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1385992309371485e-06,
+      "loss": 0.1958,
+      "step": 53185
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1382633956307827e-06,
+      "loss": 0.3686,
+      "step": 53186
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.137927560324417e-06,
+      "loss": 0.2704,
+      "step": 53187
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1375917250180516e-06,
+      "loss": 0.3735,
+      "step": 53188
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1372558897116854e-06,
+      "loss": 0.1846,
+      "step": 53189
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1369200544053196e-06,
+      "loss": 0.3312,
+      "step": 53190
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1365842190989542e-06,
+      "loss": 0.3058,
+      "step": 53191
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.136248383792588e-06,
+      "loss": 0.3772,
+      "step": 53192
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1359125484862227e-06,
+      "loss": 0.2912,
+      "step": 53193
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.135576713179857e-06,
+      "loss": 0.198,
+      "step": 53194
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.135240877873491e-06,
+      "loss": 0.1951,
+      "step": 53195
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1349050425671254e-06,
+      "loss": 0.2881,
+      "step": 53196
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1345692072607596e-06,
+      "loss": 0.3903,
+      "step": 53197
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.134233371954394e-06,
+      "loss": 0.2758,
+      "step": 53198
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.133897536648028e-06,
+      "loss": 0.2412,
+      "step": 53199
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1335617013416623e-06,
+      "loss": 0.2758,
+      "step": 53200
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1332258660352965e-06,
+      "loss": 0.2223,
+      "step": 53201
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1328900307289307e-06,
+      "loss": 0.3888,
+      "step": 53202
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.132554195422565e-06,
+      "loss": 0.1939,
+      "step": 53203
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.132218360116199e-06,
+      "loss": 0.2321,
+      "step": 53204
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1318825248098334e-06,
+      "loss": 0.1485,
+      "step": 53205
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1315466895034676e-06,
+      "loss": 0.1831,
+      "step": 53206
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.131210854197102e-06,
+      "loss": 0.2207,
+      "step": 53207
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.130875018890736e-06,
+      "loss": 0.4433,
+      "step": 53208
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1305391835843703e-06,
+      "loss": 0.2652,
+      "step": 53209
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1302033482780045e-06,
+      "loss": 0.3155,
+      "step": 53210
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.129867512971639e-06,
+      "loss": 0.211,
+      "step": 53211
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.129531677665273e-06,
+      "loss": 0.1712,
+      "step": 53212
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.129195842358907e-06,
+      "loss": 0.1334,
+      "step": 53213
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.128860007052542e-06,
+      "loss": 0.2608,
+      "step": 53214
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1285241717461756e-06,
+      "loss": 0.1661,
+      "step": 53215
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1281883364398103e-06,
+      "loss": 0.1217,
+      "step": 53216
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1278525011334445e-06,
+      "loss": 0.2669,
+      "step": 53217
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1275166658270783e-06,
+      "loss": 0.3171,
+      "step": 53218
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.127180830520713e-06,
+      "loss": 0.1782,
+      "step": 53219
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.126844995214347e-06,
+      "loss": 0.315,
+      "step": 53220
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1265091599079814e-06,
+      "loss": 0.1949,
+      "step": 53221
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1261733246016156e-06,
+      "loss": 0.1096,
+      "step": 53222
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.12583748929525e-06,
+      "loss": 0.2968,
+      "step": 53223
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.125501653988884e-06,
+      "loss": 0.2664,
+      "step": 53224
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1251658186825183e-06,
+      "loss": 0.5098,
+      "step": 53225
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1248299833761525e-06,
+      "loss": 0.221,
+      "step": 53226
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1244941480697867e-06,
+      "loss": 0.3407,
+      "step": 53227
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.124158312763421e-06,
+      "loss": 0.2932,
+      "step": 53228
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.123822477457055e-06,
+      "loss": 0.3383,
+      "step": 53229
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1234866421506894e-06,
+      "loss": 0.4411,
+      "step": 53230
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1231508068443236e-06,
+      "loss": 0.2701,
+      "step": 53231
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.122814971537958e-06,
+      "loss": 0.2754,
+      "step": 53232
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.122479136231592e-06,
+      "loss": 0.3217,
+      "step": 53233
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1221433009252267e-06,
+      "loss": 0.2484,
+      "step": 53234
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1218074656188605e-06,
+      "loss": 0.2268,
+      "step": 53235
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1214716303124947e-06,
+      "loss": 0.2885,
+      "step": 53236
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1211357950061294e-06,
+      "loss": 0.2993,
+      "step": 53237
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.120799959699763e-06,
+      "loss": 0.3461,
+      "step": 53238
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.120464124393398e-06,
+      "loss": 0.2823,
+      "step": 53239
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.120128289087032e-06,
+      "loss": 0.2351,
+      "step": 53240
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.119792453780666e-06,
+      "loss": 0.2692,
+      "step": 53241
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1194566184743005e-06,
+      "loss": 0.1736,
+      "step": 53242
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1191207831679347e-06,
+      "loss": 0.304,
+      "step": 53243
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.118784947861569e-06,
+      "loss": 0.2671,
+      "step": 53244
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.118449112555203e-06,
+      "loss": 0.2264,
+      "step": 53245
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1181132772488374e-06,
+      "loss": 0.183,
+      "step": 53246
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1177774419424716e-06,
+      "loss": 0.2476,
+      "step": 53247
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.117441606636106e-06,
+      "loss": 0.2068,
+      "step": 53248
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.11710577132974e-06,
+      "loss": 0.2685,
+      "step": 53249
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1167699360233743e-06,
+      "loss": 0.2609,
+      "step": 53250
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1164341007170085e-06,
+      "loss": 0.2511,
+      "step": 53251
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1160982654106427e-06,
+      "loss": 0.3486,
+      "step": 53252
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.115762430104277e-06,
+      "loss": 0.3024,
+      "step": 53253
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.115426594797911e-06,
+      "loss": 0.2052,
+      "step": 53254
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1150907594915454e-06,
+      "loss": 0.3062,
+      "step": 53255
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1147549241851796e-06,
+      "loss": 0.3035,
+      "step": 53256
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1144190888788143e-06,
+      "loss": 0.158,
+      "step": 53257
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.114083253572448e-06,
+      "loss": 0.1657,
+      "step": 53258
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1137474182660823e-06,
+      "loss": 0.1876,
+      "step": 53259
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.113411582959717e-06,
+      "loss": 0.1448,
+      "step": 53260
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1130757476533508e-06,
+      "loss": 0.2655,
+      "step": 53261
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1127399123469854e-06,
+      "loss": 0.3099,
+      "step": 53262
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1124040770406196e-06,
+      "loss": 0.184,
+      "step": 53263
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1120682417342534e-06,
+      "loss": 0.3638,
+      "step": 53264
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.111732406427888e-06,
+      "loss": 0.3417,
+      "step": 53265
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1113965711215223e-06,
+      "loss": 0.2052,
+      "step": 53266
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1110607358151565e-06,
+      "loss": 0.2994,
+      "step": 53267
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1107249005087908e-06,
+      "loss": 0.2256,
+      "step": 53268
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.110389065202425e-06,
+      "loss": 0.1781,
+      "step": 53269
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.110053229896059e-06,
+      "loss": 0.24,
+      "step": 53270
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1097173945896934e-06,
+      "loss": 0.5517,
+      "step": 53271
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1093815592833276e-06,
+      "loss": 0.1691,
+      "step": 53272
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.109045723976962e-06,
+      "loss": 0.2264,
+      "step": 53273
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.108709888670596e-06,
+      "loss": 0.3218,
+      "step": 53274
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1083740533642303e-06,
+      "loss": 0.2761,
+      "step": 53275
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1080382180578645e-06,
+      "loss": 0.3607,
+      "step": 53276
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1077023827514988e-06,
+      "loss": 0.2988,
+      "step": 53277
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.107366547445133e-06,
+      "loss": 0.2536,
+      "step": 53278
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1070307121387672e-06,
+      "loss": 0.2683,
+      "step": 53279
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1066948768324014e-06,
+      "loss": 0.235,
+      "step": 53280
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1063590415260357e-06,
+      "loss": 0.2827,
+      "step": 53281
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.10602320621967e-06,
+      "loss": 0.2498,
+      "step": 53282
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1056873709133045e-06,
+      "loss": 0.4191,
+      "step": 53283
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1053515356069383e-06,
+      "loss": 0.3406,
+      "step": 53284
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.105015700300573e-06,
+      "loss": 0.3154,
+      "step": 53285
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.104679864994207e-06,
+      "loss": 0.2539,
+      "step": 53286
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.104344029687841e-06,
+      "loss": 0.1804,
+      "step": 53287
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1040081943814757e-06,
+      "loss": 0.0958,
+      "step": 53288
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.10367235907511e-06,
+      "loss": 0.3133,
+      "step": 53289
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.103336523768744e-06,
+      "loss": 0.3932,
+      "step": 53290
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1030006884623783e-06,
+      "loss": 0.2621,
+      "step": 53291
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1026648531560125e-06,
+      "loss": 0.3909,
+      "step": 53292
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1023290178496468e-06,
+      "loss": 0.1729,
+      "step": 53293
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.101993182543281e-06,
+      "loss": 0.2723,
+      "step": 53294
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1016573472369152e-06,
+      "loss": 0.3119,
+      "step": 53295
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1013215119305494e-06,
+      "loss": 0.3141,
+      "step": 53296
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.1009856766241837e-06,
+      "loss": 0.2421,
+      "step": 53297
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.100649841317818e-06,
+      "loss": 0.2656,
+      "step": 53298
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.100314006011452e-06,
+      "loss": 0.1419,
+      "step": 53299
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0999781707050863e-06,
+      "loss": 0.3046,
+      "step": 53300
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0996423353987206e-06,
+      "loss": 0.2761,
+      "step": 53301
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0993065000923548e-06,
+      "loss": 0.59,
+      "step": 53302
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.098970664785989e-06,
+      "loss": 0.1949,
+      "step": 53303
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0986348294796232e-06,
+      "loss": 0.2285,
+      "step": 53304
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0982989941732575e-06,
+      "loss": 0.3241,
+      "step": 53305
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.097963158866892e-06,
+      "loss": 0.2439,
+      "step": 53306
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.097627323560526e-06,
+      "loss": 0.343,
+      "step": 53307
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0972914882541606e-06,
+      "loss": 0.5212,
+      "step": 53308
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0969556529477948e-06,
+      "loss": 0.3096,
+      "step": 53309
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0966198176414286e-06,
+      "loss": 0.2707,
+      "step": 53310
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0962839823350632e-06,
+      "loss": 0.2399,
+      "step": 53311
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0959481470286974e-06,
+      "loss": 0.1804,
+      "step": 53312
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0956123117223317e-06,
+      "loss": 0.1606,
+      "step": 53313
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.095276476415966e-06,
+      "loss": 0.42,
+      "step": 53314
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0949406411096e-06,
+      "loss": 0.3407,
+      "step": 53315
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0946048058032343e-06,
+      "loss": 0.2585,
+      "step": 53316
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0942689704968686e-06,
+      "loss": 0.32,
+      "step": 53317
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.093933135190503e-06,
+      "loss": 0.3239,
+      "step": 53318
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.093597299884137e-06,
+      "loss": 0.1863,
+      "step": 53319
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0932614645777712e-06,
+      "loss": 0.1541,
+      "step": 53320
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0929256292714055e-06,
+      "loss": 0.3844,
+      "step": 53321
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0925897939650397e-06,
+      "loss": 0.3821,
+      "step": 53322
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.092253958658674e-06,
+      "loss": 0.2403,
+      "step": 53323
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.091918123352308e-06,
+      "loss": 0.3491,
+      "step": 53324
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0915822880459424e-06,
+      "loss": 0.3602,
+      "step": 53325
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0912464527395766e-06,
+      "loss": 0.1861,
+      "step": 53326
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.090910617433211e-06,
+      "loss": 0.1927,
+      "step": 53327
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.090574782126845e-06,
+      "loss": 0.3516,
+      "step": 53328
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0902389468204797e-06,
+      "loss": 0.4307,
+      "step": 53329
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0899031115141135e-06,
+      "loss": 0.2943,
+      "step": 53330
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0895672762077477e-06,
+      "loss": 0.2853,
+      "step": 53331
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0892314409013824e-06,
+      "loss": 0.3855,
+      "step": 53332
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.088895605595016e-06,
+      "loss": 0.3801,
+      "step": 53333
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.088559770288651e-06,
+      "loss": 0.2609,
+      "step": 53334
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.088223934982285e-06,
+      "loss": 0.2453,
+      "step": 53335
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0878880996759192e-06,
+      "loss": 0.2399,
+      "step": 53336
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0875522643695535e-06,
+      "loss": 0.2589,
+      "step": 53337
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0872164290631877e-06,
+      "loss": 0.4574,
+      "step": 53338
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.086880593756822e-06,
+      "loss": 0.2467,
+      "step": 53339
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.086544758450456e-06,
+      "loss": 0.3237,
+      "step": 53340
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0862089231440904e-06,
+      "loss": 0.3437,
+      "step": 53341
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0858730878377246e-06,
+      "loss": 0.1624,
+      "step": 53342
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.085537252531359e-06,
+      "loss": 0.1301,
+      "step": 53343
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.085201417224993e-06,
+      "loss": 0.2429,
+      "step": 53344
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0848655819186273e-06,
+      "loss": 0.2194,
+      "step": 53345
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0845297466122615e-06,
+      "loss": 0.2205,
+      "step": 53346
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0841939113058957e-06,
+      "loss": 0.3251,
+      "step": 53347
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.08385807599953e-06,
+      "loss": 0.2588,
+      "step": 53348
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.083522240693164e-06,
+      "loss": 0.3462,
+      "step": 53349
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0831864053867984e-06,
+      "loss": 0.2008,
+      "step": 53350
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0828505700804326e-06,
+      "loss": 0.2608,
+      "step": 53351
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0825147347740673e-06,
+      "loss": 0.3459,
+      "step": 53352
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.082178899467701e-06,
+      "loss": 0.3176,
+      "step": 53353
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0818430641613353e-06,
+      "loss": 0.1737,
+      "step": 53354
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.08150722885497e-06,
+      "loss": 0.3009,
+      "step": 53355
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0811713935486037e-06,
+      "loss": 0.162,
+      "step": 53356
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0808355582422384e-06,
+      "loss": 0.2286,
+      "step": 53357
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0804997229358726e-06,
+      "loss": 0.2805,
+      "step": 53358
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.080163887629507e-06,
+      "loss": 0.2971,
+      "step": 53359
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.079828052323141e-06,
+      "loss": 0.2678,
+      "step": 53360
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.079492217016775e-06,
+      "loss": 0.347,
+      "step": 53361
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0791563817104095e-06,
+      "loss": 0.2403,
+      "step": 53362
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0788205464040437e-06,
+      "loss": 0.1331,
+      "step": 53363
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.078484711097678e-06,
+      "loss": 0.1555,
+      "step": 53364
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.078148875791312e-06,
+      "loss": 0.2111,
+      "step": 53365
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0778130404849464e-06,
+      "loss": 0.3451,
+      "step": 53366
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0774772051785806e-06,
+      "loss": 0.3898,
+      "step": 53367
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.077141369872215e-06,
+      "loss": 0.3314,
+      "step": 53368
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.076805534565849e-06,
+      "loss": 0.266,
+      "step": 53369
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0764696992594833e-06,
+      "loss": 0.3049,
+      "step": 53370
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0761338639531175e-06,
+      "loss": 0.1487,
+      "step": 53371
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0757980286467517e-06,
+      "loss": 0.2374,
+      "step": 53372
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.075462193340386e-06,
+      "loss": 0.2749,
+      "step": 53373
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.07512635803402e-06,
+      "loss": 0.1642,
+      "step": 53374
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.074790522727655e-06,
+      "loss": 0.2953,
+      "step": 53375
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0744546874212886e-06,
+      "loss": 0.2624,
+      "step": 53376
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.074118852114923e-06,
+      "loss": 0.2333,
+      "step": 53377
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0737830168085575e-06,
+      "loss": 0.2367,
+      "step": 53378
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0734471815021913e-06,
+      "loss": 0.2377,
+      "step": 53379
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.073111346195826e-06,
+      "loss": 0.2113,
+      "step": 53380
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.07277551088946e-06,
+      "loss": 0.3182,
+      "step": 53381
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.072439675583094e-06,
+      "loss": 0.3401,
+      "step": 53382
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0721038402767286e-06,
+      "loss": 0.2867,
+      "step": 53383
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0717680049703624e-06,
+      "loss": 0.2889,
+      "step": 53384
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.071432169663997e-06,
+      "loss": 0.274,
+      "step": 53385
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0710963343576313e-06,
+      "loss": 0.2261,
+      "step": 53386
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0707604990512655e-06,
+      "loss": 0.1319,
+      "step": 53387
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0704246637448997e-06,
+      "loss": 0.2757,
+      "step": 53388
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.070088828438534e-06,
+      "loss": 0.3319,
+      "step": 53389
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.069752993132168e-06,
+      "loss": 0.285,
+      "step": 53390
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0694171578258024e-06,
+      "loss": 0.3317,
+      "step": 53391
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0690813225194366e-06,
+      "loss": 0.4503,
+      "step": 53392
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.068745487213071e-06,
+      "loss": 0.3589,
+      "step": 53393
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.068409651906705e-06,
+      "loss": 0.2722,
+      "step": 53394
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0680738166003393e-06,
+      "loss": 0.3218,
+      "step": 53395
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0677379812939735e-06,
+      "loss": 0.3124,
+      "step": 53396
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0674021459876077e-06,
+      "loss": 0.2065,
+      "step": 53397
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0670663106812424e-06,
+      "loss": 0.1842,
+      "step": 53398
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.066730475374876e-06,
+      "loss": 0.2091,
+      "step": 53399
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0663946400685104e-06,
+      "loss": 0.2453,
+      "step": 53400
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.066058804762145e-06,
+      "loss": 0.1151,
+      "step": 53401
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.065722969455779e-06,
+      "loss": 0.2941,
+      "step": 53402
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0653871341494135e-06,
+      "loss": 0.2609,
+      "step": 53403
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0650512988430477e-06,
+      "loss": 0.2017,
+      "step": 53404
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0647154635366815e-06,
+      "loss": 0.1807,
+      "step": 53405
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.064379628230316e-06,
+      "loss": 0.3184,
+      "step": 53406
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.06404379292395e-06,
+      "loss": 0.2237,
+      "step": 53407
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0637079576175846e-06,
+      "loss": 0.4335,
+      "step": 53408
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.063372122311219e-06,
+      "loss": 0.3198,
+      "step": 53409
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0630362870048527e-06,
+      "loss": 0.2275,
+      "step": 53410
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0627004516984873e-06,
+      "loss": 0.3724,
+      "step": 53411
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0623646163921215e-06,
+      "loss": 0.3207,
+      "step": 53412
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0620287810857558e-06,
+      "loss": 0.3525,
+      "step": 53413
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.06169294577939e-06,
+      "loss": 0.4478,
+      "step": 53414
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.061357110473024e-06,
+      "loss": 0.1671,
+      "step": 53415
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0610212751666584e-06,
+      "loss": 0.247,
+      "step": 53416
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0606854398602926e-06,
+      "loss": 0.2029,
+      "step": 53417
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.060349604553927e-06,
+      "loss": 0.1547,
+      "step": 53418
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.060013769247561e-06,
+      "loss": 0.4275,
+      "step": 53419
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0596779339411953e-06,
+      "loss": 0.3211,
+      "step": 53420
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0593420986348295e-06,
+      "loss": 0.1462,
+      "step": 53421
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0590062633284638e-06,
+      "loss": 0.1952,
+      "step": 53422
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.058670428022098e-06,
+      "loss": 0.2358,
+      "step": 53423
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0583345927157326e-06,
+      "loss": 0.2392,
+      "step": 53424
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0579987574093664e-06,
+      "loss": 0.3003,
+      "step": 53425
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.057662922103001e-06,
+      "loss": 0.3177,
+      "step": 53426
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0573270867966353e-06,
+      "loss": 0.3251,
+      "step": 53427
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.056991251490269e-06,
+      "loss": 0.167,
+      "step": 53428
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0566554161839038e-06,
+      "loss": 0.1632,
+      "step": 53429
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0563195808775376e-06,
+      "loss": 0.2392,
+      "step": 53430
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.055983745571172e-06,
+      "loss": 0.2371,
+      "step": 53431
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0556479102648064e-06,
+      "loss": 0.3841,
+      "step": 53432
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0553120749584402e-06,
+      "loss": 0.2297,
+      "step": 53433
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.054976239652075e-06,
+      "loss": 0.2896,
+      "step": 53434
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.054640404345709e-06,
+      "loss": 0.2479,
+      "step": 53435
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0543045690393433e-06,
+      "loss": 0.1937,
+      "step": 53436
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0539687337329776e-06,
+      "loss": 0.4084,
+      "step": 53437
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0536328984266118e-06,
+      "loss": 0.249,
+      "step": 53438
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.053297063120246e-06,
+      "loss": 0.2794,
+      "step": 53439
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0529612278138802e-06,
+      "loss": 0.1859,
+      "step": 53440
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0526253925075144e-06,
+      "loss": 0.269,
+      "step": 53441
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0522895572011487e-06,
+      "loss": 0.2663,
+      "step": 53442
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.051953721894783e-06,
+      "loss": 0.3061,
+      "step": 53443
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.051617886588417e-06,
+      "loss": 0.1701,
+      "step": 53444
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0512820512820513e-06,
+      "loss": 0.2187,
+      "step": 53445
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0509462159756856e-06,
+      "loss": 0.4257,
+      "step": 53446
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0506103806693202e-06,
+      "loss": 0.4984,
+      "step": 53447
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.050274545362954e-06,
+      "loss": 0.1918,
+      "step": 53448
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0499387100565887e-06,
+      "loss": 0.3516,
+      "step": 53449
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.049602874750223e-06,
+      "loss": 0.297,
+      "step": 53450
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0492670394438567e-06,
+      "loss": 0.2085,
+      "step": 53451
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0489312041374913e-06,
+      "loss": 0.2063,
+      "step": 53452
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.048595368831125e-06,
+      "loss": 0.2082,
+      "step": 53453
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0482595335247598e-06,
+      "loss": 0.3809,
+      "step": 53454
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.047923698218394e-06,
+      "loss": 0.125,
+      "step": 53455
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.047587862912028e-06,
+      "loss": 0.2232,
+      "step": 53456
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0472520276056625e-06,
+      "loss": 0.4186,
+      "step": 53457
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0469161922992967e-06,
+      "loss": 0.2266,
+      "step": 53458
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.046580356992931e-06,
+      "loss": 0.5386,
+      "step": 53459
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.046244521686565e-06,
+      "loss": 0.344,
+      "step": 53460
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0459086863801993e-06,
+      "loss": 0.3525,
+      "step": 53461
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0455728510738336e-06,
+      "loss": 0.2116,
+      "step": 53462
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.045237015767468e-06,
+      "loss": 0.3537,
+      "step": 53463
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.044901180461102e-06,
+      "loss": 0.3467,
+      "step": 53464
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0445653451547362e-06,
+      "loss": 0.2377,
+      "step": 53465
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0442295098483705e-06,
+      "loss": 0.3926,
+      "step": 53466
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0438936745420047e-06,
+      "loss": 0.2114,
+      "step": 53467
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.043557839235639e-06,
+      "loss": 0.2626,
+      "step": 53468
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.043222003929273e-06,
+      "loss": 0.4137,
+      "step": 53469
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0428861686229078e-06,
+      "loss": 0.2947,
+      "step": 53470
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0425503333165416e-06,
+      "loss": 0.1531,
+      "step": 53471
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.042214498010176e-06,
+      "loss": 0.3599,
+      "step": 53472
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0418786627038105e-06,
+      "loss": 0.3167,
+      "step": 53473
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0415428273974443e-06,
+      "loss": 0.1855,
+      "step": 53474
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.041206992091079e-06,
+      "loss": 0.1915,
+      "step": 53475
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0408711567847127e-06,
+      "loss": 0.3104,
+      "step": 53476
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0405353214783474e-06,
+      "loss": 0.1874,
+      "step": 53477
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0401994861719816e-06,
+      "loss": 0.1956,
+      "step": 53478
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0398636508656154e-06,
+      "loss": 0.1922,
+      "step": 53479
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.03952781555925e-06,
+      "loss": 0.1819,
+      "step": 53480
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0391919802528842e-06,
+      "loss": 0.2869,
+      "step": 53481
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0388561449465185e-06,
+      "loss": 0.2077,
+      "step": 53482
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0385203096401527e-06,
+      "loss": 0.2503,
+      "step": 53483
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.038184474333787e-06,
+      "loss": 0.3963,
+      "step": 53484
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.037848639027421e-06,
+      "loss": 0.2814,
+      "step": 53485
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0375128037210554e-06,
+      "loss": 0.1835,
+      "step": 53486
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0371769684146896e-06,
+      "loss": 0.1508,
+      "step": 53487
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.036841133108324e-06,
+      "loss": 0.2309,
+      "step": 53488
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.036505297801958e-06,
+      "loss": 0.208,
+      "step": 53489
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0361694624955923e-06,
+      "loss": 0.2722,
+      "step": 53490
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0358336271892265e-06,
+      "loss": 0.1993,
+      "step": 53491
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0354977918828607e-06,
+      "loss": 0.2628,
+      "step": 53492
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0351619565764954e-06,
+      "loss": 0.3814,
+      "step": 53493
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.034826121270129e-06,
+      "loss": 0.4146,
+      "step": 53494
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0344902859637634e-06,
+      "loss": 0.3529,
+      "step": 53495
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.034154450657398e-06,
+      "loss": 0.2263,
+      "step": 53496
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.033818615351032e-06,
+      "loss": 0.3,
+      "step": 53497
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2.0334827800446665e-06,
+      "loss": 0.2275,
+      "step": 53498
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0331469447383003e-06,
+      "loss": 0.3103,
+      "step": 53499
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.032811109431935e-06,
+      "loss": 0.3709,
+      "step": 53500
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.032475274125569e-06,
+      "loss": 0.2533,
+      "step": 53501
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.032139438819203e-06,
+      "loss": 0.3459,
+      "step": 53502
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0318036035128376e-06,
+      "loss": 0.2843,
+      "step": 53503
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.031467768206472e-06,
+      "loss": 0.3826,
+      "step": 53504
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.031131932900106e-06,
+      "loss": 0.2951,
+      "step": 53505
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0307960975937403e-06,
+      "loss": 0.1695,
+      "step": 53506
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0304602622873745e-06,
+      "loss": 0.1763,
+      "step": 53507
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0301244269810087e-06,
+      "loss": 0.2003,
+      "step": 53508
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.029788591674643e-06,
+      "loss": 0.3137,
+      "step": 53509
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.029452756368277e-06,
+      "loss": 0.2333,
+      "step": 53510
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0291169210619114e-06,
+      "loss": 0.1589,
+      "step": 53511
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0287810857555456e-06,
+      "loss": 0.3267,
+      "step": 53512
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.02844525044918e-06,
+      "loss": 0.2331,
+      "step": 53513
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.028109415142814e-06,
+      "loss": 0.2147,
+      "step": 53514
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0277735798364483e-06,
+      "loss": 0.1927,
+      "step": 53515
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.027437744530083e-06,
+      "loss": 0.2889,
+      "step": 53516
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0271019092237167e-06,
+      "loss": 0.4232,
+      "step": 53517
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.026766073917351e-06,
+      "loss": 0.2308,
+      "step": 53518
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.026430238610985e-06,
+      "loss": 0.1873,
+      "step": 53519
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0260944033046194e-06,
+      "loss": 0.2211,
+      "step": 53520
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.025758567998254e-06,
+      "loss": 0.3262,
+      "step": 53521
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.025422732691888e-06,
+      "loss": 0.398,
+      "step": 53522
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.025086897385522e-06,
+      "loss": 0.3098,
+      "step": 53523
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0247510620791567e-06,
+      "loss": 0.1475,
+      "step": 53524
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0244152267727905e-06,
+      "loss": 0.1856,
+      "step": 53525
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.024079391466425e-06,
+      "loss": 0.3504,
+      "step": 53526
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0237435561600594e-06,
+      "loss": 0.3631,
+      "step": 53527
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0234077208536936e-06,
+      "loss": 0.2537,
+      "step": 53528
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.023071885547328e-06,
+      "loss": 0.2622,
+      "step": 53529
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.022736050240962e-06,
+      "loss": 0.1381,
+      "step": 53530
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0224002149345963e-06,
+      "loss": 0.2124,
+      "step": 53531
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0220643796282305e-06,
+      "loss": 0.3708,
+      "step": 53532
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0217285443218647e-06,
+      "loss": 0.196,
+      "step": 53533
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.021392709015499e-06,
+      "loss": 0.3098,
+      "step": 53534
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.021056873709133e-06,
+      "loss": 0.4143,
+      "step": 53535
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0207210384027674e-06,
+      "loss": 0.1737,
+      "step": 53536
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0203852030964016e-06,
+      "loss": 0.2511,
+      "step": 53537
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.020049367790036e-06,
+      "loss": 0.3119,
+      "step": 53538
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0197135324836705e-06,
+      "loss": 0.2205,
+      "step": 53539
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0193776971773043e-06,
+      "loss": 0.3151,
+      "step": 53540
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0190418618709385e-06,
+      "loss": 0.22,
+      "step": 53541
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0187060265645728e-06,
+      "loss": 0.2561,
+      "step": 53542
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.018370191258207e-06,
+      "loss": 0.4298,
+      "step": 53543
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0180343559518416e-06,
+      "loss": 0.216,
+      "step": 53544
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0176985206454754e-06,
+      "loss": 0.2098,
+      "step": 53545
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0173626853391096e-06,
+      "loss": 0.2312,
+      "step": 53546
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0170268500327443e-06,
+      "loss": 0.3258,
+      "step": 53547
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.016691014726378e-06,
+      "loss": 0.3618,
+      "step": 53548
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0163551794200127e-06,
+      "loss": 0.2151,
+      "step": 53549
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.016019344113647e-06,
+      "loss": 0.0961,
+      "step": 53550
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0156835088072808e-06,
+      "loss": 0.1985,
+      "step": 53551
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0153476735009154e-06,
+      "loss": 0.2727,
+      "step": 53552
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0150118381945496e-06,
+      "loss": 0.4573,
+      "step": 53553
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.014676002888184e-06,
+      "loss": 0.295,
+      "step": 53554
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.014340167581818e-06,
+      "loss": 0.3124,
+      "step": 53555
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0140043322754523e-06,
+      "loss": 0.278,
+      "step": 53556
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0136684969690865e-06,
+      "loss": 0.2005,
+      "step": 53557
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0133326616627208e-06,
+      "loss": 0.1883,
+      "step": 53558
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.012996826356355e-06,
+      "loss": 0.1254,
+      "step": 53559
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.012660991049989e-06,
+      "loss": 0.4276,
+      "step": 53560
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0123251557436234e-06,
+      "loss": 0.5198,
+      "step": 53561
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.011989320437258e-06,
+      "loss": 0.2719,
+      "step": 53562
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.011653485130892e-06,
+      "loss": 0.2962,
+      "step": 53563
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.011317649824526e-06,
+      "loss": 0.2437,
+      "step": 53564
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0109818145181603e-06,
+      "loss": 0.2602,
+      "step": 53565
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0106459792117945e-06,
+      "loss": 0.2252,
+      "step": 53566
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.010310143905429e-06,
+      "loss": 0.1599,
+      "step": 53567
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.009974308599063e-06,
+      "loss": 0.1931,
+      "step": 53568
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0096384732926972e-06,
+      "loss": 0.3305,
+      "step": 53569
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.009302637986332e-06,
+      "loss": 0.2104,
+      "step": 53570
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0089668026799657e-06,
+      "loss": 0.365,
+      "step": 53571
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0086309673736003e-06,
+      "loss": 0.2624,
+      "step": 53572
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0082951320672345e-06,
+      "loss": 0.1914,
+      "step": 53573
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0079592967608683e-06,
+      "loss": 0.2904,
+      "step": 53574
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.007623461454503e-06,
+      "loss": 0.3434,
+      "step": 53575
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.007287626148137e-06,
+      "loss": 0.2735,
+      "step": 53576
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0069517908417714e-06,
+      "loss": 0.3446,
+      "step": 53577
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0066159555354057e-06,
+      "loss": 0.2003,
+      "step": 53578
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.00628012022904e-06,
+      "loss": 0.2063,
+      "step": 53579
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.005944284922674e-06,
+      "loss": 0.429,
+      "step": 53580
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0056084496163083e-06,
+      "loss": 0.2378,
+      "step": 53581
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0052726143099426e-06,
+      "loss": 0.1721,
+      "step": 53582
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0049367790035768e-06,
+      "loss": 0.389,
+      "step": 53583
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.004600943697211e-06,
+      "loss": 0.3344,
+      "step": 53584
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0042651083908452e-06,
+      "loss": 0.3394,
+      "step": 53585
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0039292730844794e-06,
+      "loss": 0.2387,
+      "step": 53586
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0035934377781137e-06,
+      "loss": 0.166,
+      "step": 53587
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.003257602471748e-06,
+      "loss": 0.2456,
+      "step": 53588
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.002921767165382e-06,
+      "loss": 0.338,
+      "step": 53589
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0025859318590168e-06,
+      "loss": 0.2003,
+      "step": 53590
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0022500965526506e-06,
+      "loss": 0.2304,
+      "step": 53591
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.001914261246285e-06,
+      "loss": 0.1449,
+      "step": 53592
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0015784259399194e-06,
+      "loss": 0.1593,
+      "step": 53593
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.0012425906335532e-06,
+      "loss": 0.2167,
+      "step": 53594
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.000906755327188e-06,
+      "loss": 0.2828,
+      "step": 53595
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.000570920020822e-06,
+      "loss": 0.3423,
+      "step": 53596
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.000235084714456e-06,
+      "loss": 0.2463,
+      "step": 53597
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9998992494080906e-06,
+      "loss": 0.3686,
+      "step": 53598
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9995634141017248e-06,
+      "loss": 0.2094,
+      "step": 53599
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.999227578795359e-06,
+      "loss": 0.1421,
+      "step": 53600
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9988917434889932e-06,
+      "loss": 0.2177,
+      "step": 53601
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9985559081826275e-06,
+      "loss": 0.2391,
+      "step": 53602
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9982200728762617e-06,
+      "loss": 0.2859,
+      "step": 53603
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.997884237569896e-06,
+      "loss": 0.2659,
+      "step": 53604
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.99754840226353e-06,
+      "loss": 0.2561,
+      "step": 53605
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9972125669571643e-06,
+      "loss": 0.1259,
+      "step": 53606
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9968767316507986e-06,
+      "loss": 0.2215,
+      "step": 53607
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.996540896344433e-06,
+      "loss": 0.1931,
+      "step": 53608
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.996205061038067e-06,
+      "loss": 0.277,
+      "step": 53609
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9958692257317012e-06,
+      "loss": 0.2668,
+      "step": 53610
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9955333904253355e-06,
+      "loss": 0.3057,
+      "step": 53611
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9951975551189697e-06,
+      "loss": 0.2305,
+      "step": 53612
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.994861719812604e-06,
+      "loss": 0.2644,
+      "step": 53613
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.994525884506238e-06,
+      "loss": 0.3449,
+      "step": 53614
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9941900491998724e-06,
+      "loss": 0.2346,
+      "step": 53615
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.993854213893507e-06,
+      "loss": 0.2488,
+      "step": 53616
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.993518378587141e-06,
+      "loss": 0.2025,
+      "step": 53617
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9931825432807755e-06,
+      "loss": 0.2667,
+      "step": 53618
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9928467079744097e-06,
+      "loss": 0.3266,
+      "step": 53619
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9925108726680435e-06,
+      "loss": 0.2224,
+      "step": 53620
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.992175037361678e-06,
+      "loss": 0.3827,
+      "step": 53621
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9918392020553124e-06,
+      "loss": 0.2375,
+      "step": 53622
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9915033667489466e-06,
+      "loss": 0.3993,
+      "step": 53623
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.991167531442581e-06,
+      "loss": 0.2654,
+      "step": 53624
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.990831696136215e-06,
+      "loss": 0.3257,
+      "step": 53625
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9904958608298492e-06,
+      "loss": 0.26,
+      "step": 53626
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9901600255234835e-06,
+      "loss": 0.1668,
+      "step": 53627
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9898241902171177e-06,
+      "loss": 0.3376,
+      "step": 53628
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.989488354910752e-06,
+      "loss": 0.3239,
+      "step": 53629
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.989152519604386e-06,
+      "loss": 0.4199,
+      "step": 53630
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9888166842980204e-06,
+      "loss": 0.2637,
+      "step": 53631
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9884808489916546e-06,
+      "loss": 0.3344,
+      "step": 53632
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.988145013685289e-06,
+      "loss": 0.2178,
+      "step": 53633
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.987809178378923e-06,
+      "loss": 0.1952,
+      "step": 53634
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9874733430725573e-06,
+      "loss": 0.2988,
+      "step": 53635
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9871375077661915e-06,
+      "loss": 0.309,
+      "step": 53636
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9868016724598257e-06,
+      "loss": 0.3985,
+      "step": 53637
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.98646583715346e-06,
+      "loss": 0.3434,
+      "step": 53638
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9861300018470946e-06,
+      "loss": 0.2667,
+      "step": 53639
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9857941665407284e-06,
+      "loss": 0.2583,
+      "step": 53640
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.985458331234363e-06,
+      "loss": 0.3013,
+      "step": 53641
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9851224959279973e-06,
+      "loss": 0.2398,
+      "step": 53642
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.984786660621631e-06,
+      "loss": 0.3499,
+      "step": 53643
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9844508253152657e-06,
+      "loss": 0.195,
+      "step": 53644
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9841149900089e-06,
+      "loss": 0.3599,
+      "step": 53645
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.983779154702534e-06,
+      "loss": 0.4135,
+      "step": 53646
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9834433193961684e-06,
+      "loss": 0.1673,
+      "step": 53647
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9831074840898026e-06,
+      "loss": 0.2831,
+      "step": 53648
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.982771648783437e-06,
+      "loss": 0.3266,
+      "step": 53649
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.982435813477071e-06,
+      "loss": 0.3291,
+      "step": 53650
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9820999781707053e-06,
+      "loss": 0.2684,
+      "step": 53651
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9817641428643395e-06,
+      "loss": 0.372,
+      "step": 53652
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9814283075579737e-06,
+      "loss": 0.368,
+      "step": 53653
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.981092472251608e-06,
+      "loss": 0.1783,
+      "step": 53654
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.980756636945242e-06,
+      "loss": 0.2913,
+      "step": 53655
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9804208016388764e-06,
+      "loss": 0.2862,
+      "step": 53656
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9800849663325106e-06,
+      "loss": 0.2334,
+      "step": 53657
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.979749131026145e-06,
+      "loss": 0.2057,
+      "step": 53658
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.979413295719779e-06,
+      "loss": 0.4488,
+      "step": 53659
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9790774604134133e-06,
+      "loss": 0.2001,
+      "step": 53660
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9787416251070475e-06,
+      "loss": 0.1732,
+      "step": 53661
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.978405789800682e-06,
+      "loss": 0.2505,
+      "step": 53662
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.978069954494316e-06,
+      "loss": 0.2015,
+      "step": 53663
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.97773411918795e-06,
+      "loss": 0.2382,
+      "step": 53664
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.977398283881585e-06,
+      "loss": 0.3214,
+      "step": 53665
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9770624485752186e-06,
+      "loss": 0.2266,
+      "step": 53666
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9767266132688533e-06,
+      "loss": 0.2102,
+      "step": 53667
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9763907779624875e-06,
+      "loss": 0.2721,
+      "step": 53668
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9760549426561217e-06,
+      "loss": 0.2533,
+      "step": 53669
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.975719107349756e-06,
+      "loss": 0.3474,
+      "step": 53670
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.97538327204339e-06,
+      "loss": 0.2836,
+      "step": 53671
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9750474367370244e-06,
+      "loss": 0.2316,
+      "step": 53672
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9747116014306586e-06,
+      "loss": 0.1859,
+      "step": 53673
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.974375766124293e-06,
+      "loss": 0.2613,
+      "step": 53674
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.974039930817927e-06,
+      "loss": 0.306,
+      "step": 53675
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9737040955115613e-06,
+      "loss": 0.1912,
+      "step": 53676
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9733682602051955e-06,
+      "loss": 0.307,
+      "step": 53677
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9730324248988297e-06,
+      "loss": 0.2455,
+      "step": 53678
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.972696589592464e-06,
+      "loss": 0.2979,
+      "step": 53679
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.972360754286098e-06,
+      "loss": 0.3601,
+      "step": 53680
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9720249189797324e-06,
+      "loss": 0.4181,
+      "step": 53681
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9716890836733666e-06,
+      "loss": 0.1761,
+      "step": 53682
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.971353248367001e-06,
+      "loss": 0.2662,
+      "step": 53683
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.971017413060635e-06,
+      "loss": 0.3066,
+      "step": 53684
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9706815777542697e-06,
+      "loss": 0.1456,
+      "step": 53685
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9703457424479035e-06,
+      "loss": 0.215,
+      "step": 53686
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9700099071415378e-06,
+      "loss": 0.2041,
+      "step": 53687
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9696740718351724e-06,
+      "loss": 0.3099,
+      "step": 53688
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.969338236528806e-06,
+      "loss": 0.2663,
+      "step": 53689
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.969002401222441e-06,
+      "loss": 0.2872,
+      "step": 53690
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.968666565916075e-06,
+      "loss": 0.3647,
+      "step": 53691
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9683307306097093e-06,
+      "loss": 0.3473,
+      "step": 53692
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9679948953033435e-06,
+      "loss": 0.2362,
+      "step": 53693
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.9676590599969777e-06,
+      "loss": 0.1656,
+      "step": 53694
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.967323224690612e-06,
+      "loss": 0.274,
+      "step": 53695
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 1.966987389384246e-06,
+      "loss": 0.242,
+      "step": 53696
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9666515540778804e-06,
+      "loss": 0.2838,
+      "step": 53697
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9663157187715146e-06,
+      "loss": 0.1704,
+      "step": 53698
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.965979883465149e-06,
+      "loss": 0.3005,
+      "step": 53699
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.965644048158783e-06,
+      "loss": 0.2892,
+      "step": 53700
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9653082128524173e-06,
+      "loss": 0.2119,
+      "step": 53701
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9649723775460515e-06,
+      "loss": 0.2966,
+      "step": 53702
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9646365422396858e-06,
+      "loss": 0.3191,
+      "step": 53703
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.96430070693332e-06,
+      "loss": 0.1898,
+      "step": 53704
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.963964871626954e-06,
+      "loss": 0.3203,
+      "step": 53705
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9636290363205884e-06,
+      "loss": 0.2808,
+      "step": 53706
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9632932010142227e-06,
+      "loss": 0.1315,
+      "step": 53707
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9629573657078573e-06,
+      "loss": 0.2091,
+      "step": 53708
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.962621530401491e-06,
+      "loss": 0.2995,
+      "step": 53709
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9622856950951253e-06,
+      "loss": 0.203,
+      "step": 53710
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.96194985978876e-06,
+      "loss": 0.3009,
+      "step": 53711
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9616140244823938e-06,
+      "loss": 0.2904,
+      "step": 53712
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9612781891760284e-06,
+      "loss": 0.2092,
+      "step": 53713
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9609423538696626e-06,
+      "loss": 0.1891,
+      "step": 53714
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9606065185632964e-06,
+      "loss": 0.2962,
+      "step": 53715
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.960270683256931e-06,
+      "loss": 0.2695,
+      "step": 53716
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9599348479505653e-06,
+      "loss": 0.1486,
+      "step": 53717
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9595990126441995e-06,
+      "loss": 0.1596,
+      "step": 53718
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9592631773378338e-06,
+      "loss": 0.2394,
+      "step": 53719
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.958927342031468e-06,
+      "loss": 0.2612,
+      "step": 53720
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.958591506725102e-06,
+      "loss": 0.2464,
+      "step": 53721
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9582556714187364e-06,
+      "loss": 0.401,
+      "step": 53722
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9579198361123707e-06,
+      "loss": 0.2456,
+      "step": 53723
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.957584000806005e-06,
+      "loss": 0.2848,
+      "step": 53724
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.957248165499639e-06,
+      "loss": 0.2473,
+      "step": 53725
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9569123301932733e-06,
+      "loss": 0.2757,
+      "step": 53726
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9565764948869076e-06,
+      "loss": 0.2364,
+      "step": 53727
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9562406595805418e-06,
+      "loss": 0.4412,
+      "step": 53728
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.955904824274176e-06,
+      "loss": 0.2037,
+      "step": 53729
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9555689889678102e-06,
+      "loss": 0.1718,
+      "step": 53730
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.955233153661445e-06,
+      "loss": 0.1549,
+      "step": 53731
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9548973183550787e-06,
+      "loss": 0.4206,
+      "step": 53732
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.954561483048713e-06,
+      "loss": 0.2685,
+      "step": 53733
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9542256477423475e-06,
+      "loss": 0.316,
+      "step": 53734
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9538898124359813e-06,
+      "loss": 0.2089,
+      "step": 53735
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.953553977129616e-06,
+      "loss": 0.2998,
+      "step": 53736
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9532181418232502e-06,
+      "loss": 0.2817,
+      "step": 53737
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.952882306516884e-06,
+      "loss": 0.2087,
+      "step": 53738
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9525464712105187e-06,
+      "loss": 0.3326,
+      "step": 53739
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.952210635904153e-06,
+      "loss": 0.3393,
+      "step": 53740
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.951874800597787e-06,
+      "loss": 0.2213,
+      "step": 53741
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9515389652914213e-06,
+      "loss": 0.2156,
+      "step": 53742
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9512031299850556e-06,
+      "loss": 0.2102,
+      "step": 53743
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9508672946786898e-06,
+      "loss": 0.2295,
+      "step": 53744
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.950531459372324e-06,
+      "loss": 0.4535,
+      "step": 53745
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9501956240659582e-06,
+      "loss": 0.3161,
+      "step": 53746
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9498597887595925e-06,
+      "loss": 0.2173,
+      "step": 53747
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9495239534532267e-06,
+      "loss": 0.107,
+      "step": 53748
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.949188118146861e-06,
+      "loss": 0.2456,
+      "step": 53749
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.948852282840495e-06,
+      "loss": 0.2319,
+      "step": 53750
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9485164475341294e-06,
+      "loss": 0.1957,
+      "step": 53751
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9481806122277636e-06,
+      "loss": 0.246,
+      "step": 53752
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.947844776921398e-06,
+      "loss": 0.2454,
+      "step": 53753
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.947508941615032e-06,
+      "loss": 0.2447,
+      "step": 53754
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9471731063086662e-06,
+      "loss": 0.3361,
+      "step": 53755
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9468372710023005e-06,
+      "loss": 0.3412,
+      "step": 53756
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.946501435695935e-06,
+      "loss": 0.2821,
+      "step": 53757
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.946165600389569e-06,
+      "loss": 0.2143,
+      "step": 53758
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9458297650832036e-06,
+      "loss": 0.3001,
+      "step": 53759
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.945493929776838e-06,
+      "loss": 0.2841,
+      "step": 53760
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9451580944704716e-06,
+      "loss": 0.308,
+      "step": 53761
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9448222591641062e-06,
+      "loss": 0.2078,
+      "step": 53762
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9444864238577405e-06,
+      "loss": 0.2151,
+      "step": 53763
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9441505885513747e-06,
+      "loss": 0.2133,
+      "step": 53764
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.943814753245009e-06,
+      "loss": 0.2665,
+      "step": 53765
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.943478917938643e-06,
+      "loss": 0.0721,
+      "step": 53766
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9431430826322774e-06,
+      "loss": 0.3272,
+      "step": 53767
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9428072473259116e-06,
+      "loss": 0.228,
+      "step": 53768
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.942471412019546e-06,
+      "loss": 0.288,
+      "step": 53769
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.94213557671318e-06,
+      "loss": 0.1961,
+      "step": 53770
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9417997414068143e-06,
+      "loss": 0.2035,
+      "step": 53771
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9414639061004485e-06,
+      "loss": 0.1231,
+      "step": 53772
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9411280707940827e-06,
+      "loss": 0.2906,
+      "step": 53773
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.940792235487717e-06,
+      "loss": 0.4575,
+      "step": 53774
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.940456400181351e-06,
+      "loss": 0.2504,
+      "step": 53775
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9401205648749854e-06,
+      "loss": 0.1881,
+      "step": 53776
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9397847295686196e-06,
+      "loss": 0.4285,
+      "step": 53777
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.939448894262254e-06,
+      "loss": 0.1952,
+      "step": 53778
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.939113058955888e-06,
+      "loss": 0.258,
+      "step": 53779
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9387772236495227e-06,
+      "loss": 0.3615,
+      "step": 53780
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9384413883431565e-06,
+      "loss": 0.1916,
+      "step": 53781
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.938105553036791e-06,
+      "loss": 0.2408,
+      "step": 53782
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9377697177304254e-06,
+      "loss": 0.216,
+      "step": 53783
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.937433882424059e-06,
+      "loss": 0.2255,
+      "step": 53784
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.937098047117694e-06,
+      "loss": 0.224,
+      "step": 53785
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.936762211811328e-06,
+      "loss": 0.3424,
+      "step": 53786
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9364263765049623e-06,
+      "loss": 0.4155,
+      "step": 53787
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9360905411985965e-06,
+      "loss": 0.4451,
+      "step": 53788
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9357547058922307e-06,
+      "loss": 0.3506,
+      "step": 53789
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.935418870585865e-06,
+      "loss": 0.2854,
+      "step": 53790
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.935083035279499e-06,
+      "loss": 0.1807,
+      "step": 53791
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9347471999731334e-06,
+      "loss": 0.2092,
+      "step": 53792
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9344113646667676e-06,
+      "loss": 0.3158,
+      "step": 53793
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.934075529360402e-06,
+      "loss": 0.5611,
+      "step": 53794
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.933739694054036e-06,
+      "loss": 0.3899,
+      "step": 53795
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9334038587476703e-06,
+      "loss": 0.2427,
+      "step": 53796
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9330680234413045e-06,
+      "loss": 0.216,
+      "step": 53797
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9327321881349387e-06,
+      "loss": 0.2684,
+      "step": 53798
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.932396352828573e-06,
+      "loss": 0.2477,
+      "step": 53799
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.932060517522207e-06,
+      "loss": 0.2734,
+      "step": 53800
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9317246822158414e-06,
+      "loss": 0.1637,
+      "step": 53801
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9313888469094756e-06,
+      "loss": 0.3693,
+      "step": 53802
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9310530116031103e-06,
+      "loss": 0.2917,
+      "step": 53803
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.930717176296744e-06,
+      "loss": 0.3067,
+      "step": 53804
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9303813409903783e-06,
+      "loss": 0.3019,
+      "step": 53805
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.930045505684013e-06,
+      "loss": 0.2037,
+      "step": 53806
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9297096703776467e-06,
+      "loss": 0.2577,
+      "step": 53807
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9293738350712814e-06,
+      "loss": 0.2899,
+      "step": 53808
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9290379997649156e-06,
+      "loss": 0.2734,
+      "step": 53809
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.92870216445855e-06,
+      "loss": 0.3639,
+      "step": 53810
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.928366329152184e-06,
+      "loss": 0.2527,
+      "step": 53811
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9280304938458183e-06,
+      "loss": 0.2987,
+      "step": 53812
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9276946585394525e-06,
+      "loss": 0.3121,
+      "step": 53813
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9273588232330867e-06,
+      "loss": 0.1674,
+      "step": 53814
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.927022987926721e-06,
+      "loss": 0.2695,
+      "step": 53815
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.926687152620355e-06,
+      "loss": 0.2173,
+      "step": 53816
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9263513173139894e-06,
+      "loss": 0.2833,
+      "step": 53817
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9260154820076236e-06,
+      "loss": 0.2011,
+      "step": 53818
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.925679646701258e-06,
+      "loss": 0.254,
+      "step": 53819
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.925343811394892e-06,
+      "loss": 0.4209,
+      "step": 53820
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9250079760885263e-06,
+      "loss": 0.3274,
+      "step": 53821
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9246721407821605e-06,
+      "loss": 0.2914,
+      "step": 53822
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9243363054757947e-06,
+      "loss": 0.4102,
+      "step": 53823
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.924000470169429e-06,
+      "loss": 0.2503,
+      "step": 53824
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.923664634863063e-06,
+      "loss": 0.3147,
+      "step": 53825
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.923328799556698e-06,
+      "loss": 0.2922,
+      "step": 53826
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9229929642503316e-06,
+      "loss": 0.3467,
+      "step": 53827
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.922657128943966e-06,
+      "loss": 0.2791,
+      "step": 53828
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9223212936376005e-06,
+      "loss": 0.271,
+      "step": 53829
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9219854583312343e-06,
+      "loss": 0.1417,
+      "step": 53830
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.921649623024869e-06,
+      "loss": 0.2713,
+      "step": 53831
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.921313787718503e-06,
+      "loss": 0.1702,
+      "step": 53832
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9209779524121374e-06,
+      "loss": 0.2678,
+      "step": 53833
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9206421171057716e-06,
+      "loss": 0.2719,
+      "step": 53834
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9203062817994054e-06,
+      "loss": 0.2829,
+      "step": 53835
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.91997044649304e-06,
+      "loss": 0.3523,
+      "step": 53836
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9196346111866743e-06,
+      "loss": 0.2414,
+      "step": 53837
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9192987758803085e-06,
+      "loss": 0.1443,
+      "step": 53838
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9189629405739427e-06,
+      "loss": 0.3688,
+      "step": 53839
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.918627105267577e-06,
+      "loss": 0.2343,
+      "step": 53840
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.918291269961211e-06,
+      "loss": 0.3031,
+      "step": 53841
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9179554346548454e-06,
+      "loss": 0.2124,
+      "step": 53842
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9176195993484796e-06,
+      "loss": 0.1794,
+      "step": 53843
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.917283764042114e-06,
+      "loss": 0.3204,
+      "step": 53844
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.916947928735748e-06,
+      "loss": 0.3106,
+      "step": 53845
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9166120934293823e-06,
+      "loss": 0.3172,
+      "step": 53846
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9162762581230165e-06,
+      "loss": 0.1996,
+      "step": 53847
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9159404228166508e-06,
+      "loss": 0.1607,
+      "step": 53848
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9156045875102854e-06,
+      "loss": 0.5492,
+      "step": 53849
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.915268752203919e-06,
+      "loss": 0.3872,
+      "step": 53850
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9149329168975534e-06,
+      "loss": 0.1923,
+      "step": 53851
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.914597081591188e-06,
+      "loss": 0.2772,
+      "step": 53852
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.914261246284822e-06,
+      "loss": 0.338,
+      "step": 53853
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9139254109784565e-06,
+      "loss": 0.1265,
+      "step": 53854
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9135895756720908e-06,
+      "loss": 0.2472,
+      "step": 53855
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9132537403657246e-06,
+      "loss": 0.228,
+      "step": 53856
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.912917905059359e-06,
+      "loss": 0.3056,
+      "step": 53857
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.912582069752993e-06,
+      "loss": 0.2835,
+      "step": 53858
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9122462344466276e-06,
+      "loss": 0.1559,
+      "step": 53859
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.911910399140262e-06,
+      "loss": 0.2414,
+      "step": 53860
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.911574563833896e-06,
+      "loss": 0.362,
+      "step": 53861
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9112387285275303e-06,
+      "loss": 0.2703,
+      "step": 53862
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9109028932211645e-06,
+      "loss": 0.3389,
+      "step": 53863
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9105670579147988e-06,
+      "loss": 0.3985,
+      "step": 53864
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.910231222608433e-06,
+      "loss": 0.3479,
+      "step": 53865
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9098953873020672e-06,
+      "loss": 0.136,
+      "step": 53866
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9095595519957014e-06,
+      "loss": 0.2043,
+      "step": 53867
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9092237166893357e-06,
+      "loss": 0.2306,
+      "step": 53868
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.90888788138297e-06,
+      "loss": 0.1894,
+      "step": 53869
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.908552046076604e-06,
+      "loss": 0.2279,
+      "step": 53870
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9082162107702383e-06,
+      "loss": 0.2711,
+      "step": 53871
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.907880375463873e-06,
+      "loss": 0.5797,
+      "step": 53872
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9075445401575068e-06,
+      "loss": 0.3522,
+      "step": 53873
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9072087048511412e-06,
+      "loss": 0.2181,
+      "step": 53874
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9068728695447754e-06,
+      "loss": 0.2568,
+      "step": 53875
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9065370342384097e-06,
+      "loss": 0.2136,
+      "step": 53876
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9062011989320439e-06,
+      "loss": 0.2193,
+      "step": 53877
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9058653636256783e-06,
+      "loss": 0.3716,
+      "step": 53878
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9055295283193123e-06,
+      "loss": 0.2649,
+      "step": 53879
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9051936930129466e-06,
+      "loss": 0.3591,
+      "step": 53880
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9048578577065808e-06,
+      "loss": 0.2326,
+      "step": 53881
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.904522022400215e-06,
+      "loss": 0.2305,
+      "step": 53882
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9041861870938494e-06,
+      "loss": 0.2894,
+      "step": 53883
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9038503517874835e-06,
+      "loss": 0.1903,
+      "step": 53884
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9035145164811179e-06,
+      "loss": 0.3275,
+      "step": 53885
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9031786811747521e-06,
+      "loss": 0.1257,
+      "step": 53886
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9028428458683861e-06,
+      "loss": 0.1435,
+      "step": 53887
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9025070105620206e-06,
+      "loss": 0.2246,
+      "step": 53888
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9021711752556548e-06,
+      "loss": 0.1925,
+      "step": 53889
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.901835339949289e-06,
+      "loss": 0.3271,
+      "step": 53890
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9014995046429232e-06,
+      "loss": 0.2731,
+      "step": 53891
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9011636693365577e-06,
+      "loss": 0.3023,
+      "step": 53892
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9008278340301917e-06,
+      "loss": 0.2565,
+      "step": 53893
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.900491998723826e-06,
+      "loss": 0.1744,
+      "step": 53894
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 1.9001561634174603e-06,
+      "loss": 0.1941,
+      "step": 53895
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8998203281110944e-06,
+      "loss": 0.2107,
+      "step": 53896
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8994844928047288e-06,
+      "loss": 0.5087,
+      "step": 53897
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.899148657498363e-06,
+      "loss": 0.1943,
+      "step": 53898
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8988128221919972e-06,
+      "loss": 0.216,
+      "step": 53899
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8984769868856315e-06,
+      "loss": 0.503,
+      "step": 53900
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.898141151579266e-06,
+      "loss": 0.2417,
+      "step": 53901
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8978053162729e-06,
+      "loss": 0.2317,
+      "step": 53902
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8974694809665341e-06,
+      "loss": 0.2864,
+      "step": 53903
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8971336456601684e-06,
+      "loss": 0.3647,
+      "step": 53904
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8967978103538026e-06,
+      "loss": 0.1839,
+      "step": 53905
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.896461975047437e-06,
+      "loss": 0.3202,
+      "step": 53906
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.896126139741071e-06,
+      "loss": 0.4585,
+      "step": 53907
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8957903044347053e-06,
+      "loss": 0.2556,
+      "step": 53908
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8954544691283397e-06,
+      "loss": 0.1934,
+      "step": 53909
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8951186338219737e-06,
+      "loss": 0.3568,
+      "step": 53910
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8947827985156081e-06,
+      "loss": 0.1878,
+      "step": 53911
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8944469632092424e-06,
+      "loss": 0.3383,
+      "step": 53912
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8941111279028766e-06,
+      "loss": 0.2167,
+      "step": 53913
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8937752925965108e-06,
+      "loss": 0.2662,
+      "step": 53914
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8934394572901452e-06,
+      "loss": 0.2252,
+      "step": 53915
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8931036219837793e-06,
+      "loss": 0.2166,
+      "step": 53916
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8927677866774135e-06,
+      "loss": 0.2115,
+      "step": 53917
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.892431951371048e-06,
+      "loss": 0.3321,
+      "step": 53918
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.892096116064682e-06,
+      "loss": 0.2579,
+      "step": 53919
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8917602807583164e-06,
+      "loss": 0.2855,
+      "step": 53920
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8914244454519506e-06,
+      "loss": 0.2495,
+      "step": 53921
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8910886101455846e-06,
+      "loss": 0.2425,
+      "step": 53922
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.890752774839219e-06,
+      "loss": 0.4341,
+      "step": 53923
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8904169395328535e-06,
+      "loss": 0.2769,
+      "step": 53924
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8900811042264875e-06,
+      "loss": 0.5283,
+      "step": 53925
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8897452689201217e-06,
+      "loss": 0.3435,
+      "step": 53926
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.889409433613756e-06,
+      "loss": 0.152,
+      "step": 53927
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8890735983073902e-06,
+      "loss": 0.1809,
+      "step": 53928
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8887377630010246e-06,
+      "loss": 0.105,
+      "step": 53929
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8884019276946586e-06,
+      "loss": 0.403,
+      "step": 53930
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8880660923882928e-06,
+      "loss": 0.3223,
+      "step": 53931
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8877302570819273e-06,
+      "loss": 0.1754,
+      "step": 53932
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8873944217755613e-06,
+      "loss": 0.2793,
+      "step": 53933
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8870585864691957e-06,
+      "loss": 0.2052,
+      "step": 53934
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.88672275116283e-06,
+      "loss": 0.408,
+      "step": 53935
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8863869158564642e-06,
+      "loss": 0.2642,
+      "step": 53936
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8860510805500984e-06,
+      "loss": 0.1824,
+      "step": 53937
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8857152452437328e-06,
+      "loss": 0.1265,
+      "step": 53938
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8853794099373668e-06,
+      "loss": 0.1878,
+      "step": 53939
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.885043574631001e-06,
+      "loss": 0.2345,
+      "step": 53940
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8847077393246355e-06,
+      "loss": 0.2758,
+      "step": 53941
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8843719040182695e-06,
+      "loss": 0.2424,
+      "step": 53942
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.884036068711904e-06,
+      "loss": 0.3348,
+      "step": 53943
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8837002334055382e-06,
+      "loss": 0.1727,
+      "step": 53944
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8833643980991722e-06,
+      "loss": 0.2594,
+      "step": 53945
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8830285627928066e-06,
+      "loss": 0.1837,
+      "step": 53946
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.882692727486441e-06,
+      "loss": 0.3164,
+      "step": 53947
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.882356892180075e-06,
+      "loss": 0.2215,
+      "step": 53948
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8820210568737093e-06,
+      "loss": 0.3782,
+      "step": 53949
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8816852215673435e-06,
+      "loss": 0.3883,
+      "step": 53950
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8813493862609777e-06,
+      "loss": 0.3899,
+      "step": 53951
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8810135509546122e-06,
+      "loss": 0.2711,
+      "step": 53952
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8806777156482462e-06,
+      "loss": 0.3371,
+      "step": 53953
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8803418803418804e-06,
+      "loss": 0.325,
+      "step": 53954
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8800060450355148e-06,
+      "loss": 0.1577,
+      "step": 53955
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8796702097291488e-06,
+      "loss": 0.2277,
+      "step": 53956
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8793343744227833e-06,
+      "loss": 0.1655,
+      "step": 53957
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8789985391164175e-06,
+      "loss": 0.1998,
+      "step": 53958
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8786627038100515e-06,
+      "loss": 0.174,
+      "step": 53959
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.878326868503686e-06,
+      "loss": 0.3002,
+      "step": 53960
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8779910331973204e-06,
+      "loss": 0.2513,
+      "step": 53961
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8776551978909544e-06,
+      "loss": 0.2748,
+      "step": 53962
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8773193625845886e-06,
+      "loss": 0.3842,
+      "step": 53963
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.876983527278223e-06,
+      "loss": 0.1716,
+      "step": 53964
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.876647691971857e-06,
+      "loss": 0.2916,
+      "step": 53965
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8763118566654915e-06,
+      "loss": 0.273,
+      "step": 53966
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8759760213591257e-06,
+      "loss": 0.2416,
+      "step": 53967
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8756401860527597e-06,
+      "loss": 0.2891,
+      "step": 53968
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8753043507463942e-06,
+      "loss": 0.2241,
+      "step": 53969
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8749685154400282e-06,
+      "loss": 0.0976,
+      "step": 53970
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8746326801336626e-06,
+      "loss": 0.2851,
+      "step": 53971
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8742968448272968e-06,
+      "loss": 0.3107,
+      "step": 53972
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8739610095209309e-06,
+      "loss": 0.2423,
+      "step": 53973
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8736251742145653e-06,
+      "loss": 0.3083,
+      "step": 53974
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8732893389081997e-06,
+      "loss": 0.3551,
+      "step": 53975
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8729535036018337e-06,
+      "loss": 0.4202,
+      "step": 53976
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.872617668295468e-06,
+      "loss": 0.2362,
+      "step": 53977
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8722818329891024e-06,
+      "loss": 0.312,
+      "step": 53978
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8719459976827364e-06,
+      "loss": 0.2682,
+      "step": 53979
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8716101623763709e-06,
+      "loss": 0.3212,
+      "step": 53980
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.871274327070005e-06,
+      "loss": 0.3179,
+      "step": 53981
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.870938491763639e-06,
+      "loss": 0.2403,
+      "step": 53982
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8706026564572735e-06,
+      "loss": 0.2674,
+      "step": 53983
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8702668211509077e-06,
+      "loss": 0.3452,
+      "step": 53984
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.869930985844542e-06,
+      "loss": 0.3916,
+      "step": 53985
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8695951505381762e-06,
+      "loss": 0.2585,
+      "step": 53986
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8692593152318106e-06,
+      "loss": 0.1622,
+      "step": 53987
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8689234799254446e-06,
+      "loss": 0.2925,
+      "step": 53988
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.868587644619079e-06,
+      "loss": 0.2184,
+      "step": 53989
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8682518093127133e-06,
+      "loss": 0.1712,
+      "step": 53990
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8679159740063473e-06,
+      "loss": 0.2207,
+      "step": 53991
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8675801386999818e-06,
+      "loss": 0.3424,
+      "step": 53992
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8672443033936158e-06,
+      "loss": 0.1666,
+      "step": 53993
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8669084680872502e-06,
+      "loss": 0.5751,
+      "step": 53994
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8665726327808844e-06,
+      "loss": 0.2827,
+      "step": 53995
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8662367974745184e-06,
+      "loss": 0.2922,
+      "step": 53996
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8659009621681529e-06,
+      "loss": 0.3223,
+      "step": 53997
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8655651268617873e-06,
+      "loss": 0.2682,
+      "step": 53998
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8652292915554213e-06,
+      "loss": 0.2621,
+      "step": 53999
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8648934562490555e-06,
+      "loss": 0.4956,
+      "step": 54000
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.86455762094269e-06,
+      "loss": 0.4215,
+      "step": 54001
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.864221785636324e-06,
+      "loss": 0.192,
+      "step": 54002
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8638859503299584e-06,
+      "loss": 0.365,
+      "step": 54003
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8635501150235926e-06,
+      "loss": 0.2527,
+      "step": 54004
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8632142797172267e-06,
+      "loss": 0.2268,
+      "step": 54005
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.862878444410861e-06,
+      "loss": 0.3511,
+      "step": 54006
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8625426091044953e-06,
+      "loss": 0.234,
+      "step": 54007
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8622067737981295e-06,
+      "loss": 0.5708,
+      "step": 54008
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8618709384917638e-06,
+      "loss": 0.274,
+      "step": 54009
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8615351031853982e-06,
+      "loss": 0.3133,
+      "step": 54010
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8611992678790322e-06,
+      "loss": 0.2597,
+      "step": 54011
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8608634325726667e-06,
+      "loss": 0.405,
+      "step": 54012
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8605275972663009e-06,
+      "loss": 0.3204,
+      "step": 54013
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8601917619599349e-06,
+      "loss": 0.2909,
+      "step": 54014
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8598559266535693e-06,
+      "loss": 0.1923,
+      "step": 54015
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8595200913472033e-06,
+      "loss": 0.2985,
+      "step": 54016
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8591842560408378e-06,
+      "loss": 0.2026,
+      "step": 54017
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.858848420734472e-06,
+      "loss": 0.1912,
+      "step": 54018
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.858512585428106e-06,
+      "loss": 0.1713,
+      "step": 54019
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8581767501217404e-06,
+      "loss": 0.2474,
+      "step": 54020
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8578409148153747e-06,
+      "loss": 0.2462,
+      "step": 54021
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8575050795090089e-06,
+      "loss": 0.3119,
+      "step": 54022
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8571692442026431e-06,
+      "loss": 0.2532,
+      "step": 54023
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8568334088962775e-06,
+      "loss": 0.1731,
+      "step": 54024
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8564975735899116e-06,
+      "loss": 0.1344,
+      "step": 54025
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.856161738283546e-06,
+      "loss": 0.2027,
+      "step": 54026
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8558259029771802e-06,
+      "loss": 0.2956,
+      "step": 54027
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8554900676708142e-06,
+      "loss": 0.2593,
+      "step": 54028
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8551542323644487e-06,
+      "loss": 0.2796,
+      "step": 54029
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8548183970580829e-06,
+      "loss": 0.1863,
+      "step": 54030
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8544825617517171e-06,
+      "loss": 0.4095,
+      "step": 54031
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8541467264453513e-06,
+      "loss": 0.1888,
+      "step": 54032
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8538108911389858e-06,
+      "loss": 0.3303,
+      "step": 54033
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8534750558326198e-06,
+      "loss": 0.1829,
+      "step": 54034
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.853139220526254e-06,
+      "loss": 0.4055,
+      "step": 54035
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8528033852198884e-06,
+      "loss": 0.2341,
+      "step": 54036
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8524675499135225e-06,
+      "loss": 0.2062,
+      "step": 54037
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.852131714607157e-06,
+      "loss": 0.245,
+      "step": 54038
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.851795879300791e-06,
+      "loss": 0.2755,
+      "step": 54039
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8514600439944253e-06,
+      "loss": 0.3318,
+      "step": 54040
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8511242086880596e-06,
+      "loss": 0.1143,
+      "step": 54041
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8507883733816936e-06,
+      "loss": 0.17,
+      "step": 54042
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.850452538075328e-06,
+      "loss": 0.2195,
+      "step": 54043
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8501167027689622e-06,
+      "loss": 0.23,
+      "step": 54044
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8497808674625965e-06,
+      "loss": 0.1829,
+      "step": 54045
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8494450321562307e-06,
+      "loss": 0.23,
+      "step": 54046
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8491091968498651e-06,
+      "loss": 0.3282,
+      "step": 54047
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8487733615434991e-06,
+      "loss": 0.2104,
+      "step": 54048
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8484375262371334e-06,
+      "loss": 0.2449,
+      "step": 54049
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8481016909307678e-06,
+      "loss": 0.1197,
+      "step": 54050
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8477658556244018e-06,
+      "loss": 0.1818,
+      "step": 54051
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8474300203180362e-06,
+      "loss": 0.2519,
+      "step": 54052
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8470941850116705e-06,
+      "loss": 0.1536,
+      "step": 54053
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8467583497053047e-06,
+      "loss": 0.2332,
+      "step": 54054
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.846422514398939e-06,
+      "loss": 0.2058,
+      "step": 54055
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8460866790925733e-06,
+      "loss": 0.3145,
+      "step": 54056
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8457508437862074e-06,
+      "loss": 0.3411,
+      "step": 54057
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8454150084798416e-06,
+      "loss": 0.2879,
+      "step": 54058
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.845079173173476e-06,
+      "loss": 0.319,
+      "step": 54059
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.84474333786711e-06,
+      "loss": 0.297,
+      "step": 54060
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8444075025607445e-06,
+      "loss": 0.1573,
+      "step": 54061
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8440716672543785e-06,
+      "loss": 0.2748,
+      "step": 54062
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.843735831948013e-06,
+      "loss": 0.2221,
+      "step": 54063
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8433999966416471e-06,
+      "loss": 0.2383,
+      "step": 54064
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8430641613352812e-06,
+      "loss": 0.1514,
+      "step": 54065
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8427283260289156e-06,
+      "loss": 0.2618,
+      "step": 54066
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8423924907225498e-06,
+      "loss": 0.2648,
+      "step": 54067
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.842056655416184e-06,
+      "loss": 0.2428,
+      "step": 54068
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8417208201098183e-06,
+      "loss": 0.2112,
+      "step": 54069
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8413849848034527e-06,
+      "loss": 0.2218,
+      "step": 54070
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8410491494970867e-06,
+      "loss": 0.311,
+      "step": 54071
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.840713314190721e-06,
+      "loss": 0.2501,
+      "step": 54072
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8403774788843554e-06,
+      "loss": 0.3465,
+      "step": 54073
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8400416435779894e-06,
+      "loss": 0.3158,
+      "step": 54074
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8397058082716238e-06,
+      "loss": 0.2193,
+      "step": 54075
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.839369972965258e-06,
+      "loss": 0.1861,
+      "step": 54076
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8390341376588923e-06,
+      "loss": 0.2919,
+      "step": 54077
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8386983023525265e-06,
+      "loss": 0.1995,
+      "step": 54078
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.838362467046161e-06,
+      "loss": 0.2891,
+      "step": 54079
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.838026631739795e-06,
+      "loss": 0.3624,
+      "step": 54080
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8376907964334292e-06,
+      "loss": 0.2053,
+      "step": 54081
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8373549611270636e-06,
+      "loss": 0.3125,
+      "step": 54082
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8370191258206976e-06,
+      "loss": 0.3753,
+      "step": 54083
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.836683290514332e-06,
+      "loss": 0.2177,
+      "step": 54084
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.836347455207966e-06,
+      "loss": 0.204,
+      "step": 54085
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8360116199016003e-06,
+      "loss": 0.2018,
+      "step": 54086
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8356757845952347e-06,
+      "loss": 0.2384,
+      "step": 54087
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8353399492888687e-06,
+      "loss": 0.3467,
+      "step": 54088
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8350041139825032e-06,
+      "loss": 0.3152,
+      "step": 54089
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8346682786761374e-06,
+      "loss": 0.2129,
+      "step": 54090
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8343324433697716e-06,
+      "loss": 0.3759,
+      "step": 54091
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8339966080634058e-06,
+      "loss": 0.1439,
+      "step": 54092
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.8336607727570403e-06,
+      "loss": 0.1399,
+      "step": 54093
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8333249374506743e-06,
+      "loss": 0.3564,
+      "step": 54094
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8329891021443085e-06,
+      "loss": 0.2,
+      "step": 54095
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.832653266837943e-06,
+      "loss": 0.1244,
+      "step": 54096
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.832317431531577e-06,
+      "loss": 0.374,
+      "step": 54097
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8319815962252114e-06,
+      "loss": 0.3404,
+      "step": 54098
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8316457609188456e-06,
+      "loss": 0.2535,
+      "step": 54099
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8313099256124796e-06,
+      "loss": 0.2734,
+      "step": 54100
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.830974090306114e-06,
+      "loss": 0.2087,
+      "step": 54101
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8306382549997485e-06,
+      "loss": 0.1669,
+      "step": 54102
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8303024196933825e-06,
+      "loss": 0.232,
+      "step": 54103
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8299665843870167e-06,
+      "loss": 0.3268,
+      "step": 54104
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8296307490806512e-06,
+      "loss": 0.1623,
+      "step": 54105
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8292949137742852e-06,
+      "loss": 0.1163,
+      "step": 54106
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8289590784679196e-06,
+      "loss": 0.2723,
+      "step": 54107
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8286232431615536e-06,
+      "loss": 0.1533,
+      "step": 54108
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8282874078551878e-06,
+      "loss": 0.2494,
+      "step": 54109
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8279515725488223e-06,
+      "loss": 0.3462,
+      "step": 54110
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8276157372424563e-06,
+      "loss": 0.4284,
+      "step": 54111
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8272799019360907e-06,
+      "loss": 0.2261,
+      "step": 54112
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.826944066629725e-06,
+      "loss": 0.1003,
+      "step": 54113
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.826608231323359e-06,
+      "loss": 0.1821,
+      "step": 54114
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8262723960169934e-06,
+      "loss": 0.2055,
+      "step": 54115
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8259365607106278e-06,
+      "loss": 0.166,
+      "step": 54116
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8256007254042619e-06,
+      "loss": 0.2045,
+      "step": 54117
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.825264890097896e-06,
+      "loss": 0.3503,
+      "step": 54118
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8249290547915305e-06,
+      "loss": 0.1996,
+      "step": 54119
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8245932194851645e-06,
+      "loss": 0.1678,
+      "step": 54120
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.824257384178799e-06,
+      "loss": 0.4232,
+      "step": 54121
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8239215488724332e-06,
+      "loss": 0.1728,
+      "step": 54122
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8235857135660672e-06,
+      "loss": 0.2541,
+      "step": 54123
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8232498782597016e-06,
+      "loss": 0.2945,
+      "step": 54124
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.822914042953336e-06,
+      "loss": 0.3984,
+      "step": 54125
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.82257820764697e-06,
+      "loss": 0.128,
+      "step": 54126
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8222423723406043e-06,
+      "loss": 0.2511,
+      "step": 54127
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8219065370342385e-06,
+      "loss": 0.2991,
+      "step": 54128
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8215707017278727e-06,
+      "loss": 0.3053,
+      "step": 54129
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8212348664215072e-06,
+      "loss": 0.2849,
+      "step": 54130
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8208990311151412e-06,
+      "loss": 0.2136,
+      "step": 54131
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8205631958087754e-06,
+      "loss": 0.3427,
+      "step": 54132
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8202273605024099e-06,
+      "loss": 0.328,
+      "step": 54133
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8198915251960439e-06,
+      "loss": 0.1922,
+      "step": 54134
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8195556898896783e-06,
+      "loss": 0.26,
+      "step": 54135
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8192198545833125e-06,
+      "loss": 0.2918,
+      "step": 54136
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8188840192769465e-06,
+      "loss": 0.2047,
+      "step": 54137
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.818548183970581e-06,
+      "loss": 0.2963,
+      "step": 54138
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8182123486642154e-06,
+      "loss": 0.2313,
+      "step": 54139
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8178765133578494e-06,
+      "loss": 0.2976,
+      "step": 54140
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8175406780514836e-06,
+      "loss": 0.1655,
+      "step": 54141
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.817204842745118e-06,
+      "loss": 0.2607,
+      "step": 54142
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.816869007438752e-06,
+      "loss": 0.1481,
+      "step": 54143
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8165331721323865e-06,
+      "loss": 0.318,
+      "step": 54144
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8161973368260208e-06,
+      "loss": 0.2608,
+      "step": 54145
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8158615015196548e-06,
+      "loss": 0.3861,
+      "step": 54146
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8155256662132892e-06,
+      "loss": 0.2971,
+      "step": 54147
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8151898309069234e-06,
+      "loss": 0.3681,
+      "step": 54148
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8148539956005577e-06,
+      "loss": 0.3441,
+      "step": 54149
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8145181602941919e-06,
+      "loss": 0.253,
+      "step": 54150
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8141823249878259e-06,
+      "loss": 0.3331,
+      "step": 54151
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8138464896814603e-06,
+      "loss": 0.5871,
+      "step": 54152
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8135106543750948e-06,
+      "loss": 0.1931,
+      "step": 54153
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8131748190687288e-06,
+      "loss": 0.2452,
+      "step": 54154
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.812838983762363e-06,
+      "loss": 0.258,
+      "step": 54155
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8125031484559974e-06,
+      "loss": 0.4051,
+      "step": 54156
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8121673131496314e-06,
+      "loss": 0.3369,
+      "step": 54157
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8118314778432659e-06,
+      "loss": 0.3924,
+      "step": 54158
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8114956425369e-06,
+      "loss": 0.3902,
+      "step": 54159
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8111598072305341e-06,
+      "loss": 0.1941,
+      "step": 54160
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8108239719241685e-06,
+      "loss": 0.2549,
+      "step": 54161
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8104881366178028e-06,
+      "loss": 0.3917,
+      "step": 54162
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.810152301311437e-06,
+      "loss": 0.1479,
+      "step": 54163
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8098164660050712e-06,
+      "loss": 0.3896,
+      "step": 54164
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8094806306987057e-06,
+      "loss": 0.1935,
+      "step": 54165
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8091447953923397e-06,
+      "loss": 0.2956,
+      "step": 54166
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.808808960085974e-06,
+      "loss": 0.3319,
+      "step": 54167
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8084731247796083e-06,
+      "loss": 0.3786,
+      "step": 54168
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8081372894732423e-06,
+      "loss": 0.2275,
+      "step": 54169
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8078014541668768e-06,
+      "loss": 0.1752,
+      "step": 54170
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.807465618860511e-06,
+      "loss": 0.2364,
+      "step": 54171
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8071297835541452e-06,
+      "loss": 0.1987,
+      "step": 54172
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8067939482477794e-06,
+      "loss": 0.256,
+      "step": 54173
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8064581129414135e-06,
+      "loss": 0.184,
+      "step": 54174
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.806122277635048e-06,
+      "loss": 0.2584,
+      "step": 54175
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8057864423286821e-06,
+      "loss": 0.3339,
+      "step": 54176
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8054506070223163e-06,
+      "loss": 0.3115,
+      "step": 54177
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8051147717159506e-06,
+      "loss": 0.2142,
+      "step": 54178
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.804778936409585e-06,
+      "loss": 0.1823,
+      "step": 54179
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.804443101103219e-06,
+      "loss": 0.3307,
+      "step": 54180
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8041072657968534e-06,
+      "loss": 0.2093,
+      "step": 54181
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8037714304904877e-06,
+      "loss": 0.3081,
+      "step": 54182
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8034355951841217e-06,
+      "loss": 0.3372,
+      "step": 54183
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8030997598777561e-06,
+      "loss": 0.3392,
+      "step": 54184
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8027639245713903e-06,
+      "loss": 0.1858,
+      "step": 54185
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8024280892650246e-06,
+      "loss": 0.2617,
+      "step": 54186
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8020922539586588e-06,
+      "loss": 0.2459,
+      "step": 54187
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8017564186522932e-06,
+      "loss": 0.3369,
+      "step": 54188
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8014205833459272e-06,
+      "loss": 0.2878,
+      "step": 54189
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8010847480395615e-06,
+      "loss": 0.1905,
+      "step": 54190
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.800748912733196e-06,
+      "loss": 0.2253,
+      "step": 54191
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.80041307742683e-06,
+      "loss": 0.325,
+      "step": 54192
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.8000772421204643e-06,
+      "loss": 0.3374,
+      "step": 54193
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7997414068140986e-06,
+      "loss": 0.1493,
+      "step": 54194
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7994055715077328e-06,
+      "loss": 0.1335,
+      "step": 54195
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.799069736201367e-06,
+      "loss": 0.3529,
+      "step": 54196
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.798733900895001e-06,
+      "loss": 0.3664,
+      "step": 54197
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7983980655886355e-06,
+      "loss": 0.1838,
+      "step": 54198
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7980622302822697e-06,
+      "loss": 0.2215,
+      "step": 54199
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.797726394975904e-06,
+      "loss": 0.135,
+      "step": 54200
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7973905596695381e-06,
+      "loss": 0.2954,
+      "step": 54201
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7970547243631726e-06,
+      "loss": 0.1564,
+      "step": 54202
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7967188890568066e-06,
+      "loss": 0.1733,
+      "step": 54203
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.796383053750441e-06,
+      "loss": 0.1969,
+      "step": 54204
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7960472184440752e-06,
+      "loss": 0.1617,
+      "step": 54205
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7957113831377093e-06,
+      "loss": 0.1932,
+      "step": 54206
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7953755478313437e-06,
+      "loss": 0.3839,
+      "step": 54207
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.795039712524978e-06,
+      "loss": 0.4044,
+      "step": 54208
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7947038772186121e-06,
+      "loss": 0.3037,
+      "step": 54209
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7943680419122464e-06,
+      "loss": 0.2274,
+      "step": 54210
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7940322066058808e-06,
+      "loss": 0.2587,
+      "step": 54211
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7936963712995148e-06,
+      "loss": 0.167,
+      "step": 54212
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.793360535993149e-06,
+      "loss": 0.3506,
+      "step": 54213
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7930247006867835e-06,
+      "loss": 0.1925,
+      "step": 54214
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7926888653804175e-06,
+      "loss": 0.3948,
+      "step": 54215
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.792353030074052e-06,
+      "loss": 0.2209,
+      "step": 54216
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7920171947676861e-06,
+      "loss": 0.2166,
+      "step": 54217
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7916813594613204e-06,
+      "loss": 0.2803,
+      "step": 54218
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7913455241549546e-06,
+      "loss": 0.3475,
+      "step": 54219
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7910096888485886e-06,
+      "loss": 0.2358,
+      "step": 54220
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.790673853542223e-06,
+      "loss": 0.2291,
+      "step": 54221
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7903380182358573e-06,
+      "loss": 0.3315,
+      "step": 54222
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7900021829294915e-06,
+      "loss": 0.5085,
+      "step": 54223
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7896663476231257e-06,
+      "loss": 0.2094,
+      "step": 54224
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7893305123167601e-06,
+      "loss": 0.1423,
+      "step": 54225
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7889946770103942e-06,
+      "loss": 0.2461,
+      "step": 54226
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7886588417040284e-06,
+      "loss": 0.2248,
+      "step": 54227
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7883230063976628e-06,
+      "loss": 0.2804,
+      "step": 54228
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7879871710912968e-06,
+      "loss": 0.199,
+      "step": 54229
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7876513357849313e-06,
+      "loss": 0.246,
+      "step": 54230
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7873155004785655e-06,
+      "loss": 0.2493,
+      "step": 54231
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7869796651721997e-06,
+      "loss": 0.2456,
+      "step": 54232
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.786643829865834e-06,
+      "loss": 0.2422,
+      "step": 54233
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7863079945594684e-06,
+      "loss": 0.2817,
+      "step": 54234
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7859721592531024e-06,
+      "loss": 0.3316,
+      "step": 54235
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7856363239467366e-06,
+      "loss": 0.3344,
+      "step": 54236
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.785300488640371e-06,
+      "loss": 0.1895,
+      "step": 54237
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.784964653334005e-06,
+      "loss": 0.1871,
+      "step": 54238
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7846288180276395e-06,
+      "loss": 0.3469,
+      "step": 54239
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7842929827212737e-06,
+      "loss": 0.2089,
+      "step": 54240
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7839571474149077e-06,
+      "loss": 0.2207,
+      "step": 54241
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7836213121085422e-06,
+      "loss": 0.3337,
+      "step": 54242
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7832854768021762e-06,
+      "loss": 0.2356,
+      "step": 54243
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7829496414958106e-06,
+      "loss": 0.345,
+      "step": 54244
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7826138061894448e-06,
+      "loss": 0.3278,
+      "step": 54245
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.782277970883079e-06,
+      "loss": 0.194,
+      "step": 54246
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7819421355767133e-06,
+      "loss": 0.2179,
+      "step": 54247
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7816063002703477e-06,
+      "loss": 0.3102,
+      "step": 54248
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7812704649639817e-06,
+      "loss": 0.2495,
+      "step": 54249
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.780934629657616e-06,
+      "loss": 0.1843,
+      "step": 54250
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7805987943512504e-06,
+      "loss": 0.2496,
+      "step": 54251
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7802629590448844e-06,
+      "loss": 0.203,
+      "step": 54252
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7799271237385188e-06,
+      "loss": 0.3253,
+      "step": 54253
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.779591288432153e-06,
+      "loss": 0.2752,
+      "step": 54254
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.779255453125787e-06,
+      "loss": 0.2768,
+      "step": 54255
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7789196178194215e-06,
+      "loss": 0.3844,
+      "step": 54256
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.778583782513056e-06,
+      "loss": 0.2519,
+      "step": 54257
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.77824794720669e-06,
+      "loss": 0.5593,
+      "step": 54258
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7779121119003242e-06,
+      "loss": 0.2682,
+      "step": 54259
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7775762765939586e-06,
+      "loss": 0.177,
+      "step": 54260
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7772404412875926e-06,
+      "loss": 0.1878,
+      "step": 54261
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.776904605981227e-06,
+      "loss": 0.3247,
+      "step": 54262
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7765687706748613e-06,
+      "loss": 0.1543,
+      "step": 54263
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7762329353684953e-06,
+      "loss": 0.2357,
+      "step": 54264
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7758971000621297e-06,
+      "loss": 0.2652,
+      "step": 54265
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7755612647557637e-06,
+      "loss": 0.1821,
+      "step": 54266
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7752254294493982e-06,
+      "loss": 0.1992,
+      "step": 54267
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7748895941430324e-06,
+      "loss": 0.2527,
+      "step": 54268
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7745537588366666e-06,
+      "loss": 0.1268,
+      "step": 54269
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7742179235303009e-06,
+      "loss": 0.3006,
+      "step": 54270
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7738820882239353e-06,
+      "loss": 0.2727,
+      "step": 54271
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7735462529175693e-06,
+      "loss": 0.3363,
+      "step": 54272
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7732104176112035e-06,
+      "loss": 0.2081,
+      "step": 54273
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.772874582304838e-06,
+      "loss": 0.1677,
+      "step": 54274
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.772538746998472e-06,
+      "loss": 0.123,
+      "step": 54275
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7722029116921064e-06,
+      "loss": 0.3376,
+      "step": 54276
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7718670763857406e-06,
+      "loss": 0.2884,
+      "step": 54277
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7715312410793746e-06,
+      "loss": 0.1595,
+      "step": 54278
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.771195405773009e-06,
+      "loss": 0.3963,
+      "step": 54279
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7708595704666435e-06,
+      "loss": 0.3609,
+      "step": 54280
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7705237351602775e-06,
+      "loss": 0.2468,
+      "step": 54281
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7701878998539118e-06,
+      "loss": 0.274,
+      "step": 54282
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7698520645475462e-06,
+      "loss": 0.286,
+      "step": 54283
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7695162292411802e-06,
+      "loss": 0.2814,
+      "step": 54284
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7691803939348146e-06,
+      "loss": 0.2246,
+      "step": 54285
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7688445586284486e-06,
+      "loss": 0.3272,
+      "step": 54286
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7685087233220829e-06,
+      "loss": 0.2235,
+      "step": 54287
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7681728880157173e-06,
+      "loss": 0.2314,
+      "step": 54288
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7678370527093513e-06,
+      "loss": 0.2215,
+      "step": 54289
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.7675012174029858e-06,
+      "loss": 0.359,
+      "step": 54290
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.76716538209662e-06,
+      "loss": 0.1316,
+      "step": 54291
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 1.766829546790254e-06,
+      "loss": 0.337,
+      "step": 54292
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7664937114838884e-06,
+      "loss": 0.1502,
+      "step": 54293
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7661578761775229e-06,
+      "loss": 0.299,
+      "step": 54294
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7658220408711569e-06,
+      "loss": 0.2605,
+      "step": 54295
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.765486205564791e-06,
+      "loss": 0.2589,
+      "step": 54296
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7651503702584255e-06,
+      "loss": 0.2908,
+      "step": 54297
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7648145349520595e-06,
+      "loss": 0.2148,
+      "step": 54298
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.764478699645694e-06,
+      "loss": 0.1253,
+      "step": 54299
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7641428643393282e-06,
+      "loss": 0.5281,
+      "step": 54300
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7638070290329622e-06,
+      "loss": 0.1386,
+      "step": 54301
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7634711937265967e-06,
+      "loss": 0.1696,
+      "step": 54302
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7631353584202309e-06,
+      "loss": 0.2129,
+      "step": 54303
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.762799523113865e-06,
+      "loss": 0.3611,
+      "step": 54304
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7624636878074993e-06,
+      "loss": 0.2368,
+      "step": 54305
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7621278525011338e-06,
+      "loss": 0.1952,
+      "step": 54306
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7617920171947678e-06,
+      "loss": 0.1701,
+      "step": 54307
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7614561818884022e-06,
+      "loss": 0.146,
+      "step": 54308
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7611203465820362e-06,
+      "loss": 0.2396,
+      "step": 54309
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7607845112756704e-06,
+      "loss": 0.3995,
+      "step": 54310
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7604486759693049e-06,
+      "loss": 0.3003,
+      "step": 54311
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7601128406629389e-06,
+      "loss": 0.2653,
+      "step": 54312
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7597770053565733e-06,
+      "loss": 0.2568,
+      "step": 54313
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7594411700502076e-06,
+      "loss": 0.3473,
+      "step": 54314
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7591053347438416e-06,
+      "loss": 0.2188,
+      "step": 54315
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.758769499437476e-06,
+      "loss": 0.2097,
+      "step": 54316
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7584336641311102e-06,
+      "loss": 0.2797,
+      "step": 54317
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7580978288247444e-06,
+      "loss": 0.1387,
+      "step": 54318
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7577619935183787e-06,
+      "loss": 0.3077,
+      "step": 54319
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7574261582120131e-06,
+      "loss": 0.4748,
+      "step": 54320
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7570903229056471e-06,
+      "loss": 0.2855,
+      "step": 54321
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7567544875992816e-06,
+      "loss": 0.3065,
+      "step": 54322
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7564186522929158e-06,
+      "loss": 0.1421,
+      "step": 54323
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7560828169865498e-06,
+      "loss": 0.2388,
+      "step": 54324
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7557469816801842e-06,
+      "loss": 0.3868,
+      "step": 54325
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7554111463738185e-06,
+      "loss": 0.4646,
+      "step": 54326
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7550753110674527e-06,
+      "loss": 0.2367,
+      "step": 54327
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.754739475761087e-06,
+      "loss": 0.2876,
+      "step": 54328
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7544036404547213e-06,
+      "loss": 0.3926,
+      "step": 54329
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7540678051483553e-06,
+      "loss": 0.1962,
+      "step": 54330
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7537319698419898e-06,
+      "loss": 0.2717,
+      "step": 54331
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7533961345356238e-06,
+      "loss": 0.144,
+      "step": 54332
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.753060299229258e-06,
+      "loss": 0.4282,
+      "step": 54333
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7527244639228925e-06,
+      "loss": 0.3139,
+      "step": 54334
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7523886286165265e-06,
+      "loss": 0.4055,
+      "step": 54335
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.752052793310161e-06,
+      "loss": 0.1672,
+      "step": 54336
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7517169580037951e-06,
+      "loss": 0.2357,
+      "step": 54337
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7513811226974291e-06,
+      "loss": 0.3335,
+      "step": 54338
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7510452873910636e-06,
+      "loss": 0.1596,
+      "step": 54339
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7507094520846978e-06,
+      "loss": 0.1967,
+      "step": 54340
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.750373616778332e-06,
+      "loss": 0.1516,
+      "step": 54341
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7500377814719662e-06,
+      "loss": 0.1155,
+      "step": 54342
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7497019461656007e-06,
+      "loss": 0.3112,
+      "step": 54343
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7493661108592347e-06,
+      "loss": 0.2529,
+      "step": 54344
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7490302755528691e-06,
+      "loss": 0.2379,
+      "step": 54345
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7486944402465034e-06,
+      "loss": 0.1593,
+      "step": 54346
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7483586049401374e-06,
+      "loss": 0.2639,
+      "step": 54347
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7480227696337718e-06,
+      "loss": 0.2223,
+      "step": 54348
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.747686934327406e-06,
+      "loss": 0.2858,
+      "step": 54349
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7473510990210402e-06,
+      "loss": 0.2348,
+      "step": 54350
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7470152637146745e-06,
+      "loss": 0.2939,
+      "step": 54351
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.746679428408309e-06,
+      "loss": 0.3079,
+      "step": 54352
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.746343593101943e-06,
+      "loss": 0.3678,
+      "step": 54353
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7460077577955771e-06,
+      "loss": 0.2216,
+      "step": 54354
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7456719224892114e-06,
+      "loss": 0.3286,
+      "step": 54355
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7453360871828456e-06,
+      "loss": 0.3532,
+      "step": 54356
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.74500025187648e-06,
+      "loss": 0.1257,
+      "step": 54357
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.744664416570114e-06,
+      "loss": 0.2111,
+      "step": 54358
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7443285812637485e-06,
+      "loss": 0.1881,
+      "step": 54359
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7439927459573827e-06,
+      "loss": 0.2598,
+      "step": 54360
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7436569106510167e-06,
+      "loss": 0.3829,
+      "step": 54361
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7433210753446511e-06,
+      "loss": 0.267,
+      "step": 54362
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7429852400382854e-06,
+      "loss": 0.4843,
+      "step": 54363
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7426494047319196e-06,
+      "loss": 0.1721,
+      "step": 54364
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7423135694255538e-06,
+      "loss": 0.3918,
+      "step": 54365
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7419777341191883e-06,
+      "loss": 0.3625,
+      "step": 54366
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7416418988128223e-06,
+      "loss": 0.2993,
+      "step": 54367
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7413060635064565e-06,
+      "loss": 0.2708,
+      "step": 54368
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.740970228200091e-06,
+      "loss": 0.2876,
+      "step": 54369
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.740634392893725e-06,
+      "loss": 0.2325,
+      "step": 54370
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7402985575873594e-06,
+      "loss": 0.1718,
+      "step": 54371
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7399627222809936e-06,
+      "loss": 0.2232,
+      "step": 54372
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7396268869746278e-06,
+      "loss": 0.208,
+      "step": 54373
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.739291051668262e-06,
+      "loss": 0.263,
+      "step": 54374
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7389552163618965e-06,
+      "loss": 0.3116,
+      "step": 54375
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7386193810555305e-06,
+      "loss": 0.35,
+      "step": 54376
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7382835457491647e-06,
+      "loss": 0.2545,
+      "step": 54377
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.737947710442799e-06,
+      "loss": 0.3601,
+      "step": 54378
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7376118751364332e-06,
+      "loss": 0.2464,
+      "step": 54379
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7372760398300676e-06,
+      "loss": 0.1595,
+      "step": 54380
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7369402045237016e-06,
+      "loss": 0.3326,
+      "step": 54381
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7366043692173358e-06,
+      "loss": 0.1278,
+      "step": 54382
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7362685339109703e-06,
+      "loss": 0.2899,
+      "step": 54383
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7359326986046043e-06,
+      "loss": 0.2125,
+      "step": 54384
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7355968632982387e-06,
+      "loss": 0.3283,
+      "step": 54385
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.735261027991873e-06,
+      "loss": 0.2308,
+      "step": 54386
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7349251926855072e-06,
+      "loss": 0.1384,
+      "step": 54387
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7345893573791414e-06,
+      "loss": 0.2724,
+      "step": 54388
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7342535220727758e-06,
+      "loss": 0.2185,
+      "step": 54389
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7339176867664098e-06,
+      "loss": 0.3063,
+      "step": 54390
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.733581851460044e-06,
+      "loss": 0.2738,
+      "step": 54391
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7332460161536785e-06,
+      "loss": 0.2581,
+      "step": 54392
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7329101808473125e-06,
+      "loss": 0.3983,
+      "step": 54393
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.732574345540947e-06,
+      "loss": 0.2681,
+      "step": 54394
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7322385102345812e-06,
+      "loss": 0.279,
+      "step": 54395
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7319026749282154e-06,
+      "loss": 0.1664,
+      "step": 54396
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7315668396218496e-06,
+      "loss": 0.4241,
+      "step": 54397
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.731231004315484e-06,
+      "loss": 0.1486,
+      "step": 54398
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.730895169009118e-06,
+      "loss": 0.3488,
+      "step": 54399
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7305593337027523e-06,
+      "loss": 0.2243,
+      "step": 54400
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7302234983963865e-06,
+      "loss": 0.1478,
+      "step": 54401
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7298876630900207e-06,
+      "loss": 0.3704,
+      "step": 54402
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7295518277836552e-06,
+      "loss": 0.3205,
+      "step": 54403
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7292159924772892e-06,
+      "loss": 0.2059,
+      "step": 54404
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7288801571709234e-06,
+      "loss": 0.2255,
+      "step": 54405
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7285443218645578e-06,
+      "loss": 0.1052,
+      "step": 54406
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7282084865581919e-06,
+      "loss": 0.1966,
+      "step": 54407
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7278726512518263e-06,
+      "loss": 0.2436,
+      "step": 54408
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7275368159454605e-06,
+      "loss": 0.3646,
+      "step": 54409
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7272009806390947e-06,
+      "loss": 0.2833,
+      "step": 54410
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.726865145332729e-06,
+      "loss": 0.2441,
+      "step": 54411
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7265293100263634e-06,
+      "loss": 0.2978,
+      "step": 54412
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7261934747199974e-06,
+      "loss": 0.3491,
+      "step": 54413
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7258576394136316e-06,
+      "loss": 0.2417,
+      "step": 54414
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.725521804107266e-06,
+      "loss": 0.3449,
+      "step": 54415
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7251859688009e-06,
+      "loss": 0.2874,
+      "step": 54416
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7248501334945345e-06,
+      "loss": 0.1954,
+      "step": 54417
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7245142981881687e-06,
+      "loss": 0.3506,
+      "step": 54418
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7241784628818028e-06,
+      "loss": 0.2244,
+      "step": 54419
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7238426275754372e-06,
+      "loss": 0.3843,
+      "step": 54420
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7235067922690716e-06,
+      "loss": 0.253,
+      "step": 54421
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7231709569627056e-06,
+      "loss": 0.1974,
+      "step": 54422
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7228351216563399e-06,
+      "loss": 0.2639,
+      "step": 54423
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.722499286349974e-06,
+      "loss": 0.2637,
+      "step": 54424
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7221634510436083e-06,
+      "loss": 0.3823,
+      "step": 54425
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7218276157372427e-06,
+      "loss": 0.2657,
+      "step": 54426
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7214917804308768e-06,
+      "loss": 0.2714,
+      "step": 54427
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.721155945124511e-06,
+      "loss": 0.2923,
+      "step": 54428
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7208201098181454e-06,
+      "loss": 0.3151,
+      "step": 54429
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7204842745117794e-06,
+      "loss": 0.1771,
+      "step": 54430
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7201484392054139e-06,
+      "loss": 0.17,
+      "step": 54431
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.719812603899048e-06,
+      "loss": 0.1678,
+      "step": 54432
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.719476768592682e-06,
+      "loss": 0.2091,
+      "step": 54433
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7191409332863165e-06,
+      "loss": 0.2133,
+      "step": 54434
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.718805097979951e-06,
+      "loss": 0.3966,
+      "step": 54435
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.718469262673585e-06,
+      "loss": 0.238,
+      "step": 54436
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7181334273672192e-06,
+      "loss": 0.2769,
+      "step": 54437
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7177975920608536e-06,
+      "loss": 0.1289,
+      "step": 54438
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7174617567544877e-06,
+      "loss": 0.2126,
+      "step": 54439
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.717125921448122e-06,
+      "loss": 0.1848,
+      "step": 54440
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7167900861417563e-06,
+      "loss": 0.183,
+      "step": 54441
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7164542508353903e-06,
+      "loss": 0.2986,
+      "step": 54442
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7161184155290248e-06,
+      "loss": 0.361,
+      "step": 54443
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7157825802226588e-06,
+      "loss": 0.2613,
+      "step": 54444
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7154467449162932e-06,
+      "loss": 0.2155,
+      "step": 54445
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7151109096099274e-06,
+      "loss": 0.1372,
+      "step": 54446
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7147750743035614e-06,
+      "loss": 0.211,
+      "step": 54447
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7144392389971959e-06,
+      "loss": 0.3363,
+      "step": 54448
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7141034036908303e-06,
+      "loss": 0.2959,
+      "step": 54449
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7137675683844643e-06,
+      "loss": 0.2475,
+      "step": 54450
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7134317330780986e-06,
+      "loss": 0.2425,
+      "step": 54451
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.713095897771733e-06,
+      "loss": 0.2436,
+      "step": 54452
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.712760062465367e-06,
+      "loss": 0.1872,
+      "step": 54453
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7124242271590014e-06,
+      "loss": 0.1554,
+      "step": 54454
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7120883918526357e-06,
+      "loss": 0.2102,
+      "step": 54455
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7117525565462697e-06,
+      "loss": 0.2214,
+      "step": 54456
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7114167212399041e-06,
+      "loss": 0.365,
+      "step": 54457
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7110808859335385e-06,
+      "loss": 0.2305,
+      "step": 54458
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7107450506271726e-06,
+      "loss": 0.2237,
+      "step": 54459
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7104092153208068e-06,
+      "loss": 0.2684,
+      "step": 54460
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7100733800144412e-06,
+      "loss": 0.218,
+      "step": 54461
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7097375447080752e-06,
+      "loss": 0.2873,
+      "step": 54462
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7094017094017097e-06,
+      "loss": 0.4437,
+      "step": 54463
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7090658740953439e-06,
+      "loss": 0.2885,
+      "step": 54464
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.708730038788978e-06,
+      "loss": 0.3737,
+      "step": 54465
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7083942034826123e-06,
+      "loss": 0.3024,
+      "step": 54466
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7080583681762463e-06,
+      "loss": 0.1915,
+      "step": 54467
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7077225328698808e-06,
+      "loss": 0.2835,
+      "step": 54468
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.707386697563515e-06,
+      "loss": 0.2526,
+      "step": 54469
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.707050862257149e-06,
+      "loss": 0.1724,
+      "step": 54470
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7067150269507835e-06,
+      "loss": 0.2009,
+      "step": 54471
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7063791916444179e-06,
+      "loss": 0.2554,
+      "step": 54472
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.706043356338052e-06,
+      "loss": 0.2528,
+      "step": 54473
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7057075210316861e-06,
+      "loss": 0.2114,
+      "step": 54474
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7053716857253206e-06,
+      "loss": 0.3172,
+      "step": 54475
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7050358504189546e-06,
+      "loss": 0.3232,
+      "step": 54476
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.704700015112589e-06,
+      "loss": 0.4585,
+      "step": 54477
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7043641798062232e-06,
+      "loss": 0.4372,
+      "step": 54478
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7040283444998572e-06,
+      "loss": 0.389,
+      "step": 54479
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7036925091934917e-06,
+      "loss": 0.2594,
+      "step": 54480
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.703356673887126e-06,
+      "loss": 0.1373,
+      "step": 54481
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7030208385807601e-06,
+      "loss": 0.1285,
+      "step": 54482
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7026850032743944e-06,
+      "loss": 0.3332,
+      "step": 54483
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7023491679680288e-06,
+      "loss": 0.1977,
+      "step": 54484
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7020133326616628e-06,
+      "loss": 0.2638,
+      "step": 54485
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7016774973552972e-06,
+      "loss": 0.2743,
+      "step": 54486
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7013416620489315e-06,
+      "loss": 0.4765,
+      "step": 54487
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7010058267425655e-06,
+      "loss": 0.2658,
+      "step": 54488
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.7006699914362e-06,
+      "loss": 0.4023,
+      "step": 54489
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 1.700334156129834e-06,
+      "loss": 0.129,
+      "step": 54490
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6999983208234684e-06,
+      "loss": 0.3485,
+      "step": 54491
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6996624855171026e-06,
+      "loss": 0.2952,
+      "step": 54492
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6993266502107366e-06,
+      "loss": 0.2448,
+      "step": 54493
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.698990814904371e-06,
+      "loss": 0.2961,
+      "step": 54494
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6986549795980053e-06,
+      "loss": 0.1191,
+      "step": 54495
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6983191442916395e-06,
+      "loss": 0.3526,
+      "step": 54496
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6979833089852737e-06,
+      "loss": 0.2593,
+      "step": 54497
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6976474736789081e-06,
+      "loss": 0.2206,
+      "step": 54498
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6973116383725421e-06,
+      "loss": 0.2721,
+      "step": 54499
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6969758030661766e-06,
+      "loss": 0.3484,
+      "step": 54500
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6966399677598108e-06,
+      "loss": 0.3358,
+      "step": 54501
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6963041324534448e-06,
+      "loss": 0.2008,
+      "step": 54502
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6959682971470793e-06,
+      "loss": 0.3554,
+      "step": 54503
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6956324618407135e-06,
+      "loss": 0.2893,
+      "step": 54504
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6952966265343477e-06,
+      "loss": 0.3221,
+      "step": 54505
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.694960791227982e-06,
+      "loss": 0.4088,
+      "step": 54506
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6946249559216164e-06,
+      "loss": 0.2758,
+      "step": 54507
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6942891206152504e-06,
+      "loss": 0.3365,
+      "step": 54508
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6939532853088846e-06,
+      "loss": 0.1702,
+      "step": 54509
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.693617450002519e-06,
+      "loss": 0.3337,
+      "step": 54510
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.693281614696153e-06,
+      "loss": 0.1649,
+      "step": 54511
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6929457793897875e-06,
+      "loss": 0.1871,
+      "step": 54512
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6926099440834215e-06,
+      "loss": 0.222,
+      "step": 54513
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.692274108777056e-06,
+      "loss": 0.1505,
+      "step": 54514
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6919382734706902e-06,
+      "loss": 0.2992,
+      "step": 54515
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6916024381643242e-06,
+      "loss": 0.2345,
+      "step": 54516
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6912666028579586e-06,
+      "loss": 0.243,
+      "step": 54517
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6909307675515928e-06,
+      "loss": 0.1422,
+      "step": 54518
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.690594932245227e-06,
+      "loss": 0.256,
+      "step": 54519
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6902590969388613e-06,
+      "loss": 0.2581,
+      "step": 54520
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6899232616324957e-06,
+      "loss": 0.1199,
+      "step": 54521
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6895874263261297e-06,
+      "loss": 0.1978,
+      "step": 54522
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.689251591019764e-06,
+      "loss": 0.3641,
+      "step": 54523
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6889157557133984e-06,
+      "loss": 0.3074,
+      "step": 54524
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6885799204070324e-06,
+      "loss": 0.2637,
+      "step": 54525
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6882440851006668e-06,
+      "loss": 0.1915,
+      "step": 54526
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.687908249794301e-06,
+      "loss": 0.4046,
+      "step": 54527
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6875724144879353e-06,
+      "loss": 0.3603,
+      "step": 54528
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6872365791815695e-06,
+      "loss": 0.3697,
+      "step": 54529
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.686900743875204e-06,
+      "loss": 0.1217,
+      "step": 54530
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.686564908568838e-06,
+      "loss": 0.2357,
+      "step": 54531
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6862290732624722e-06,
+      "loss": 0.389,
+      "step": 54532
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6858932379561066e-06,
+      "loss": 0.2764,
+      "step": 54533
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6855574026497406e-06,
+      "loss": 0.1075,
+      "step": 54534
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.685221567343375e-06,
+      "loss": 0.2461,
+      "step": 54535
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.684885732037009e-06,
+      "loss": 0.3652,
+      "step": 54536
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6845498967306435e-06,
+      "loss": 0.2063,
+      "step": 54537
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6842140614242777e-06,
+      "loss": 0.2492,
+      "step": 54538
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6838782261179117e-06,
+      "loss": 0.2787,
+      "step": 54539
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6835423908115462e-06,
+      "loss": 0.1895,
+      "step": 54540
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6832065555051804e-06,
+      "loss": 0.3333,
+      "step": 54541
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6828707201988146e-06,
+      "loss": 0.2139,
+      "step": 54542
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6825348848924488e-06,
+      "loss": 0.1594,
+      "step": 54543
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6821990495860833e-06,
+      "loss": 0.2956,
+      "step": 54544
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6818632142797173e-06,
+      "loss": 0.511,
+      "step": 54545
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6815273789733515e-06,
+      "loss": 0.3763,
+      "step": 54546
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.681191543666986e-06,
+      "loss": 0.2989,
+      "step": 54547
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.68085570836062e-06,
+      "loss": 0.2565,
+      "step": 54548
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6805198730542544e-06,
+      "loss": 0.2813,
+      "step": 54549
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6801840377478886e-06,
+      "loss": 0.2817,
+      "step": 54550
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6798482024415228e-06,
+      "loss": 0.1691,
+      "step": 54551
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.679512367135157e-06,
+      "loss": 0.3444,
+      "step": 54552
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6791765318287915e-06,
+      "loss": 0.2131,
+      "step": 54553
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6788406965224255e-06,
+      "loss": 0.1913,
+      "step": 54554
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6785048612160597e-06,
+      "loss": 0.2621,
+      "step": 54555
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6781690259096942e-06,
+      "loss": 0.2845,
+      "step": 54556
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6778331906033282e-06,
+      "loss": 0.1804,
+      "step": 54557
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6774973552969626e-06,
+      "loss": 0.2367,
+      "step": 54558
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6771615199905966e-06,
+      "loss": 0.2712,
+      "step": 54559
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6768256846842309e-06,
+      "loss": 0.1261,
+      "step": 54560
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6764898493778653e-06,
+      "loss": 0.3291,
+      "step": 54561
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6761540140714993e-06,
+      "loss": 0.2458,
+      "step": 54562
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6758181787651337e-06,
+      "loss": 0.3577,
+      "step": 54563
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.675482343458768e-06,
+      "loss": 0.2657,
+      "step": 54564
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6751465081524022e-06,
+      "loss": 0.3345,
+      "step": 54565
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6748106728460364e-06,
+      "loss": 0.204,
+      "step": 54566
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6744748375396709e-06,
+      "loss": 0.478,
+      "step": 54567
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6741390022333049e-06,
+      "loss": 0.4374,
+      "step": 54568
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.673803166926939e-06,
+      "loss": 0.2096,
+      "step": 54569
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6734673316205735e-06,
+      "loss": 0.243,
+      "step": 54570
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6731314963142075e-06,
+      "loss": 0.2875,
+      "step": 54571
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.672795661007842e-06,
+      "loss": 0.2461,
+      "step": 54572
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6724598257014762e-06,
+      "loss": 0.2082,
+      "step": 54573
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6721239903951102e-06,
+      "loss": 0.2518,
+      "step": 54574
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6717881550887446e-06,
+      "loss": 0.2407,
+      "step": 54575
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.671452319782379e-06,
+      "loss": 0.4222,
+      "step": 54576
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.671116484476013e-06,
+      "loss": 0.3307,
+      "step": 54577
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6707806491696473e-06,
+      "loss": 0.4312,
+      "step": 54578
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6704448138632815e-06,
+      "loss": 0.3819,
+      "step": 54579
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6701089785569158e-06,
+      "loss": 0.3119,
+      "step": 54580
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6697731432505502e-06,
+      "loss": 0.4976,
+      "step": 54581
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6694373079441842e-06,
+      "loss": 0.2466,
+      "step": 54582
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6691014726378184e-06,
+      "loss": 0.2111,
+      "step": 54583
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6687656373314529e-06,
+      "loss": 0.2401,
+      "step": 54584
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6684298020250869e-06,
+      "loss": 0.3802,
+      "step": 54585
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6680939667187213e-06,
+      "loss": 0.2216,
+      "step": 54586
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6677581314123555e-06,
+      "loss": 0.2189,
+      "step": 54587
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6674222961059896e-06,
+      "loss": 0.1778,
+      "step": 54588
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.667086460799624e-06,
+      "loss": 0.2306,
+      "step": 54589
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6667506254932584e-06,
+      "loss": 0.318,
+      "step": 54590
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6664147901868924e-06,
+      "loss": 0.2637,
+      "step": 54591
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6660789548805267e-06,
+      "loss": 0.3015,
+      "step": 54592
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.665743119574161e-06,
+      "loss": 0.2246,
+      "step": 54593
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6654072842677951e-06,
+      "loss": 0.4006,
+      "step": 54594
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6650714489614295e-06,
+      "loss": 0.2833,
+      "step": 54595
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6647356136550638e-06,
+      "loss": 0.1091,
+      "step": 54596
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6643997783486978e-06,
+      "loss": 0.2471,
+      "step": 54597
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6640639430423322e-06,
+      "loss": 0.1836,
+      "step": 54598
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6637281077359667e-06,
+      "loss": 0.3259,
+      "step": 54599
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6633922724296007e-06,
+      "loss": 0.2512,
+      "step": 54600
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6630564371232349e-06,
+      "loss": 0.2268,
+      "step": 54601
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6627206018168691e-06,
+      "loss": 0.1224,
+      "step": 54602
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6623847665105033e-06,
+      "loss": 0.3768,
+      "step": 54603
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6620489312041378e-06,
+      "loss": 0.2049,
+      "step": 54604
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6617130958977718e-06,
+      "loss": 0.148,
+      "step": 54605
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.661377260591406e-06,
+      "loss": 0.3047,
+      "step": 54606
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6610414252850404e-06,
+      "loss": 0.1751,
+      "step": 54607
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6607055899786745e-06,
+      "loss": 0.3009,
+      "step": 54608
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6603697546723089e-06,
+      "loss": 0.2817,
+      "step": 54609
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6600339193659431e-06,
+      "loss": 0.1629,
+      "step": 54610
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6596980840595771e-06,
+      "loss": 0.2115,
+      "step": 54611
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6593622487532116e-06,
+      "loss": 0.3723,
+      "step": 54612
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.659026413446846e-06,
+      "loss": 0.2539,
+      "step": 54613
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.65869057814048e-06,
+      "loss": 0.2097,
+      "step": 54614
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6583547428341142e-06,
+      "loss": 0.3217,
+      "step": 54615
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6580189075277487e-06,
+      "loss": 0.2683,
+      "step": 54616
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6576830722213827e-06,
+      "loss": 0.327,
+      "step": 54617
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6573472369150171e-06,
+      "loss": 0.2105,
+      "step": 54618
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6570114016086513e-06,
+      "loss": 0.4469,
+      "step": 54619
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6566755663022854e-06,
+      "loss": 0.2238,
+      "step": 54620
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6563397309959198e-06,
+      "loss": 0.2187,
+      "step": 54621
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.656003895689554e-06,
+      "loss": 0.2042,
+      "step": 54622
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6556680603831882e-06,
+      "loss": 0.4124,
+      "step": 54623
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6553322250768225e-06,
+      "loss": 0.192,
+      "step": 54624
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6549963897704565e-06,
+      "loss": 0.2233,
+      "step": 54625
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.654660554464091e-06,
+      "loss": 0.0996,
+      "step": 54626
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6543247191577253e-06,
+      "loss": 0.2273,
+      "step": 54627
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6539888838513594e-06,
+      "loss": 0.2455,
+      "step": 54628
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6536530485449936e-06,
+      "loss": 0.2,
+      "step": 54629
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.653317213238628e-06,
+      "loss": 0.2632,
+      "step": 54630
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.652981377932262e-06,
+      "loss": 0.4863,
+      "step": 54631
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6526455426258965e-06,
+      "loss": 0.266,
+      "step": 54632
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6523097073195307e-06,
+      "loss": 0.2493,
+      "step": 54633
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6519738720131647e-06,
+      "loss": 0.1875,
+      "step": 54634
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6516380367067991e-06,
+      "loss": 0.2517,
+      "step": 54635
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6513022014004334e-06,
+      "loss": 0.2473,
+      "step": 54636
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6509663660940676e-06,
+      "loss": 0.3008,
+      "step": 54637
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6506305307877018e-06,
+      "loss": 0.2426,
+      "step": 54638
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6502946954813362e-06,
+      "loss": 0.311,
+      "step": 54639
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6499588601749703e-06,
+      "loss": 0.3985,
+      "step": 54640
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6496230248686047e-06,
+      "loss": 0.2255,
+      "step": 54641
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.649287189562239e-06,
+      "loss": 0.165,
+      "step": 54642
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.648951354255873e-06,
+      "loss": 0.1919,
+      "step": 54643
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6486155189495074e-06,
+      "loss": 0.1467,
+      "step": 54644
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6482796836431416e-06,
+      "loss": 0.3077,
+      "step": 54645
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6479438483367758e-06,
+      "loss": 0.198,
+      "step": 54646
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.64760801303041e-06,
+      "loss": 0.2223,
+      "step": 54647
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.647272177724044e-06,
+      "loss": 0.2506,
+      "step": 54648
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6469363424176785e-06,
+      "loss": 0.1912,
+      "step": 54649
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6466005071113127e-06,
+      "loss": 0.2291,
+      "step": 54650
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.646264671804947e-06,
+      "loss": 0.2988,
+      "step": 54651
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6459288364985812e-06,
+      "loss": 0.2657,
+      "step": 54652
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6455930011922156e-06,
+      "loss": 0.4046,
+      "step": 54653
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6452571658858496e-06,
+      "loss": 0.2481,
+      "step": 54654
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.644921330579484e-06,
+      "loss": 0.1716,
+      "step": 54655
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6445854952731183e-06,
+      "loss": 0.3705,
+      "step": 54656
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6442496599667523e-06,
+      "loss": 0.4446,
+      "step": 54657
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6439138246603867e-06,
+      "loss": 0.335,
+      "step": 54658
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.643577989354021e-06,
+      "loss": 0.143,
+      "step": 54659
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6432421540476552e-06,
+      "loss": 0.2508,
+      "step": 54660
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6429063187412894e-06,
+      "loss": 0.3677,
+      "step": 54661
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6425704834349238e-06,
+      "loss": 0.2223,
+      "step": 54662
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6422346481285578e-06,
+      "loss": 0.1454,
+      "step": 54663
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6418988128221923e-06,
+      "loss": 0.3131,
+      "step": 54664
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6415629775158265e-06,
+      "loss": 0.1536,
+      "step": 54665
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6412271422094605e-06,
+      "loss": 0.1654,
+      "step": 54666
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.640891306903095e-06,
+      "loss": 0.198,
+      "step": 54667
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6405554715967292e-06,
+      "loss": 0.2747,
+      "step": 54668
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6402196362903634e-06,
+      "loss": 0.296,
+      "step": 54669
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6398838009839976e-06,
+      "loss": 0.1903,
+      "step": 54670
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6395479656776316e-06,
+      "loss": 0.2688,
+      "step": 54671
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.639212130371266e-06,
+      "loss": 0.1502,
+      "step": 54672
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6388762950649003e-06,
+      "loss": 0.2704,
+      "step": 54673
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6385404597585345e-06,
+      "loss": 0.1878,
+      "step": 54674
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6382046244521687e-06,
+      "loss": 0.2673,
+      "step": 54675
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6378687891458032e-06,
+      "loss": 0.3262,
+      "step": 54676
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6375329538394372e-06,
+      "loss": 0.4067,
+      "step": 54677
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6371971185330716e-06,
+      "loss": 0.3343,
+      "step": 54678
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6368612832267058e-06,
+      "loss": 0.1895,
+      "step": 54679
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6365254479203398e-06,
+      "loss": 0.1919,
+      "step": 54680
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6361896126139743e-06,
+      "loss": 0.169,
+      "step": 54681
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6358537773076085e-06,
+      "loss": 0.1511,
+      "step": 54682
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6355179420012427e-06,
+      "loss": 0.2192,
+      "step": 54683
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.635182106694877e-06,
+      "loss": 0.2279,
+      "step": 54684
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6348462713885114e-06,
+      "loss": 0.1685,
+      "step": 54685
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6345104360821454e-06,
+      "loss": 0.1698,
+      "step": 54686
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.6341746007757796e-06,
+      "loss": 0.2569,
+      "step": 54687
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.633838765469414e-06,
+      "loss": 0.2118,
+      "step": 54688
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.633502930163048e-06,
+      "loss": 0.3317,
+      "step": 54689
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6331670948566825e-06,
+      "loss": 0.3108,
+      "step": 54690
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6328312595503167e-06,
+      "loss": 0.1775,
+      "step": 54691
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.632495424243951e-06,
+      "loss": 0.3202,
+      "step": 54692
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6321595889375852e-06,
+      "loss": 0.2588,
+      "step": 54693
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6318237536312192e-06,
+      "loss": 0.1679,
+      "step": 54694
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6314879183248536e-06,
+      "loss": 0.24,
+      "step": 54695
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6311520830184878e-06,
+      "loss": 0.3944,
+      "step": 54696
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.630816247712122e-06,
+      "loss": 0.2066,
+      "step": 54697
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6304804124057563e-06,
+      "loss": 0.2223,
+      "step": 54698
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6301445770993907e-06,
+      "loss": 0.2457,
+      "step": 54699
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6298087417930247e-06,
+      "loss": 0.1966,
+      "step": 54700
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.629472906486659e-06,
+      "loss": 0.2022,
+      "step": 54701
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6291370711802934e-06,
+      "loss": 0.1959,
+      "step": 54702
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6288012358739274e-06,
+      "loss": 0.2918,
+      "step": 54703
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6284654005675619e-06,
+      "loss": 0.1587,
+      "step": 54704
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.628129565261196e-06,
+      "loss": 0.3269,
+      "step": 54705
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6277937299548303e-06,
+      "loss": 0.2397,
+      "step": 54706
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6274578946484645e-06,
+      "loss": 0.2393,
+      "step": 54707
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.627122059342099e-06,
+      "loss": 0.3784,
+      "step": 54708
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.626786224035733e-06,
+      "loss": 0.22,
+      "step": 54709
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6264503887293672e-06,
+      "loss": 0.2598,
+      "step": 54710
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6261145534230016e-06,
+      "loss": 0.1756,
+      "step": 54711
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6257787181166356e-06,
+      "loss": 0.0979,
+      "step": 54712
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.62544288281027e-06,
+      "loss": 0.2211,
+      "step": 54713
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6251070475039043e-06,
+      "loss": 0.1991,
+      "step": 54714
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6247712121975383e-06,
+      "loss": 0.2316,
+      "step": 54715
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6244353768911727e-06,
+      "loss": 0.2222,
+      "step": 54716
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6240995415848068e-06,
+      "loss": 0.2094,
+      "step": 54717
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6237637062784412e-06,
+      "loss": 0.2951,
+      "step": 54718
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6234278709720754e-06,
+      "loss": 0.3531,
+      "step": 54719
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6230920356657096e-06,
+      "loss": 0.1563,
+      "step": 54720
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6227562003593439e-06,
+      "loss": 0.2804,
+      "step": 54721
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6224203650529783e-06,
+      "loss": 0.2967,
+      "step": 54722
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6220845297466123e-06,
+      "loss": 0.3443,
+      "step": 54723
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6217486944402465e-06,
+      "loss": 0.2161,
+      "step": 54724
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.621412859133881e-06,
+      "loss": 0.2773,
+      "step": 54725
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.621077023827515e-06,
+      "loss": 0.3336,
+      "step": 54726
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6207411885211494e-06,
+      "loss": 0.2188,
+      "step": 54727
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6204053532147836e-06,
+      "loss": 0.333,
+      "step": 54728
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6200695179084179e-06,
+      "loss": 0.2075,
+      "step": 54729
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.619733682602052e-06,
+      "loss": 0.4737,
+      "step": 54730
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6193978472956865e-06,
+      "loss": 0.1134,
+      "step": 54731
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6190620119893205e-06,
+      "loss": 0.4329,
+      "step": 54732
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6187261766829548e-06,
+      "loss": 0.2174,
+      "step": 54733
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6183903413765892e-06,
+      "loss": 0.2706,
+      "step": 54734
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6180545060702232e-06,
+      "loss": 0.2659,
+      "step": 54735
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6177186707638576e-06,
+      "loss": 0.2854,
+      "step": 54736
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6173828354574917e-06,
+      "loss": 0.1828,
+      "step": 54737
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6170470001511259e-06,
+      "loss": 0.3203,
+      "step": 54738
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6167111648447603e-06,
+      "loss": 0.2148,
+      "step": 54739
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6163753295383943e-06,
+      "loss": 0.3481,
+      "step": 54740
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6160394942320288e-06,
+      "loss": 0.2657,
+      "step": 54741
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.615703658925663e-06,
+      "loss": 0.3096,
+      "step": 54742
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6153678236192972e-06,
+      "loss": 0.2355,
+      "step": 54743
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6150319883129314e-06,
+      "loss": 0.3485,
+      "step": 54744
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6146961530065659e-06,
+      "loss": 0.2339,
+      "step": 54745
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6143603177001999e-06,
+      "loss": 0.2283,
+      "step": 54746
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6140244823938341e-06,
+      "loss": 0.1899,
+      "step": 54747
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6136886470874685e-06,
+      "loss": 0.358,
+      "step": 54748
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6133528117811026e-06,
+      "loss": 0.1702,
+      "step": 54749
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.613016976474737e-06,
+      "loss": 0.3473,
+      "step": 54750
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6126811411683712e-06,
+      "loss": 0.2621,
+      "step": 54751
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6123453058620052e-06,
+      "loss": 0.3188,
+      "step": 54752
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6120094705556397e-06,
+      "loss": 0.3266,
+      "step": 54753
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.611673635249274e-06,
+      "loss": 0.2821,
+      "step": 54754
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6113377999429081e-06,
+      "loss": 0.3653,
+      "step": 54755
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6110019646365423e-06,
+      "loss": 0.2094,
+      "step": 54756
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6106661293301768e-06,
+      "loss": 0.3122,
+      "step": 54757
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6103302940238108e-06,
+      "loss": 0.175,
+      "step": 54758
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6099944587174452e-06,
+      "loss": 0.3535,
+      "step": 54759
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6096586234110792e-06,
+      "loss": 0.1749,
+      "step": 54760
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6093227881047135e-06,
+      "loss": 0.2378,
+      "step": 54761
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6089869527983479e-06,
+      "loss": 0.2854,
+      "step": 54762
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.608651117491982e-06,
+      "loss": 0.2944,
+      "step": 54763
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6083152821856163e-06,
+      "loss": 0.2155,
+      "step": 54764
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6079794468792506e-06,
+      "loss": 0.2562,
+      "step": 54765
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6076436115728846e-06,
+      "loss": 0.2442,
+      "step": 54766
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.607307776266519e-06,
+      "loss": 0.1596,
+      "step": 54767
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6069719409601534e-06,
+      "loss": 0.1812,
+      "step": 54768
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6066361056537875e-06,
+      "loss": 0.3273,
+      "step": 54769
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6063002703474217e-06,
+      "loss": 0.253,
+      "step": 54770
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6059644350410561e-06,
+      "loss": 0.2626,
+      "step": 54771
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6056285997346901e-06,
+      "loss": 0.1446,
+      "step": 54772
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6052927644283246e-06,
+      "loss": 0.1603,
+      "step": 54773
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6049569291219588e-06,
+      "loss": 0.1862,
+      "step": 54774
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6046210938155928e-06,
+      "loss": 0.1607,
+      "step": 54775
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6042852585092272e-06,
+      "loss": 0.1608,
+      "step": 54776
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6039494232028615e-06,
+      "loss": 0.3039,
+      "step": 54777
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6036135878964957e-06,
+      "loss": 0.2627,
+      "step": 54778
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.60327775259013e-06,
+      "loss": 0.2731,
+      "step": 54779
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6029419172837643e-06,
+      "loss": 0.1682,
+      "step": 54780
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6026060819773984e-06,
+      "loss": 0.4054,
+      "step": 54781
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6022702466710328e-06,
+      "loss": 0.2338,
+      "step": 54782
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6019344113646668e-06,
+      "loss": 0.2899,
+      "step": 54783
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.601598576058301e-06,
+      "loss": 0.2839,
+      "step": 54784
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6012627407519355e-06,
+      "loss": 0.3502,
+      "step": 54785
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6009269054455695e-06,
+      "loss": 0.343,
+      "step": 54786
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.600591070139204e-06,
+      "loss": 0.0809,
+      "step": 54787
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.6002552348328381e-06,
+      "loss": 0.1839,
+      "step": 54788
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5999193995264721e-06,
+      "loss": 0.2137,
+      "step": 54789
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5995835642201066e-06,
+      "loss": 0.39,
+      "step": 54790
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.599247728913741e-06,
+      "loss": 0.555,
+      "step": 54791
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.598911893607375e-06,
+      "loss": 0.1843,
+      "step": 54792
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5985760583010093e-06,
+      "loss": 0.3025,
+      "step": 54793
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5982402229946437e-06,
+      "loss": 0.2296,
+      "step": 54794
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5979043876882777e-06,
+      "loss": 0.3521,
+      "step": 54795
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5975685523819121e-06,
+      "loss": 0.2903,
+      "step": 54796
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5972327170755464e-06,
+      "loss": 0.3374,
+      "step": 54797
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5968968817691804e-06,
+      "loss": 0.2309,
+      "step": 54798
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5965610464628148e-06,
+      "loss": 0.3314,
+      "step": 54799
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.596225211156449e-06,
+      "loss": 0.3114,
+      "step": 54800
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5958893758500833e-06,
+      "loss": 0.2278,
+      "step": 54801
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5955535405437175e-06,
+      "loss": 0.2661,
+      "step": 54802
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.595217705237352e-06,
+      "loss": 0.2874,
+      "step": 54803
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.594881869930986e-06,
+      "loss": 0.3093,
+      "step": 54804
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5945460346246204e-06,
+      "loss": 0.2908,
+      "step": 54805
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5942101993182544e-06,
+      "loss": 0.3511,
+      "step": 54806
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5938743640118886e-06,
+      "loss": 0.3257,
+      "step": 54807
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.593538528705523e-06,
+      "loss": 0.194,
+      "step": 54808
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.593202693399157e-06,
+      "loss": 0.3086,
+      "step": 54809
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5928668580927915e-06,
+      "loss": 0.2031,
+      "step": 54810
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5925310227864257e-06,
+      "loss": 0.227,
+      "step": 54811
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5921951874800597e-06,
+      "loss": 0.1933,
+      "step": 54812
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5918593521736942e-06,
+      "loss": 0.3109,
+      "step": 54813
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5915235168673284e-06,
+      "loss": 0.1011,
+      "step": 54814
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5911876815609626e-06,
+      "loss": 0.2887,
+      "step": 54815
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5908518462545968e-06,
+      "loss": 0.3295,
+      "step": 54816
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5905160109482313e-06,
+      "loss": 0.2675,
+      "step": 54817
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5901801756418653e-06,
+      "loss": 0.1511,
+      "step": 54818
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5898443403354997e-06,
+      "loss": 0.2639,
+      "step": 54819
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.589508505029134e-06,
+      "loss": 0.1945,
+      "step": 54820
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.589172669722768e-06,
+      "loss": 0.5341,
+      "step": 54821
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5888368344164024e-06,
+      "loss": 0.2302,
+      "step": 54822
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5885009991100366e-06,
+      "loss": 0.2395,
+      "step": 54823
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5881651638036708e-06,
+      "loss": 0.2027,
+      "step": 54824
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.587829328497305e-06,
+      "loss": 0.4145,
+      "step": 54825
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5874934931909395e-06,
+      "loss": 0.4851,
+      "step": 54826
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5871576578845735e-06,
+      "loss": 0.2953,
+      "step": 54827
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5868218225782077e-06,
+      "loss": 0.2851,
+      "step": 54828
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.586485987271842e-06,
+      "loss": 0.2394,
+      "step": 54829
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5861501519654762e-06,
+      "loss": 0.3576,
+      "step": 54830
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5858143166591106e-06,
+      "loss": 0.2085,
+      "step": 54831
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5854784813527446e-06,
+      "loss": 0.2845,
+      "step": 54832
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.585142646046379e-06,
+      "loss": 0.2555,
+      "step": 54833
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5848068107400133e-06,
+      "loss": 0.21,
+      "step": 54834
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5844709754336473e-06,
+      "loss": 0.2963,
+      "step": 54835
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5841351401272817e-06,
+      "loss": 0.4208,
+      "step": 54836
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.583799304820916e-06,
+      "loss": 0.2809,
+      "step": 54837
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5834634695145502e-06,
+      "loss": 0.33,
+      "step": 54838
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5831276342081844e-06,
+      "loss": 0.2701,
+      "step": 54839
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5827917989018188e-06,
+      "loss": 0.1532,
+      "step": 54840
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5824559635954528e-06,
+      "loss": 0.2548,
+      "step": 54841
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.582120128289087e-06,
+      "loss": 0.3586,
+      "step": 54842
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5817842929827215e-06,
+      "loss": 0.1439,
+      "step": 54843
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5814484576763555e-06,
+      "loss": 0.2589,
+      "step": 54844
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.58111262236999e-06,
+      "loss": 0.2643,
+      "step": 54845
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5807767870636242e-06,
+      "loss": 0.3746,
+      "step": 54846
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5804409517572584e-06,
+      "loss": 0.4022,
+      "step": 54847
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5801051164508926e-06,
+      "loss": 0.2152,
+      "step": 54848
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.579769281144527e-06,
+      "loss": 0.4273,
+      "step": 54849
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.579433445838161e-06,
+      "loss": 0.2104,
+      "step": 54850
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5790976105317953e-06,
+      "loss": 0.286,
+      "step": 54851
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5787617752254295e-06,
+      "loss": 0.1868,
+      "step": 54852
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5784259399190637e-06,
+      "loss": 0.2032,
+      "step": 54853
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5780901046126982e-06,
+      "loss": 0.2761,
+      "step": 54854
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5777542693063322e-06,
+      "loss": 0.2867,
+      "step": 54855
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5774184339999664e-06,
+      "loss": 0.1563,
+      "step": 54856
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5770825986936009e-06,
+      "loss": 0.2359,
+      "step": 54857
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5767467633872349e-06,
+      "loss": 0.2791,
+      "step": 54858
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5764109280808693e-06,
+      "loss": 0.2632,
+      "step": 54859
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5760750927745035e-06,
+      "loss": 0.2287,
+      "step": 54860
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5757392574681378e-06,
+      "loss": 0.2572,
+      "step": 54861
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.575403422161772e-06,
+      "loss": 0.1913,
+      "step": 54862
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5750675868554064e-06,
+      "loss": 0.3272,
+      "step": 54863
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5747317515490404e-06,
+      "loss": 0.1531,
+      "step": 54864
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5743959162426746e-06,
+      "loss": 0.1508,
+      "step": 54865
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.574060080936309e-06,
+      "loss": 0.3201,
+      "step": 54866
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.573724245629943e-06,
+      "loss": 0.2423,
+      "step": 54867
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5733884103235775e-06,
+      "loss": 0.3288,
+      "step": 54868
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5730525750172118e-06,
+      "loss": 0.2049,
+      "step": 54869
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.572716739710846e-06,
+      "loss": 0.2667,
+      "step": 54870
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5723809044044802e-06,
+      "loss": 0.1678,
+      "step": 54871
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5720450690981146e-06,
+      "loss": 0.1924,
+      "step": 54872
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5717092337917486e-06,
+      "loss": 0.2739,
+      "step": 54873
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5713733984853829e-06,
+      "loss": 0.2607,
+      "step": 54874
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.571037563179017e-06,
+      "loss": 0.2983,
+      "step": 54875
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5707017278726513e-06,
+      "loss": 0.3005,
+      "step": 54876
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5703658925662858e-06,
+      "loss": 0.2556,
+      "step": 54877
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5700300572599198e-06,
+      "loss": 0.365,
+      "step": 54878
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.569694221953554e-06,
+      "loss": 0.2915,
+      "step": 54879
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5693583866471884e-06,
+      "loss": 0.2876,
+      "step": 54880
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5690225513408224e-06,
+      "loss": 0.2005,
+      "step": 54881
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5686867160344569e-06,
+      "loss": 0.1899,
+      "step": 54882
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.568350880728091e-06,
+      "loss": 0.2055,
+      "step": 54883
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5680150454217253e-06,
+      "loss": 0.2016,
+      "step": 54884
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5676792101153595e-06,
+      "loss": 0.3215,
+      "step": 54885
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.567343374808994e-06,
+      "loss": 0.3146,
+      "step": 54886
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.567007539502628e-06,
+      "loss": 0.248,
+      "step": 54887
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 1.5666717041962622e-06,
+      "loss": 0.2908,
+      "step": 54888
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5663358688898967e-06,
+      "loss": 0.2245,
+      "step": 54889
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5660000335835307e-06,
+      "loss": 0.1778,
+      "step": 54890
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.565664198277165e-06,
+      "loss": 0.364,
+      "step": 54891
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5653283629707993e-06,
+      "loss": 0.1729,
+      "step": 54892
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5649925276644333e-06,
+      "loss": 0.3432,
+      "step": 54893
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5646566923580678e-06,
+      "loss": 0.3948,
+      "step": 54894
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5643208570517018e-06,
+      "loss": 0.2447,
+      "step": 54895
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5639850217453362e-06,
+      "loss": 0.3288,
+      "step": 54896
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5636491864389704e-06,
+      "loss": 0.1094,
+      "step": 54897
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5633133511326047e-06,
+      "loss": 0.1961,
+      "step": 54898
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5629775158262389e-06,
+      "loss": 0.2758,
+      "step": 54899
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5626416805198733e-06,
+      "loss": 0.1571,
+      "step": 54900
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5623058452135073e-06,
+      "loss": 0.2075,
+      "step": 54901
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5619700099071416e-06,
+      "loss": 0.2759,
+      "step": 54902
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.561634174600776e-06,
+      "loss": 0.2895,
+      "step": 54903
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.56129833929441e-06,
+      "loss": 0.3646,
+      "step": 54904
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5609625039880444e-06,
+      "loss": 0.4426,
+      "step": 54905
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5606266686816787e-06,
+      "loss": 0.4354,
+      "step": 54906
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5602908333753127e-06,
+      "loss": 0.3105,
+      "step": 54907
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5599549980689471e-06,
+      "loss": 0.1133,
+      "step": 54908
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5596191627625816e-06,
+      "loss": 0.1976,
+      "step": 54909
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5592833274562156e-06,
+      "loss": 0.3438,
+      "step": 54910
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5589474921498498e-06,
+      "loss": 0.2253,
+      "step": 54911
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5586116568434842e-06,
+      "loss": 0.1788,
+      "step": 54912
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5582758215371182e-06,
+      "loss": 0.3221,
+      "step": 54913
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5579399862307527e-06,
+      "loss": 0.2745,
+      "step": 54914
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.557604150924387e-06,
+      "loss": 0.2488,
+      "step": 54915
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.557268315618021e-06,
+      "loss": 0.3476,
+      "step": 54916
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5569324803116553e-06,
+      "loss": 0.269,
+      "step": 54917
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5565966450052894e-06,
+      "loss": 0.2101,
+      "step": 54918
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5562608096989238e-06,
+      "loss": 0.2924,
+      "step": 54919
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.555924974392558e-06,
+      "loss": 0.2603,
+      "step": 54920
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.555589139086192e-06,
+      "loss": 0.2501,
+      "step": 54921
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5552533037798265e-06,
+      "loss": 0.2377,
+      "step": 54922
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.554917468473461e-06,
+      "loss": 0.1792,
+      "step": 54923
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.554581633167095e-06,
+      "loss": 0.2424,
+      "step": 54924
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5542457978607291e-06,
+      "loss": 0.4524,
+      "step": 54925
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5539099625543636e-06,
+      "loss": 0.2574,
+      "step": 54926
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5535741272479976e-06,
+      "loss": 0.1734,
+      "step": 54927
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.553238291941632e-06,
+      "loss": 0.2262,
+      "step": 54928
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5529024566352662e-06,
+      "loss": 0.2036,
+      "step": 54929
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5525666213289003e-06,
+      "loss": 0.1951,
+      "step": 54930
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5522307860225347e-06,
+      "loss": 0.2276,
+      "step": 54931
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5518949507161691e-06,
+      "loss": 0.3223,
+      "step": 54932
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5515591154098031e-06,
+      "loss": 0.361,
+      "step": 54933
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5512232801034374e-06,
+      "loss": 0.3116,
+      "step": 54934
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5508874447970718e-06,
+      "loss": 0.3898,
+      "step": 54935
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5505516094907058e-06,
+      "loss": 0.2191,
+      "step": 54936
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5502157741843402e-06,
+      "loss": 0.301,
+      "step": 54937
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5498799388779745e-06,
+      "loss": 0.2748,
+      "step": 54938
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5495441035716085e-06,
+      "loss": 0.3147,
+      "step": 54939
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.549208268265243e-06,
+      "loss": 0.1867,
+      "step": 54940
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.548872432958877e-06,
+      "loss": 0.1816,
+      "step": 54941
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5485365976525114e-06,
+      "loss": 0.2684,
+      "step": 54942
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5482007623461456e-06,
+      "loss": 0.4642,
+      "step": 54943
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5478649270397796e-06,
+      "loss": 0.2096,
+      "step": 54944
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.547529091733414e-06,
+      "loss": 0.2997,
+      "step": 54945
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5471932564270485e-06,
+      "loss": 0.4096,
+      "step": 54946
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5468574211206825e-06,
+      "loss": 0.5389,
+      "step": 54947
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5465215858143167e-06,
+      "loss": 0.3037,
+      "step": 54948
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5461857505079511e-06,
+      "loss": 0.2059,
+      "step": 54949
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5458499152015852e-06,
+      "loss": 0.1153,
+      "step": 54950
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5455140798952196e-06,
+      "loss": 0.2653,
+      "step": 54951
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5451782445888538e-06,
+      "loss": 0.3222,
+      "step": 54952
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5448424092824878e-06,
+      "loss": 0.2474,
+      "step": 54953
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5445065739761223e-06,
+      "loss": 0.1828,
+      "step": 54954
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5441707386697565e-06,
+      "loss": 0.2831,
+      "step": 54955
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5438349033633907e-06,
+      "loss": 0.3609,
+      "step": 54956
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.543499068057025e-06,
+      "loss": 0.3079,
+      "step": 54957
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5431632327506594e-06,
+      "loss": 0.3613,
+      "step": 54958
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5428273974442934e-06,
+      "loss": 0.3834,
+      "step": 54959
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5424915621379278e-06,
+      "loss": 0.3355,
+      "step": 54960
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.542155726831562e-06,
+      "loss": 0.2256,
+      "step": 54961
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.541819891525196e-06,
+      "loss": 0.3978,
+      "step": 54962
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5414840562188305e-06,
+      "loss": 0.4615,
+      "step": 54963
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5411482209124645e-06,
+      "loss": 0.2683,
+      "step": 54964
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.540812385606099e-06,
+      "loss": 0.1303,
+      "step": 54965
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5404765502997332e-06,
+      "loss": 0.3269,
+      "step": 54966
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5401407149933672e-06,
+      "loss": 0.3582,
+      "step": 54967
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5398048796870016e-06,
+      "loss": 0.1238,
+      "step": 54968
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5394690443806358e-06,
+      "loss": 0.2249,
+      "step": 54969
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.53913320907427e-06,
+      "loss": 0.294,
+      "step": 54970
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5387973737679043e-06,
+      "loss": 0.2943,
+      "step": 54971
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5384615384615387e-06,
+      "loss": 0.4168,
+      "step": 54972
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5381257031551727e-06,
+      "loss": 0.3944,
+      "step": 54973
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5377898678488072e-06,
+      "loss": 0.2295,
+      "step": 54974
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5374540325424414e-06,
+      "loss": 0.3807,
+      "step": 54975
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5371181972360754e-06,
+      "loss": 0.3163,
+      "step": 54976
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5367823619297098e-06,
+      "loss": 0.3391,
+      "step": 54977
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.536446526623344e-06,
+      "loss": 0.3175,
+      "step": 54978
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5361106913169783e-06,
+      "loss": 0.2986,
+      "step": 54979
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5357748560106125e-06,
+      "loss": 0.4415,
+      "step": 54980
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.535439020704247e-06,
+      "loss": 0.4738,
+      "step": 54981
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.535103185397881e-06,
+      "loss": 0.3071,
+      "step": 54982
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5347673500915152e-06,
+      "loss": 0.4403,
+      "step": 54983
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5344315147851496e-06,
+      "loss": 0.3266,
+      "step": 54984
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5340956794787836e-06,
+      "loss": 0.3367,
+      "step": 54985
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.533759844172418e-06,
+      "loss": 0.4686,
+      "step": 54986
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.533424008866052e-06,
+      "loss": 0.3124,
+      "step": 54987
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5330881735596865e-06,
+      "loss": 0.138,
+      "step": 54988
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5327523382533207e-06,
+      "loss": 0.438,
+      "step": 54989
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5324165029469547e-06,
+      "loss": 0.2981,
+      "step": 54990
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5320806676405892e-06,
+      "loss": 0.3589,
+      "step": 54991
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5317448323342234e-06,
+      "loss": 0.3002,
+      "step": 54992
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5314089970278576e-06,
+      "loss": 0.2203,
+      "step": 54993
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5310731617214919e-06,
+      "loss": 0.3627,
+      "step": 54994
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5307373264151263e-06,
+      "loss": 0.3345,
+      "step": 54995
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5304014911087603e-06,
+      "loss": 0.1629,
+      "step": 54996
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5300656558023947e-06,
+      "loss": 0.0752,
+      "step": 54997
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.529729820496029e-06,
+      "loss": 0.2268,
+      "step": 54998
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.529393985189663e-06,
+      "loss": 0.2752,
+      "step": 54999
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.5290581498832974e-06,
+      "loss": 0.3349,
+      "step": 55000
+    },
+    {
+      "epoch": 2.77,
+      "eval_accuracy": 0.8849976858484121,
+      "eval_f1_microaverage": 0.8849976858484121,
+      "eval_loss": 0.3068837523460388,
+      "eval_precision_macroaverage": 0.7689999171592804,
+      "eval_precision_microaverage": 0.8849976858484121,
+      "eval_recall_macroaverage": 0.6882016737904029,
+      "eval_recall_microaverage": 0.8849976858484121,
+      "eval_runtime": 1173.3779,
+      "eval_samples_per_second": 232.012,
+      "eval_steps_per_second": 3.625,
+      "step": 55000
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 59553,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 5000,
+  "total_flos": 9.232543646493112e+17,
+  "train_batch_size": 64,
+  "trial_name": null,
+  "trial_params": null
+}